Guest Post by Willis Eschenbach
I’m 74, and I’ve been programming computers nearly as long as anyone alive.
When I was 15, I’d been reading about computers in pulp science fiction magazines like Amazing Stories, Analog, and Galaxy for a while. I wanted one so badly. Why? I figured it could do my homework for me. Hey, I was 15, wad’ja expect?
I was always into math, it came easy to me. In 1963, the summer after my junior year in high school, nearly sixty years ago now, I was one of the kids selected from all over the US to participate in the National Science Foundation summer school in mathematics. It was held up in Corvallis, Oregon, at Oregon State University.
It was a wonderful time. I got to study math with a bunch of kids my age who were as excited as I was about math. Bizarrely, one of the other students turned out to be a second cousin of mine I’d never even heard of. Seems math runs in the family. My older brother is a genius mathematician, inventor of the first civilian version of the GPS. What a curious world.
The best news about the summer school was, in addition to the math classes, marvel of marvels, they taught us about computers … and they had a real live one that we could write programs for!
They started out by having us design and build logic circuits using wires, relays, the real-world stuff. They were for things like AND gates, OR gates, and flip-flops. Great fun!
Then they introduced us to Algol. Algol is a long-dead computer language, designed in 1958, but it was a standard for a long time. It was very similar to but an improvement on Fortran in that it used less memory.
Once we had learned something about Algol, they took us to see the computer. It was huge old CDC 3300, standing about as high as a person’s chest, taking up a good chunk of a small room. The back of it looked like this.
It had a memory composed of small ring-shaped magnets with wires running through them, like the photo below. The computer energized a combination of the wires to “flip” the magnetic state of each of the small rings. This allowed each small ring to represent a binary 1 or a 0.
How much memory did it have? A whacking great 768 kilobytes. Not gigabytes. Not megabytes. Kilobytes. Thats one ten-thousandth of the memory of the ten-year-old Mac I’m writing this on.
It was programmed using Hollerith punch cards. They didn’t let us anywhere near the actual computer, of course. We sat at the card punch machines and typed in our program. Here’s a punch card, 7 3/8 inches wide by 3 1/4 inches high by 0.007 inches thick. (187 x 83 x.018 mm).
The program would end up as a stack of cards with holes punched in them, usually 25-50 cards or so. I’d give my stack to the instructors, and a couple of days later I’d get a note saying “Problem on card 11”. So I’d rewrite card 11, resubmit them, and get a note saying “Problem on card 19” … debugging a program written on punch cards was a slooow process, I can assure you
And I loved it. It was amazing. My first program was the “Sieve of Eratosthenes“, and I was over the moon when it finally compiled and ran. I was well and truly hooked, and I never looked back.
The rest of that summer I worked as a bicycle messenger in San Francisco, riding a one-speed bike up and down the hills delivering blueprints. I gave all the money I made to our mom to help support the family. But I couldn’t get the computer out of my mind.
Ten years later, after graduating from high school and then dropping out of college after one year, I went back to college specifically so I could study computers. I enrolled in Laney College in Oakland. It was a great school, about 80% black, 10% Hispanic, and the rest a mixed bag of melanin-deficient folks. (I’m told than nowadays the polically-correct term is “melanin-challenged”, to avoid offending anyone.) The Laney College Computer Department had a Datapoint 2200 computer, the first desktop computer.
It had only 8 kilobytes of memory … but the advantage was that you could program it directly. The disadvantage was that only one student could work on it at any time. However, the computer teacher saw my love of the machine, so he gave me a key to the computer room so I could come in before or after hours and program to my heart’s content. I spent every spare hour there. It used a language called Databus, my second computer language.
The first program I wrote for this computer? You’ll laugh. It was a test to see if there was “precognition”. You know, seeing the future. My first version, I punched a key from 0 to 9. Then the computer picked a random number, and recorded if I was right or not.
Finding I didn’t have precognition, I re-wrote the program. In version 2, the computer picked the number before, rather than after, I made the selection. No precognition needed. Guess what?
No better than random chance. And sadly, that one-semester course was all that Laney College offered. That’s the extent of my formal computer education. The rest I taught myself, year after year, language after language, concept after concept, program after program.
Ten years after that, I bought the first computer I ever owned — the Radio Shack TRS-80, AKA the “Trash Eighty”. It was the first notebook-style computer. I took that sucker all over the world. I wrote endless programs on it, including marine celestial navigation programs that I used to navigate by the stars between islands the South Pacific. It was also my first introduction to Basic, my third computer language.
And by then IBM had released the IBM PC, the first personal computer. When I returned to the US I bought one. I learned my fourth computer language, CPM. I wrote all kinds of programs for it. But then a couple years later Apple came out with the Macintosh. I bought one of those as well, because of the mouse and the art and music programs. I figured I’d use the Mac for creating my art and my music and such, and the PC for serious work.
But after a year or so, I found I was using nothing but the Mac, and there was a quarter-inch of dust on my IBM PC. So I traded the PC for a piano, the very piano here in our house that I played last night for my 19-month-old granddaughter, and I never looked back at the IBM side of computing.
I taught myself C and C++ when I needed speed to run blackjack simulations … see, I’d learned to play professional blackjack along the way, counting cards. And when my player friends told me how much it cost for them to test their new betting and counting systems, I wrote a blackjack simulation program to test the new ideas. You need to run about a hundred thousand hands for a solid result. That took several days in Basic, but in C, I’d start the run at night, and when I got up the next morning, the run would be done. I charged $100 per test, and I thought “This is what I wanted a computer for … to make me a hundred bucks a night while I’m asleep.”
Since then, I’ve never been without a computer. I’ve written literally thousands and thousands of programs. On my current computer, a ten-year-old Macbook Pro, a quick check shows that there are well over 4,000 programs I’ve written. I’ve written programs in Algol, Datacom, 68000 Machine Language, Basic, C/C++, Hypertalk, Forth, Logo, Lisp, Mathematica (3 languages), Vectorscript, Pascal, VBA, Stella computer modeling language, and these days, R.
I had the immense good fortune to be directed to R by Steve McIntyre of ClimateAudit. It’s the best language I’ve ever used—free, cross-platform, fast, with a killer user interface and free “packages” to do just about anything you can name. If you do any serious programming, I can’t recommend it enough.
Oh, yeah, somewhere in there I spent a year as the Service Manager for an Apple Dealership. As you might guess given my checkered history, it wasn’t in some logical location … it was in downtown Suva, in Fiji. There I fixed a lot of computers and I learned immense patience dealing with good folks who truly thought that the CD tray that came out of the front of their computer when they did something by accident was a coffee cup holder … oh, and I also installed the Macintosh hardware for the Fiji Government Printers and trained the employees how to use Photoshop. I also taught two semesters of Computers 101 at the Fiji Institute of Technology.
I bring all of this up to let you know that I’m far, far from being a novice, a beginner, or even a journeyman programmer. I was working with “computer based evolution” to try to analyze the stock market before most folks even heard of it. I’m a master of the art, able to do things like write “hooks” into Excel that let Excel transparently call a separate program in C for its wicked-fast speed, and then return the answer to a cell in Excel …
Now, folks who’ve read my work know that I am far from enamored of computer climate models. I’ve been asked “What do you have against computer models?” and “How can you not trust models, we use them for everything?”
Well, based on a lifetime’s experience in the field, I can assure you of a few things about computer climate models and computer models in general. Here’s the short course.
• A computer model is nothing more than a physical realization of the beliefs, understandings, wrong ideas, and misunderstandings of whoever wrote the model. Therefore, the results it produces are going to support, bear out, and instantiate the programmer’s beliefs, understandings, wrong ideas, and misunderstandings. All that the computer does is make those under- and misunder-standings look official and reasonable. Oh, and make mistakes really, really fast. Been there, done that.
• Computer climate models are members of a particular class of models called “Iterative” computer models. In this class of models, the output of one timestep is fed back into the computer as the input of the next timestep. Members of his class of models are notoriously cranky, unstable, and prone to internal oscillations and generally falling off the perch. They usually need to be artificially “fenced in” in some sense to keep them from spiraling out of control.
• As anyone who has ever tried to model say the stock market can tell you, a model which can reproduce the past absolutely flawlessly may, and in fact very likely will, give totally incorrect predictions of the future. Been there, done that too. As the brokerage advertisements in the US are required to say, “Past performance is no guarantee of future success”.
• This means that the fact that a climate model can hindcast the past climate perfectly does NOT mean that it is an accurate representation of reality. And in particular, it does NOT mean it can accurately predict the future.
• Chaotic systems like weather and climate are notoriously difficult to model, even in the short term. That’s why projections of a cyclone’s future path over say the next 48 hours are in the shape of a cone and not a straight line.
• There is an entire branch of computer science called “V&V”, which stands for validation and verification. It’s how you can be assured that your software is up to the task it was designed for. Here’s a description from the web
What is software verification and validation (V&V)?
Verification
820.3(a) Verification means confirmation by examination and provision of objective evidence that specified requirements have been fulfilled.
“Documented procedures, performed in the user environment, for obtaining, recording, and interpreting the results required to establish that predetermined specifications have been met” (AAMI).
Validation
820.3(z) Validation means confirmation by examination and provision of objective evidence that the particular requirements for a specific intended use can be consistently fulfilled.
Process Validation means establishing by objective evidence that a process consistently produces a result or product meeting its predetermined specifications.
Design Validation means establishing by objective evidence that device specifications conform with user needs and intended use(s).
“Documented procedure for obtaining, recording, and interpreting the results required to establish that a process will consistently yield product complying with predetermined specifications” (AAMI).
Further V&V information here.
• Your average elevator control software has been subjected to more V&V than the computer climate models. And unless a computer model’s software has been subjected to extensive and rigorous V&V. the fact that the model says that something happens in modelworld is NOT evidence that it actually happens in the real world … and even then, as they say, “Excrement occurs”. We lost a Mars probe because someone didn’t convert a single number to metric from Imperial measurements … and you can bet that JPL subjects their programs to extensive and rigorous V&V.
• Computer modelers, myself included at times, are all subject to a nearly irresistible desire to mistake Modelworld for the real world. They say things like “We’ve determined that climate phenomenon X is caused by forcing Y”. But a true statement would be “We’ve determined that in our model, the modeled climate phenomenon X is caused by our modeled forcing Y”. Unfortunately, the modelers are not the only ones fooled in this process.
• The more tunable parameters a model has, the less likely it is to accurately represent reality. Climate models have dozens of tunable parameters. Here are 25 of them, there are plenty more.
What’s wrong with parameters in a model? Here’s an oft-repeated story about the famous physicist Freeman Dyson getting schooled on the subject by the even more famous Enrico Fermi …
By the spring of 1953, after heroic efforts, we had plotted theoretical graphs of meson–proton scattering. We joyfully observed that our calculated numbers agreed pretty well with Fermi’s measured numbers. So I made an appointment to meet with Fermi and show him our results. Proudly, I rode the Greyhound bus from Ithaca to Chicago with package of our theoretical graphs to show to Fermi.
When I arrived in Fermi’s office, I handed the graphs to Fermi, but he hardly glanced at them. He invited me to sit down, and asked me in a friendly way about the health of my wife and our new-born baby son, now fifty years old. Then he delivered his verdict in a quiet, even voice. “There are two ways of doing calculations in theoretical physics”, he said. “One way, and this is the way I prefer, is to have a clear physical picture of the process that you are calculating. The other way is to have a precise and self- consistent mathematical formalism. You have neither.”
I was slightly stunned, but ventured to ask him why he did not consider the pseudoscalar meson theory to be a self- consistent mathematical formalism. He replied, “Quantum electrodynamics is a good theory because the forces are weak, and when the formalism is ambiguous we have a clear physical picture to guide us. With the pseudoscalar meson theory there is no physical picture, and the forces are so strong that nothing converges. To reach your calculated results, you had to introduce arbitrary cut-off procedures that are not based either on solid physics or on solid mathematics.”
In desperation I asked Fermi whether he was not impressed by the agreement between our calculated numbers and his measured numbers. He replied, “How many arbitrary parameters did you use for your calculations?” I thought for a moment about our cut-off procedures and said, “Four.” He said, “I remember my friend Johnny von Neumann used to say, with four parameters I can fit an elephant, and with five I can make him wiggle his trunk.” With that, the conver- sation was over. I thanked Fermi for his time and trouble, and sadly took the next bus back to Ithaca to tell the bad news to the students.
• The climate is arguably the most complex system that humans have tried to model. It has no less than six major subsystems—the ocean, atmosphere, lithosphere, cryosphere, biosphere, and electrosphere. None of these subsystems is well understood on its own, and we have only spotty, gap-filled rough measurements of each of them. Each of them has its own internal cycles, mechanisms, phenomena, resonances, and feedbacks. Each one of the subsystems interacts with every one of the others. There are important phenomena occurring at all time scales from nanoseconds to millions of years, and at all spatial scales from nanometers to planet-wide. Finally, there are both internal and external forcings of unknown extent and effect. For example, how does the solar wind affect the biosphere? Not only that, but we’ve only been at the project for a few decades. Our models are … well … to be generous I’d call them Tinkertoy representations of real-world complexity.
• Many runs of climate models end up on the cutting room floor because they don’t agree with the aforesaid programmer’s beliefs, understandings, wrong ideas, and misunderstandings. They will only show us the results of the model runs that they agree with, not the results from the runs where the model either went off the rails or simply gave an inconvenient result. Here are two thousand runs from 414 versions of a model running first a control and then a doubled-CO2 simulation. You can see that many of the results go way out of bounds.
As a result of all of these considerations, anyone who thinks that the climate models can “prove” or “establish” or “verify” something that happened five hundred years ago or a hundred years from now is living in a fool’s paradise. These models are in no way up to that task. They may offer us insights, or make us consider new ideas, but they can only “prove” things about what happens in modelworld, not the real world.
Be clear that having written dozens of models myself, I’m not against models. I’ve written and used them my whole life. However, there are models, and then there are models. Some models have been tested and subjected to extensive V&V and their output has been compared to the real world and found to be very accurate. So we use them to navigate interplanetary probes and design new aircraft wings and the like.
Climate models, sadly, are not in that class of models. Heck, if they were, we’d only need one of them, instead of the dozens that exist today and that all give us different answers … leading to the ultimate in modeler hubris, the idea that averaging those dozens of models will get rid of the “noise” and leave only solid results behind.
Finally, as a lifelong computer programmer, I couldn’t disagree more with the claim that “All models are wrong but some are useful.” Consider the CFD models that the Boeing engineers use to design wings on jumbo jets or the models that run our elevators. Are you going to tell me with a straight face that those models are wrong? If you truly believed that, you’d never fly or get on an elevator again. Sure, they’re not exact reproductions of reality, that’s what “model” means … but they are right enough to be depended on in life-and-death situations.
Now, let me be clear on this question. While models that are right are absolutely useful, it certainly is also possible for a model that is wrong to be useful.
But for a model that is wrong to be useful, we absolutely need to understand WHY it is wrong. Once we know where it went wrong we can fix the mistake. But with the complex iterative climate models with dozens of parameters required, where the output of one cycle is used as the input to the next cycle, and where a hundred-year run with a half-hour timestep involves 1.75 million steps, determining where a climate model went off the track is nearly impossible. Was it an error in the parameter that specifies the ice temperature at 10,000 feet elevation? Was it an error in the parameter that limits the formation of melt ponds on sea ice to only certain months? There’s no way to tell, so there’s no way to learn from our mistakes.
Next, all of these models are “tuned” to represent the past slow warming trend. And generally, they do it well … because the various parameters have been adjusted and the model changed over time until they do so. So it’s not a surprise that they can do well at that job … at least on the parts of the past that they’ve been tuned to reproduce.
But then, the modelers will pull out the modeled “anthropogenic forcings” like CO2, and proudly proclaim that since the model no longer can reproduce the past gradual warming, that demostrates that the anthropogenic forcings are the cause of the warming … I assume you can see the problem with that claim.
In addition, the gridsize of the computer models are far larger than important climate phenomena like thunderstorms, dust devils, and tornados. If the climate model is wrong, is it because it doesn’t contain those phenomena? I say yes … computer climate modelers say nothing.
Heck, we don’t even know if the Navier-Stokes fluid dynamics equations as they are used in the climate models converge to the right answer, and near as I can tell, there’s no way to determine that.
To close the circle, let me return to where I started—a computer model is nothing more than my ideas made solid. That’s it. That’s all.
So if I think CO2 is the secret control knob for the global temperature, the output of any model I create will reflect and verify that assumption.
But if I think (as I do) that the temperature is kept within narrow bounds by emergent phenomena, then the output of my new model will reflect and verify that assumption.
Now, would the outputs of either of those very different models be “evidence” about the real world?
Not on this planet.
And that is the short list of things that are wrong with computer models … there’s more, but as Pierre said, “the margins of this page are too small to contain them” …
My very best to everyone, stay safe in these curious times,
w.
PS—When you comment, quote what you’re talking about. If you don’t, misunderstandings multiply.
H/T to Wim Röst for suggesting I write up what started as a comment on my last post.
Wow! Talk about memory lane!!
(and thanks for that first picture, makes my Wire Spaghetti seem less horrendous now)
– The more tunable parameters a model has, the less likely it is to accurately represent reality.
My thesis advisor in College had a similar conclusion essentially- the more parameters your algorithm needed, the less real it was. This was for Computer Vision Algorithms, but I’ve found it fits everywhere….
Spaghetti wire wrap was fairly common, and repairable too. I built an 8085 based computer using wire wrap. I also fixed a radio observatory ADC board that used wire wrap all in the early 80s. In regards to an old computer with 768KB of core memory, I can only imagine that would take up an enormous amount of space. I saw 4KB of core memory in an old BASIC-4 machine and the core memory was 1/2″ thick and the depth and width of the steel cabinet it was in.
The problem with wire wrap is that as clock frequencies go up, the wires stop being wires, and start being antennas.
Mark, you actually have the same problem with long traces on PC boards. Especially on single layer boards. Multi-layer boards make that less of a problem now.
True, however PC traces are both more controllable as well as being completely repeatable. In PC traces you can lay down guard traces surrounding problematic traces. This is not possible with wire wrap.
Sure, but I have had to wire wrapped a few twisted pair to fix problems that had no easier solution and rerouting didn’t take care of ’em.
In 1973, I wire wrapped several 11 inch by 8 inch boards populated with sockets to hold 7400 series gate ICs. It was a 4 bit microcontroller, complete with magnetic core memory. We got the thing going, but only briefly and only if it never moved.
As soon as we put it into a truck to gather data, it stopped working forever.
The problem was sockets and wire wrap, and connectors between boards.
The reason that computers work today is the massive integration of the transistors, gates and functions…all done on one chip. Interconnects are the killer. Connections are unreliable. Put in thousands of socket pins and wire wrap connections and you have junk.
Indeed, wire wrap was often used in small volume electronics. Even big things like computers, but when only a few were made, it was much cheaper than designing PCBs.
Wire wrap got really interesting back in the early 60’s when they were using Teflon for insulation. Took ’em a while to figure out that it suffered from cold flow and when wrapped under even slight pressure it eventually developed highly intermittent shorts. We had to field replace a whole lot of backplanes when it was finally discovered.
Yes, but core memory was great fun. I had the job of running a PDP11 and you could halt the processor, switch the power off overnight, go back and switch it on the next morning and it simply continued from where it had stopped.
Great post, Willis, as usual.
Wow, that Datapoint 2200 brought back memories for me. I worked with Datapoint systems in the late 70s and 80s, then for Datapoint themselves. All the Ford dealers in Europe used business systems written in Databus. Sadly, Datapoint lost its way along with most other minicomputer manufacturers, but for a while led the industry in networking (ARCnet) and desktop videoconferencing, used by the US military and I think NATO.
Anyway, back on topic, you are so right about models. The only 100% accurate model of reality is reality itself. Everything else is a more or less useful approximation and subject to significant amplification of human error.
https://www.edify.org/steve-james/
https://www.nytimes.com/1982/05/14/business/company-news-executives-resign-posts-at-datapoint.html
http://www.datapointremembered.org/in-memoriam/
You sure that was not Datashare?
Our Datapoint 2200s used Datashare.
R
Datashare was the multi-user evironment. Databus was the programming language. Datashare allowed multiple intelligent terminals to communicate with each other without a host. Datashare was the multi user interpreter.
They had a DOS also, and to go with it a sort of primitive Bash type scripting language called Chain.
Took a lot of hard work to destroy Datapoint.
They were the original contractors with Intel for the 8008, I think I remember. Didn’t like the result, came up with their own instruction set (and I think they did a better job, especially with the followon 5500). They told us once we were the second biggest customer after Safeway, and every once in a while they’d ship us an extra printer or processor, then take it back with apologies a week later. A year or two later, they imploded when a new accountant got curious about a bunch of hotel rooms being rented long-term, and found that someone had wanted to continue their streak of profitable quarters, ordered and shipped a few extra units, then took them back. Eventually it got out of hand, and when the financial skulduggery went public, they fell apart. Or so I heard. It did explain the extra equipment they’d send and take back. Might just be confirmation bias….
Great post Willis. It is nice to be reminded of the old days with computers and models and numerical solutions to the models.
I still have a “Trash 80” somewhere in my office. Fun little machine.
About models and their stability, I have a short story that happened in a partial differential equation (PDE) class I took in a class on Engineering Mathematics.
One of our homework assignments was to use a particular numerical method to solve a given PDE, Well it turned out later this tricky professor asked us to use an unstable numerical scheme that was unconditional unstable for any (delta x, delta t) we selected. In previous assignments we had learned that too large of (dx,dt) choices could be improved by using smaller (dx,dt) values.
When we turned our homework in, he took a few minutes to go through our homework papers. Most of the class gave up when the procedure blew up and blew up sooner and more with smaller delta t values. However there was one clever student who turned in some results that appeared to be a solution.
The next class day the professor handed back our homework papers and everyone who gave up and said that a numerical solution could not be found got an A on the assignment. He proceed to ask the one guy who got a solution to stay after class. He then put the problem and the numerical method on the blackboard and demonstrated to us that no solution could be obtained with that numerical scheme, and that the smaller delta t we used the quicker and more violently the “solution” blew up. And he wrote on the blackboard the numerical scheme and labeled it the “unstable scheme”. He then demonstrated that what weighting factors we had to use to make the scheme conditionally stable.
That was over 50 years ago and I still remember the lessons (not the details) of checking (V&V) our numerical solutions under a variety of circumstances. Here is a small list of the lessons learned that day but appreciated more and more with the passing years.
1) When presented with a numerical solution to differential equation(s) the solution is always an approximation.
2) Good engineers, scientists always include some analyses of the likely range of errors in the results.
3) Producing reasonable numerical results to PDES is not for the novice, tricky, or dishonest person.
4) Numerical analyses solutions to a single PDE is often difficult and quantifying the errors is also an approximation.
5) God save us from those who produce numerical solutions to a large number of coupled or linked PDES and claim they know the solutions to the models are correct and accurate to a specified range of uncertainty, or worse, they give the solutions as tested and perfect.
6) In hindsight, it was a blessing to learn mathematics and ways to apply and developing computer solutions to mathematical problems in the dawning age of digital computers.
Thanks again Willis.
I hope to see more posts from you on these modeling/computer topics.
That’s not technically true.
The models are true enough to commit to building engineering models. Those models are tested thoroughly. Then full sized units are built, and those are again tested thoroughly. Only at that point do you commit to production.
We actually have certain configurations that we know the model output will not be up to snuff. Most of them work great if you keep the flow in the laminar configuration, but they can get a little hinky when turbulent flow emerges, particular in the transonic arena. That is why P-51 and P-38s that broke the sound barrier in a dive tended to lose their wings (the models don’t show that). The forces were definitely not linear. There are models for transonic flight, but the physical models still get put through their paces in a wind tunnel to double check the output and surprises still occur.
The F-86 Sabre could also go supersonic in a dive — they fared better w/the swept wings & fully-movable elevators.
I can relate to MarkW’s comment:
Back around late 1980s to early 1990s, I worked on a hybrid AI and numerical based code to “tune” an high current, pulsed particle accelerator. The code needed to adjust various “steering” coils to keep the beam going down the (near) center of the beam tube. The accelerator pulsed at approximately 1 Hz, with the pulses lasting 10s of nanoseconds. The beam tube could not withstand continued strikes of the multi mega-volt by few kilo-amp electron beam. The beam position was measured at intervals. Thus, the task was to adjust the coils to center the next beam pulse based on the previous pulse positions down the beam tube.
One day, I was talking with the chief numerical modeling physicist about issues we had encountered between the actual beamline performance vs the modeled predictions. He made the remark (which I have to paraphrase as its been to long to provide an actual quote): “Our models are good enough to design a beamline, but not good enough to predict exactly how they will operate.” At the time, I was taken aback. After all, this was just particles (albeit relativistic ones) interacting with electric and magnetic fields. But then, as I considered the issues with all the manufacturing imperfections of all the magnets and electric coils, cascaded with all their alignment errors, his statement made complete sense; this was the whole point of the “tuning” system we were constructing.
Mr. Eschenbach,
Brilliant. Thanks.
It’s good to be reminded of things like this. We forget— and we also forget that the young have little or no knowledge or experience of history.
CDC = Control Data Corporation
DataPoint Corporation
DEC = Digital Equipment Corporation
Data General Corporation
Like you, I started young. I’ve been through punched tape and Hollerith cards and Visicalc and Lotus 1-2-3 and Excel and Reverse Polish Notation (RPN) and Fortran and Cobol and Basic.
Throughout it all, I’ve watched with astonishment and amazement as people treated the output from computer models as if it were the revealed truth.
God bless von Neumann. He knew— and he communicated that knowledge better than anybody ever did or ever will:
“Give me four parameters, and I can fit an elephant. Give me five, and I can wiggle its trunk”.
Started with Algol on an Elliot 503. Most fun language was Forth (still used to control telescopes, I understand). You could buy the core on a plugin cartridge for the Commodore 64 I got for the kids. It’s a threaded compact language that runs faster than interpreted languages like the Basic in the C64.
The C64 was the last machine I ever bothered getting down to machine code. The Cannon SE-100 100 memory desktop programmable calculator (early 70s) was probably the penultimate. That’s all you got.
The IPCC process for climate modelling is part of the problem, but if you must have it then how about a requirement to hold out all information on 1930-1990 in their construction (including parameter estimation), and then verify out of sample against that (plus report actual propagation of uncertainty)?
I too recall the horrors of dropping a stack of Hollerith cards and trying to splice together punched paper tape.
Phillip, I wondered if someone would mention the dreaded calamity of a dropped deck of punch cards.
Back in ’69 we had to suffer the boss’ high-school son coming in to work during his school vacation breaks.
Nice kid, but he had feet 4 sizes larger than his adolescent frame required, and we used to have bets on just when he would stumble while carrying a tray of card decks.
Of course it happened, and as Murphy’s Law dictates, it was month-end data entry.
His Dad the boss accepted the overtime cost philosophically.
Willis, thanks for the memories. I recall from card days that there was a command for the card punch to put in sequential card numbering in columns out in right(?) field. The sequential numbers would print out along the tops of the cards. If dropped, the deck could be reconstructed. I failed to find the command for you. As a consolation, if your computer does not have “Pi”, try ARCTAN(one radian)*4. You will get Pi to the limit of the computer innards instead of what you might define Pi to be, like 3.14.
For a reasonable approximation try 355/113
That’s why we drew a line across the top of the deck in its box. If you DID drop it the line helped to show if any of the cards were out of sequence when you put them back in the box.
You meant a diagonal line, right? Otherwise it wouldn’t be a help.
I recall always wearing a rubber band around my wrist, like many other co-workers, to always have one at hand to keep card decks together.
I remember a closet with paper tape subroutine rolls… Fortran I believe. Card stacks, Algol60 and Algol68 came later.
Might I add:
Bunker Ramo (used on the Ikara system I describe further down) and
Perkin-Elmer (which became, in part, Concurrent Computer Corp that I worked for over 17 years)
More memories
My first computer was an Apple11 which only had 16k bytes but had lots of mathematical operators and you could save a program and data on a tape recorder. One of the my first programs was a spreadsheet before Visicalc was available. Spreadsheets normally have only two dimensions x &y but each sheet can be piled up and referenced eg monthly accounts which can add to a year end sheet. My spreadsheet program using to for loops could have as many dimensions as one wished. That is a feature of mathematics which is not bound by physical limits. I have a book on string theory (The Shape of Inner Space by Shing-Tung Yau) which mentions 10 dimensions for the universe. Most (if not all) climate models ignore physical reality eg they ignore the 2nd law of thermodynamics, they ignore that above a wavelength of 10 micron which is about the peak emitted from the Earth’s surface CO2 only absorbs and emitts at a wavelength around 14.8 micron (ie the absorptivity is close to zero not 1 as claimed). I do not program any more. I can do with Excel all I want in my old age.
> Your average elevator control software has been subjected to more V&V than the computer climate models.
That’s because elevator software has direct impacts on people’s expected comfort, livelihoods and behavior unlike clima… oh… wait. Never mind.
Hi Willis, nice article! I am 4 years younger (less old?) than you but had somewhat of a similar start with computers, and therefore really enjoyed your article. However I am surprised you called MS-DOS a language. I first used it as QDOS (Quick and Dirty Operating System) from Seattle Computers before Bill turned it into MSDOS, but is was never a language, just an OS. Why do you call it a language?
I think DOS stood for Disk Operating System. Datapoint had its own DOS operating system. It’s first microprocessor – used in the 2200 – was claimed by some to be the direct forerunner of the first Intel chip – the 8086 I think.
It was possible to make batch programming, the later versions got more complexity, since Win NT, if I remeber well.
Yes, I was taken back a little by calling a Disk Operating System a computer language. However, it was the ability to write batch files to tell the OS what to do and when, particularly when booting, that I think qualifies it as a language.
Batch processing made a big difference, especially when you started the job just before leaving for lunch.
I thought the very first Intel microprocessor chip was the 4004. That’s what we studied in college right from the Intel manual.
It was; the four indicated it was a 4-bit processor. Intel made it big with the later 8-bit 8080. Motorola had the 6800 8-bit processor.
There was an 8008 that was I believe, an 8 bit version of the 4004. The op-code set was fairly limited.
I thought the 6500 line was Motorola’s first microprocessor.
Commodore Vic-20 had the Motorola 6502 8-bit processor — my first computer. Ran on “Vic–Basic” or straight machine-language if you could do it.
I developed some of the earliest electronic ticketing systems for public transport by writing assembly code for the 6502. Amazing what we could achieve with a 2k EEPROM.
That’s what we programmed in machine code in college – had to ‘jam’ the instructions.
Oh wow, my first computer was also an SCP machine, though the OS was called 86-DOS by the time I bought it. The Seattle small assembler was really fast.
Willis’s comments about computer models reminds me of Bob Pease’s comments on SPICE, “SPICE will lie to you”. Modeling circuitry is a much simpler problem than modeling than climate and also orders of magnitude easier to verify. This also goes along with “all models are wrong, but some are useful” in that while SPICE does not give a perfect answer to circuit simulation, the answer it gives is generally close enough to be useful – provided that the circuit model was sufficiently detailed.
Like, Willis, my first exposures to computer programming CDC machines, first being a CDC 1700 and second being a CDC 6400.
I’m right with you. I remember taking a stack of computer punch cards to be processed (FORTRAN) and dropping them on the floor in 1969. About 4 hrs of work scrambled beyond use. My first company I started was providing computer services to IBM midrange customers. The second was based on OS2 servers and dialup. Other than hardware, the biggest step forward has been object based programming languages. You can hire an experienced programmer and turn out the work of literally 60 programmers from the 1980’s.
A classic story of a guy with the same scrambled mishap. He used a version of Fortran that allowed more than one statement on a card. He then developed a programming style where each card had a label, and read like
860 X=X+1; GOTO 870
Now he could shuffle cards and the program still ran ..
Very quickly started putting a commented out sequence number at the end of each card.
Just remember to not number sequentially. Just in case you discover you have to add a couple of lines in the middle of your program.
I soon learnt that a diagonal line drawn across the edge of the stack made re-sorting a lot easier 🙂
And I too was a taken aback by the statement that “Microsoft Disk Operation System” (aka MS-DOS) was a programming language
Maybe it’s a reference to the scripted .bat files you can create for Ms Dos?
We actually had a sorter in the computer room that would sort the cards if you put a sequence number on the card.
This is why a big magic marker was an essential accessory for punch card programming—with a diagonal stripe across the top it was way easier to recover from disaster.
Ah yes – the fat felt-tip marker-pen.
I remember now…
RE
You can program in DOS using script or .bat files
I was going to comment about MSDOS being an OS rather than a language. It often came with Basic bundled with it.
A few years before SB/MS/PC DOS, there was CP/M 80.
The CP/M distribution typically came with a macro assembler – that was useful as common word processors often needed some customization to deal with particular printers. WordStar for example came with source listing for Diablo 630 daisywheel printers and Epson ‘graphtrax’ augmented dot matrix printers.
Yep. Used them to make instruction manuals in the 80’s.
WordStar and Superalc. Supercalc was an early speadsheet. It had a feature I still haven’teen in Excel – being able to copy a formula or group of formulas with eoither relative or absolute references without having written the formula specifically with either type of reference.
My ex-wife worked for a lawyer who had purchased some kind of word processing system. Might have been a Wang. What I remember about it was that it used a non-standard formatting for it’s floppy disks. It also didn’t include a formatting program. If you wanted new floppy’s, you had to buy them from the company. Something like $10 per disk.
My first experience was also in college. A CDC, I do not remember the number.
In my first class we used punch cards. A few months later they were using teletype machines for data entry. Used a single line editor to do text entry. Might have been ed.
The first four languages that I learned were Fortran, Pascal, PL/M and ASM86.
For the ASM86 we used an Intel development station. It had 4 8 inch floppy drives, and the OS disk was always put in the first drive. The editor/compiler went in the second slot and the third and fourth slots were for your program and data.
I never tried to move one of those development stations, but it looked like it would take at least 2 people.
My first job out of college we used 6510 assembly. I also redesigned the circuitry and relaid out the circuit boards. There was something very satisfying about laying out tape. Both puzzle and art. In my second job I learned C, and that’s pretty much all I’ve used since then. Have done some C++ and Python though.
Interestingly enough, a few years back I interviewed with an elevator company that was still using PL/M for most of their code. All the new stuff was being done in C++, but they had decades worth of code that hadn’t been retired yet that still had to be maintained. There PL/M guy had indicated a desire to retire.
I thought I was a shoe in for that job. After all, how many people have ever worked with PL/M? I was wrong.
Back when all of us were learning programming along with being keypunch operators, a friend (a joker, no less) inserted his stack of cards into the reader and headed out the door, just a moment before the wall of printers erupted, tractoring their entire box of paper onto the floor.
Of course, he said it was an unplanned event.
I’m only 2 years younger and started with Algol on a Burroughs B5500, which I’m guessing ran around 2 MHz (that’s 0.002 GHz). I recall it had a couple megabytes of hand-woven-in-Haiti core memory cards. Punch card programs and one run overnight to get results from a chain printer. Graduated to BASIC, then C, then 286/287 ASM. My masterpiece was loading the Mandelbrot algorithm entirely onto the 287 math coprocessor. Aaaannd, I have proof:
My bad, I was thinking about CPM. Fixed.
w.
CPM is not a programming language.
.
https://en.wikipedia.org/wiki/CP/M
,
Computer professionals reading this see it as a heck of a lot of BS.
From your link:
Commands[edit]
The following list of built-in commands are supported by the CP/M Console Command Processor:[14]
Transient commands in CP/M include:[14]
CP/M Plus (CP/M Version 3) includes the following built-in commands:[15]
I call that a language, although it is indeed an operating system. Me, I used to use it regularly, so I thought of it as a language. So sue me.
w.
Consider yourself sued. All your are doing is bullshitting the non-programmers on this site.
.
“I call that a language”
…
That shows how ignorant you are of computer programming.
Brian, you are hilarious. You come in, and the biggest thing you can find to whine about is that I called CP/M a “computer language”, since it had words that I could use to make the computer do what I wanted??
That’s it? That’s your evidence that I’m not a “real programmer”?
As to “bullshitting non-programmers”, there have been dozens of programmers with much more experience than I who have commented on this thread … but none with your combination of arrogance, vitriol, and unpleasantness. And guess what?
Not one of them has accused me of “bullshitting” anyone. You’re the only one.
More to the point, you STILL have not said one word or noted one complaint or pointed out one error regarding the actual subject of this post, which as the title makes clear is computer models. You just want to accuse me of meaningless “wrongdoing” like calling CP/M a language. What’s next? You gonna blame me for misgendering a potato?
All you’re doing at this point is further destroying your own reputation. Whether or not I’m good at programming computers has NOTHING to do with what term I might use to describe CP/M. In addition, whether I am right about the computer climate models has NOTHING to do with whether I can program. You’re off on a meaningless rant about trivia.
Get a life, Brian! Do you want your obituary to read something along the lines of “Brian Jackson was really good at dragging things down, sidetracking interesting conversations, ignoring the actual issues, accusing people of meaningless offenses, and was widely known for his amazing ability to piss on birthday cakes?”
w.
Willis
What Jackson did is called “nit picking,” a form of red herring.
Brian does things using grunts and groans and huffing and puffing !
That is his language. !
As to his knowledge of computer programming….. roflmao !!
Brainless, you are the lowest sort of troll, and full of hatred.
Mr. Jackson: Thank you so much, I am a non-programmer and I was totally bull-shitted into thinking CPM was a language. Since it was so critical to the article, you sure showed him, huh? Anyway I sure would have been embarrassed at my next soiree, where I planned to impress the ladies with my “what’s your sign? Did you know CPM is a language?” patter. Shame on Mr. E for bull-shitting like that.
Mr. Jackson, THAT is bull-shitting. Hope it helps you recognize.
Willis, were you sufficiently masochistic to use MP/M?
Never.
w.
From that, one might expect that your tendencies to recognize usefulness vs. time spent, to have kicked in when confronted with an Altair 8800.
CP/M (Control Language for microprocessor, trying to make /M looking like Greek mu, for micro) was not a language either, but still an operating system made by Digital Research for the 8-bit microprocessor 8080 made by Intel.
Don’t confuse simple commands in an OS (DIR, etc.) with instructions in a programming language which can be used to write any sort of program.
I am sorry to sound pedantic but with your programming skills (way superior to mine) I thought this was understood.
Thanks, Michael. Using CP/M I can tell a computer to do a variety of things, and it will do them, but only simple things. It has a very small vocabulary.
Using Basic I can tell a computer to do more complex things.
Using Stella I can tell a computer to do even more complex things.
I have a 19-month old granddaughter. She has a very small vocabulary. She can tell me to do a variety of things, and I’ll do them … but only simple things.
So your claim is, she’s not using a language, just running my operating system?
You do sound pedantic. A “language” is, inter alia, a way to give commands, and some have larger vocabularies than others. I speak three versions of Pijin—Tok Pisin, Bislama, and Solomon Islands Pijin. None of them have more than about a thousand words, while English has about 170,000 words … does that mean that Bislama is not a language because it doesn’t have a large vocabulary?
Finally, so freaking what? I’m self-taught in everything, so it’s not unusual that I don’t use some ivory-tower-approved term. Here’s an example.
For thirty years I called the moon’s “terminator wind” the “moon wind”. Why? Because I’d often observed it during the long nights I’d spent at sea, and wondered about it, and eventually I figured out what caused it all by myself, and that’s what I named it years before I’d ever heard of a terminator.
So … does that make me more, or less, of an expert on the moon wind than some guy who learned the official name of it in a book? I think more, although of course YMMV.
However, when a man starts focusing on what I call things, rather than on my ideas and accomplishments, all I can do is shake my head and wonder why he’s wasting his time on such trivial BS … call it what you like, my friend, and let’s get back to the real issues.
w.
Michael,
Depending on the hardware design it doesn’t take but a single instruction to build any program (without I/O of course): https://en.wikipedia.org/wiki/One-instruction_set_computer.
Add some form of input and output instruction and you can then build a 3-instruction computer that will perform most programming tasks.
I too noticed the DOS/language comment and my first thought was that I didn’t agree. However since it was neither relevant, nor important, I said nothing.
However as certain individuals kept trying to turn this molehill into a mountain, I thought more about it.
The more I thought about the difference between an operating system and a computer language, the more vague it became. Eventually I decided that it didn’t really matter and that the whole issue was rapidly becoming a debate on how many angels could dance on a pin head type argument.
Brian, your apparent belief that you, and only you, have the right and intelligence to decide such issues for everyone is rapidly becoming your trademark.
This re-enforces my perception that computer modeling of climate is the ultimate intellectual hubris
Modeling can be useful in helping you figure out what it is you don’t know yet.
It’s not useful for anything beyond that.
Thank you Willis, I enjoyed your post and learned something too….
Thanks for the enjoyable read/ride Willis. I am reminded of a Kraftwerk song listened to while punching holes in cards back in the day. “Das Modell” from the album, yep, album “Die Mensch-Maschine”
I was not a Fortran Fan.
Hey, Fortran was limited but useful. (It was a lot younger in 1959 when I started on an IBM 704, as were all the other languages.) But the best programming language I have used was a strange hybrid that ran on the Control Data 3100 in our lab in the middle Sixties. You could throw both Fortran and assembly language into the same program if you knew what you were doing. I reworked Spacewar to run on the 3100 using that hybrid.
And the spaceships in the first version got bigger and bigger as the game went on. I used too large a dt, and not enough terms in the expression. The second version was better, and the third good.
Ben Franklin, asked in 1747 what electricity was good for, replied, “If there is no other Use discover’d of Electricity, this, however, is something considerable, that it may help to make a vain Man humble.” Computer programming and modeling can serve the same purpose.
.. and how many hours were spent looking for the bug just knowing it can’t be a result of your code.. only to have someone come in behind you and find it in one minute.
I didn’t spend much time at all – it was a compact section of code. In those days, mathematical operations ate up cycles, so the program used scaling, which was a shift operation. The location of the problem was intuitively obvious to even the casual observer.
Most languages that I know of will let you mix assembly language into the code.
There are a few versions of C that use pragmas to allow you to insert the assembly language directly in-line. (You have to be careful doing this as it will make porting your code a lot more difficult.) In more general circumstances, you need to create subroutines in a different module, compile the module as a library and then make a library call to pull in the ASM.
Logically one conclusion is obvious.
Isn;t that what AOC wants to do? At least the first part…
Isn’t that what was tried on January 6th, 2021?
Liar.
Yes Pelosi did
Not even close, but I’m sure that thinking this way will help you keep your sense of superiority intact.
But AOC (et al) don’t want to plug the US back in.
That sign is ahead of its time.
I have had IT support jobs in the past where actually I have had to ask ‘have you tried switching it of and on again?’ and that worked…!
Griff, I had to drive 27 miles one time to plug in a keypunch for a customer.
Griff, have you tried just switching off?
It would make no difference to the worth of your comments.
Great story, I remember some of that early on almost all watching on the side. Saw many, including family, immediately adapted to it. I played (real) softball with the brilliant man that started our university computer program, cautioned me about such. Huge machine, fortunately got smaller, didn’t cure the caution.
Why is it not understood that that is why we call them models? Cars, planes, once had one of the WWII B-17 models that were used for teaching identification. Never could get it to glide, but it had been very useful for its purpose. Ok, that’s too simple.
There’s an old movie about a plane crash in the desert. The crash survivors decide to use the parts of the crashed plane to build a new plane to fly them out of the desert.
The guy who did the design near the end admitted that while he was trained as an aeronautical engineer, he had spent his career designing model planes. When the other passengers started getting upset, he told them that models have to be better designed than full sized planes. In his words, a model has to fly without the benefit of a pilot.
Flight of the Phoenix, based on a true story.
“Flight of the Phoenix”
I recommend the 1965 version with Jimmy Stewart.
Yep, that’s it, a classic; never bothered with the modern remake.
In my experience, remakes are rarely as good as the origninal.
yes, the remake was rubbish!
My recollections:
Adjusting the tone and volume controls to get the program saved on cassette tape to load.
The evil words: Syntax Error
dBase II
Our first 5-1/4 floppy drive for the Apple.
Wow! Fast and loaded every time
Adding a Z80 CPM card and 8″ floppies to the Apple II.
A 14″ plate 10meg hard drive (Altos 4 user CPM system.
Punch tapes for a wire edm.
Love the HP-15C
Still have mine from 1983.
Lou Ottens, inventor of the cassette tape, died yesterday aged 94<a href=” https://www.theguardian.com/world/2021/mar/11/lou-ottens-inventor-of-the-cassette-tape-dies-aged-94 “> link </a>
The sound of an 8” hard sector floppy disk (on a Cromemco S-100 Z-80) click click click click …. click click click click. Microfocus Cobol. Z-80 assembler.
http://bubek.net/pics/disklocher1.jpg
I remember I got a hardware tool to upgrade a 5-1/4 180 KB memory to 360 KB make it usable both sides in a C64 floppy drive.
Amazing times 😀
“I remember I got a hardware tool to upgrade a 5-1/4 180 KB memory to 360 KB make it usable both sides ”
One of the early computers I worked on was an IBM System/3 Mod 12 mini computer. It used 8 inch singled sided floppy diskettes as input in place of a card reader. The Diskettes had a notch on one site, that if covered, prevented you from writing on the diskette. We learned that if you cut a similar notch on the other side of the disk a flipped the disk over you could also write on the back sided of the diskette. IBM frowned on this but we never had a problem with it. We used “Dykes” (Diagonal Wire Cutting Pliers) to cut the notch.
Thanks, Willis!
I had always thought that one of the biggest problems with climate models was trying to work with two massive, chaotic systems simultaneously! Per usual your posts make understanding easier; a skill that far too few ‘professional’ educators possess!
One of the few regrets I have in my life besides turning down a scholarship to Stanford was not accepting my parents offer of piano lessons when I was seven! I’ve added playing the first movement of “Moonlight Sonata” to my bucket list; Beethoven being one of the original rock stars in my opinion. Guitars are great fun, but if you’re serious about composition and songwriting piano is unbeatable! Stay safe and healthy!
I spent a good part of my career in computer service management and can attest to your observations about programming and programmers. What I noticed about Climate modeling is a tendency to either compromise, discount, ignore, or be selective about the data to fit a desired outcome. “Forcings” are the elephant in the room. When doing compute intensive programs with finite element modeling that affect design parameters of say airplanes, bridges, cars etc. one always assumes the underlying material data is correct because it has been time tested. Not so much with climate modeling. It’s the result of producing a desired outcome vs. an accurate outcome.
I was also in the hardware side of computers and found programmers to have a narrow view of how their programs would be used.
They never allowed for a cat walking over a keyboard and the subsequent effect of random inputs.
Back in the days when I was doing a lot of programming, I spent more time trapping unreasonable input than writing the core program, so that the person using it wouldn’t crash it with out of range values. I also developed Computer Assisted Instruction software to supplement my labs. I considered the input trapping to be part of the learning experience because the students (if they were paying attention) would soon learn what wouldn’t work if they didn’t know what they were doing.
Now, now – SOME of us did. Although in my professional career, I didn’t concern myself with cats; <i>users</i>, particularly from the marketing department or (shudder) HR, were much, much worse hazards.
In my day, we considered software development experience as limited by the amount of time spent supporting code you had developed in multiple installations. Until then you didn’t have the foggiest idea of how reliable, usable or functional it was.
So true! Long, long, ago I worked for Big Blue as a (rare) professional hire. I was running an in-house expected resource capacity planning (modeling) program. The program kept crashing and pissed me off so I tried to see why. Low and behold the program had no input bounds checking. At the company, EBCDIC was the encoding of choice and bounds checking non-trivial, but still necessary. To make matters worse, the program was written in APL version 1. I found two locations where bounds checking was making the program crash and filed bug reports.
The point of emphasizing that I was a professional hire, was that I had a lot of experience in programming and hardware design so picking up on the problem was relatively easy for me. The normal hires for this division were non-engineering/non-hard science types, for example my immediate supervisor was a history major. You get the picture. Everyone working with the capacity planner who hit a wrong key would worry that they had “broken” the program and restart without telling anyone.
Shortly thereafter I was offered a position with another company and since BB couldn’t/wouldn’t match it I resigned. The day I left the Branch Manager conducted the exit interview and told me the program had been used for two years before the first bug was reported. That bug was one of the two I reported, and reported just a week before I reported it. The other bug had never been reported.
Funny that many of the CVSs reported on programs to this day are essentially bounds checking problems. BUFFER OVERRUN is a bounds checking problem!
My language knowledge parallels many of others in this thread. Maybe the TI 9900 assembly, and being in the first programming class in the world, taught by Nikolaus Wirth required to learn Pascal might also be interesting.
William
You said,
When I was a young man working at Lockheed MSC, they were so desperate to hire college graduates for managers that they had a music major managing engineers.
Yea, Big Blue went through a phase where too many engineers promoted to management weren’t working out and had to be reassigned back to engineering jobs. So, they started hiring non-engineers, mostly with degrees in Business Management, for those positions. I left in the early phase of that trial. Don’t know whether it eventually worked better or not. Fortunately other companies and start-ups I work for after that still had mostly engineers and/or scientists in management. At least you could have a reasonable discussion with them.
In my company we have a core product that we use to support multiple customers.
For each new customer, we develop translators to convert the customer’s files into a standardized format that our core program processes.
We start when the customer sends us their documentation that describes their data formats. We write the translators and test it with internally generated test data.
Then we have the customer send us some of their data
There’s nothing like counting bytes in a record to try and figure out which field was increased or decreased in size, or where an extra field was squeezed in, and what kind of data is in that new field.
One of these days I may get a file from a customer that actually matches their documentation, but I’m not optimistic.
In most cases, we have to adapt our code to match the actual data, because the guy who wrote the customers code has retired, and nobody in that company knows the code well enough to make changes to it.
My cat knows more shortcuts than I do. I’ve never been able to manipulate windoze better than she can.
We’ve got a few things in common:
1. age
2. same first laptop , except I also got the serial FDD and barcode wand, upgraded memory to 32KB (voiding the warranty), but never learned Basic, or any other programming language)
3. dropped out of high school, then dropped out of grad school twice, but only diminished my earning power through these academic feats
4. have (putative) grandchildren, but have never met or talked to them, and don’t know their birthdate (twins) or location
I guess I’m like a Bizarro to your Superman (except I only have a cat)
I wish there were some way for (non-billionaire) programming-challenged people like me to access the programming skills of people like you.
For example, I’ve been looking for years for a simple database or even just a spreadsheet template to let me enter the details of my various nutritional supplements and medications so I can see if I’m getting too much or too little of each, how much each product is costing me per day and when I need to order more, etc.
A simpler problem concerns digital micrometers. I’ve bought half a dozen over the past two decades, and every one of them has a serial port. But none of the various vendors offer a data cable or software to permit them to be used with a PC to quickly record a series of measurements (as in checking neck expansion of cartridge brass). Things are actually going backwards for consumers in this respect.
Thirty years ago I could buy a multimeter from Radio Shack Canada that had a serial port and software for DOS and Windows. Ten years ago, or so, you couldn’t buy such a thing in Canada anymore AND you could no longer import it from Radio Shack USA. I had to get someone in the USA to buy one for me in a RS store and mail it to me here in Canada (thanks so much NAFTA!).
Get Livecode. Successor to Hypercard. Very, very easy to learn, and very powerful. Mac, Windows, Linux, Android. Here is the opensource version:
https://livecode.org/download-member-offer/
Thanks for the tip. I AM impressed that Livecode will run under Windows 7 with only 256MB of RAM and 150MB of disk space, and runs in compatibility mode in Windows 10, but otherwise the FAQ/FAQ left me mostly scratching my head.
Easy for you, maybe…
Very very off topic, but OK, here is how to get started. You create what LC refers to as a ‘stack’, which contains ‘cards’. A card is basically just a graphical interface background, its something to put your components on.
You then start by dragging components across to it. These are of two sorts, purely graphical elements, which are not part of the programming, just look and feel. More important, you also place things like buttons, menus, fields which are part of the programming.
Start with something really simple, like the usual ‘Hello World’. You will need to drag over two components, a field and a button. Give each of them a name, like eg ‘field1’ and ‘calcbutton’.
Next you write little scripts for each of your active components. In this case you have only one, your button. So your program will consist of a script of that button, and it will be something like
on mouseup, put ‘Hello World’ into field1
generally speaking, programming in LC consists of writing scripts for objects and events. Events can be stuff like mouseup, mousedown. Objects can be stacks, cards, buttons, fields…
Buy a copy of Mark Schonewille’s book ‘Programming Livecode for the real beginner’. I guarantee if you work through it, you’ll be able to program. Its about 200 pages and very clear and gradually leads you into how to use all the different features.
Use tab separated text files to hold any data. Don’t worry if you find his section on arrays difficult. Its just about the only section of the book that is less than clear to a beginner, but if you get to arrays in your programming, you can see the finish line and can find someone to explain them to you more fully than he does.
There are tutorials on the LiveCode site also.
The great thing about it, apart from the ease of configuring the gui, is that its almost automatically structured at some level. For any given script, you can insist on writing spaghetti. But the underlying structure of scripting components means that your overall program will be in manageable self contained blocks.
Its not R, as recommended by Willis. Its not C++. But it will get you well started.
Good luck!
Should have given a link. Buy it here:
https://www3.economy-x-talk.com/file.php?node=programming-livecode-for-the-real-beginner
I can write you a program to do this. What caliper are you using… a quick internet search doesn’t show one with serial or usb port.
That would be greatly appreciated. Of all that I’ve owned over the past 30 years, only the most recent purchase (a couple of years ago) had a name and it’s shown here:
https://www.amazon.com/Mastercraft-Digital-Caliper/dp/B07K8JVS5W
The serial port is neither shown nor described, but you can see its sliding cover at the top right corner of the display case. I’ve attached an image of it. BTW all of the digital calipers I’ve owned, and all that I’ve examined (ie. cheap ones) have the identical flat board with four flat contact strips. The second picture is of the port of a no-name caliper that’s at least 20 years older than the Mastercraft one. Their measurements differ by less than a tenth of a millimeter. Oops! The site will only let me attach one image.
Komerade,
After writing my first response to your kind offer, I found a link offering something close.
http://robotroom.com/Caliper-Digital-Data-Port.html
I haven’t studied the contents closely, but most of it is clearly beyond my technical competence. The cable he proposes would have to be modified by replacing the DIN connector with a usb one – something I think I could manage if I knew which of the USB pins to connect. And his software is meant to send the caliper data to a milling machine(?), not a PC. But I assume his notes would help in writing a suitable program for sending data to a PC.
cheers,
Otro
Nice, Willis. How anybody less than 70? and with a modicum of science/engineering training can appreciate the modern age, nor its vast complexity I don’t know. I regularly pay homage to the billions of transistors between here and there -created by human beings, not to mention the billions of lines of code -written by fallible human beings. My first experience was programming an HP 2116? to play ping pong with the lights and switches at Stanford in 1970. I think we programmed it with the switches….. IBM cards came later. Who will appreciate this when we are gone? Certainly not the modern crowd I think.
John
Yep.
I remember sitting at a pc with one of my nieces who recently got her degree in IT.
I brought up the DOS command screen.
She said – “what’s THAT?”
I had a on and off relationshipwith computers. I had a couple of visits to CDC in Bloomington Minneapolis, It was a fascinating place and made a big impression on me. As a result I still take an interest in the Vikings, Twins and Timberwolves.
My experience with computers and languages is in the field of test engineering, finding other people’s dropoffs. Most had bespoke systems and languages. One was a 10 bit system by Elliott Automation, I think, that used two and a half 54 series logic circuits.
As a fellow programmer, though not nearly as seasoned as you, I wholeheartedly agree. I’ve built models for artillery shells and other well understood, narrowly defined physical processes. Even these can go horribly wrong. Garbage in, Garbage out. Computers are finite machines. Regardless of how slick the interface or how many processors you cram into it, it’s just a really fast Abacus. There’s no way you can predict the future of the weather, the stock market, or any other chaotic, nondeterministic process with a finite machine. That does not compute!
An abacus can’t do logic tests and alter what it does based on the outcome of the logic test.
Also has non-existent undo support…
Do NOT fold, spindle, or mutilate.
Think or thwim.
Sigh, all this reminds me of all the web pages I’m getting close to writing but haven’t had time to for 25 years yet.
Brief notes:
Dad taught me the binary number system when I was about 7, then how to count in binary on my fingers – they go from 0 to 1023.
In 1963 he had designed the Bailey Meter 756, the first commercially successful parallel processor. No one knows that because it was a process control computer. It was successful because it was decimal machine and that made it a lot easier for power plant people to deal with. He said it was so easy to program a 12 year old could do it, and taught me how to program the I/O processor. That was before core memory was common (your photo), the system executed off a drum – think spinning cylinder with a magnetic coating on the outside.
It wasn’t until I got to CMU in 1968 that I realized I was born to be a programmer. The first course was using Algol on the Univac 1108. The lecturer was Alan Perlis, one of the inventors. In meeting the keypunch, which did not have a backspace key to put the chads back in, you had to duplicate the card to the point of the error and deal with it there.
The duplicate function brought an epiphany – copying audio tape or paper lost information. Here I could type a card, duplicate it 100 times, and the data on the first and last cards would be identical. Absolutely amazing.
Near the end of the semester my first on my own, for the heck of it program had a goal to simulate the trajectory of an Apollo module on its flight to the moon. First came Earth – print filled in circles and open ellipses on the line printer. Then Mercury – an orbit around the Earth. Those orbits kept spiraling inward. Thinking that might have been related to discontinuities in the atan() function, I changed to atan2() [hmm, very fuzzy memory, I should check the code], but the spiraling still happened. The problem might have been single precision floating point, but there’s little reason to explore that.
I set it aside as the end of the semester approached and never got back to it, but I learned a tremendous amount about simulations, cometary (parabolic and hyperbolic) trajectories did bizarre things until I adjusted the time steps as a function of distance between Earth and the comet, and all the “1 plus epsilon” issues with floating point.
In my sophomore year I got a parttime job operating the Computer Science Dept’s new DEC PDP-10. It remains my most favorite computer ever, a position that no computer today or in the future can ever equal. OTOH, I can’t go back to it either. Progress marches on, but at least we can look back!
One thing I did write recently was in response to the Y2K transition – can you believe there are actually people who don’t remember it?
Before that, there was the DATE-75 problem on the PDP-10’s “TOPS-10” operating system. I figured I better write that up as a reply. Old PDP-10 fans might like http://wermenh.com/folklore/dirlop.html – covers that, but also has examples about why assembler programming on the -10 was more productive than pretty much any other system I’ve encountered. A wonderful machine on many levels.
I never verified it but was told by several friends familiar it that DEC’s Fortran compiler wrote tighter code than most assembler programmers could.
Ahem. I was at DEC and I was told that by one of the compiler’s authors. That was an utterly ridiculous claim, as TOPS-10 and most user level programs dedicated several of the 16 general purpose registers to purposes like point to important data structure. In assembly code, they were “just there” across a module’s subroutines whereas Fortran had to pass them on the stack or other memory. I think my reply started out with “Well maybe….”
Around 2005, I was back at DEC and decided to see if I could make the IP checksum routine faster on DEC’s Alpha processor. It wasn’t much code, but I had already come up with better byte swapping routines, and well, manually writing RISC assembler code is difficult, what with dealing with caches, pipelines and other cruft that original (discreet transistors!) PDP-10 didn’t implement.
I came up with code that looked pretty good to me, nicely unrolled, did 32 bit math instead of the 16 bit inherent in the checksum. (I couldn’t do 64 bit because I had to handle overflows.) There were a couple wait states, but better than the old code.
As a lark (and recalling a CMU case where someone found his LISP code was faster than his compiled code), I rewrote things in C to try out. The C code was faster. The C compiler used a different computation that allowed some intermediate results to make it through the pipeline and the generated code had no wait states.
With my 30 year sense of superiority over compilers nearly completely smashed, I rewrote my code to use the compiler’s algorithm and came up with something that was a little faster checksumming data that was not in the processor’s cache.
Between multiple levels of caching, data and instruction pipelines, look ahead processing you need to be an absolute ASM guru to have a chance of outperforming most modern compilers.
Beyond that, every time you code gets ported to a new computer, you have to go through it all over again.
Yes, and I was absolute ASM guru back in the day. And the IP checksum code is heavily used by NFS, and I saw I could do a better job than the original engineer did. The exercise was worthwhile.
Ever played with Grey code?
All the time. But that is a thing more applicable to those of us that design the hardware.
I didn’t learn about that until I got to college. I was pleased to realize the PDP-10’s line printer had an optical drum position sensor built around a spinning disk that used Gray code. I didn’t try hard to do that on my fingers. Even if I figured it out it would be a challenge to convert between the two.
https://www.allaboutcircuits.com/technical-articles/gray-code-basics/
Then you know what binary 4 looks like on your hand. Not many people will know what you really mean if you yell, “Four!”
Yeah, I kinda try to keep my fingers moving as I go past four and also aim them elsewhere.
I think I wrote a simple Algol to list all possible four letter words. Umm, 26 would fit on a 132 character line printer line, two blocks of 26 would fit on a page, that still requires 26 x 26 x 13 pages. I recall only printing 10 pages or, maybe starting ‘d’ or ‘f’.
That was probably inspired by Arthur C Clarke’s The Nine Billion Names of God and fulfilled a need to see for myself that a trivial subset of that grander goal was just a simple matter of programming.
https://urbigenous.net/library/nine_billion_names_of_god.html
Also, “four” is appropriate. Apparently some Audi Quattro owners have gotten license plate “QQQQ” approved since it sort of looks like the Audi logo.
I have a computer model of the second world war written by a team in Eastern Europe. I’ve run it several times (it takes months to run a simulation on a PC.) In my virtual ensemble of model results I have Germany winning the war most frequently, which is why I am writing this in German.
Fortunately for the rest of the world, Hitler was no where near as good a general as he thought he was.
Kind of reminds me of climate scientists.
MarkW: Or you could say the model inputs for Hitler’s military aptitude were….not validated. Hitler’s model ran into reality at Stalingrad. AGW ran into its own Stalingrad some time back, but our press hasn’t noticed.
Ever watch “The Man in the High Castle?” (Amazon Prime Video)
Isn’t that about the fraud Mann?
Did you ever play Eastern Front on the Atari?
Or the paper version from SPI with 3000 individual cardboard counters and 5 foot square paper map?
Hi Willis,
This is a very good post, and leads to a couple of things that you might be able to clarify for me.
1) are the parameters used truly tuned within the main model or are they determined by independent experimentation?
2)Do the equations used within models satisfy the basic physical laws of mass,momentum, energy and and do these models balance all nodes within the grid to satisfy these laws at each grid point (viscous and non viscous flow, natural convection/advection, enthalpy states, partial pressure analysis to derive latent heat transfer, etc) ?
3) what is the general gridding methodologies and are models done in Finite element, volume, or difference methodologies?
The purpose behind my questions is that I suspect most models are exercises in curve fitting and not iterative solutions of the applicable physics equations.
But as in all things computer related… GIGO.
With respect to (1), both. There are two basic parameter tuning processes. I wrote about and illustrated both in my years ago climate models post here, ‘The trouble with climate models’.
I found it at https://wattsupwiththat.com/2015/08/09/the-trouble-with-global-climate-models/
Rud, you hit the nail on the head, and the post that Joseph found is EXACTLY what I was referring to. I have to admit I asked leading questions like a prosecuting attorney because I knew these answers existed and wanted to bring them into the record again.
From way back in my CFD days, I learned that true models were never “tuned” in themselves. parameterizations were ok, but you absolutely had to determine those coefficients in independent experiments My specific experience utilized the Hazen-Williams pipe flow approximations). Once you built your model, the only things you should do are to change boundary conditions or add additional equation terms and rerun the model. and by terms I mean a complete new equation set.
As soon as you start tweaking coefficients within the model itself because it didn’t predict the expiremental values you meaured you left the realm of reality and entered “curve fit land” and lost all ability to trust any “prediciton” your model made. Curve fitting is great at interpolating between know data points but absolutely sucks at extrapolation outside of the measured data set. Essentially you are trying to empirically fit a curve to take care of every possible variable, but some of those variables are codependent with other variables, or non linear/multi order, and so require their own set of equations to balance to come to the correct “parameter”.
I didn’t see them in the list of parameters that Willlis provided, but I’ve been told that the parameters include things like amount of aerosols in the atmosphere, this parameter is allowed to vary over time, but I don’t believe it can vary spatially. It just takes one number for the entire world.
Works for really big volcanic eruptions, not so much for urban smog.
The problem is that even today, the total amount of aerosols in the atmosphere is a bit of a guess. As you go back in time, the total amount becomes an even bigger guess.
This allows the modelers to put in whatever the amount of aerosols they need to get the results they are looking for. The ultimate Finagle’s constant.
Even better, we don’t have a good understanding of how to map the industrial aerosol concentration into temperature change. The uncertainty is so high, at least a couple decades ago, we aren’t certain about the sign of the effect.
With respect to 1), what Rud Istvan said. When he talks, I listen.
Regarding 2), yes, in theory, and within limitations. The limitations are round-off errors. I discuss this question here.
However,
a) there is no guarantee that all relevant physical laws and limitations are included,
b) if all of them were included they’d all get the same answer, but they don’t, and
c) far too many things are parameterized rather than the result of underlying calculations, including lots of important sub-grid-scale phenomena.
Regarding 3), I don’t know. I’ve stayed away from climate models as much as possible.
w.
And the real problem is that the programmers and their apologists say that the models are all based on physics — except when it comes to parameterizations. They are akin to what engineers call fudge factors. Subjective estimates of how things work. Sort of like E = b(mC^2 + e)
Willis, a well developed model satisfies the physical constraints at all nodes. When I wrote CFD code back in the day, equations for grid point state and flow would balance to match basic physical laws like conservation of momentum and energy. The energy flux from into a specific node would match the energy flux out of a specific node. the same for momentum and such. If a physcial manifestation , say aerosols, could not be described by an energy or momentum calculation, then it could be parametrized ONLY in terms of an equation that could be, and those parameters had to be calculated through independent experimentation. Your a), b), c) responses above indicate exactly what I am referring to. Your link to your article noting Gavin Schmidt had ne idea if his models could satisfy these requirements is exactly what I expected to find.
Your “emergent phenomenon” explanations are absolutely spot on, and until we can solve the navier-stokes equations without simplification we are unable to predict these phenomenon. Engineers come close with our general understanding and expressions like Nusselt, Prandtl, Rayleigh, and Grashof numbers, but they are approximate relations that require an critical mind to evaluate each and every application and the assumptions made in each case.
As far as point 3), I suspect that most models are finite difference which simply seek to measure the change in each grid point over time and drive that to zero with each successive iteration. This can be good for many generalized area, but anyone who deals in physics knows that when working with fluxes (energy, mass, etc) then a finiate volume approach neeads to be used, which is a much more complicated calculation set.
In the end, and to be simple, I suspect I agree with you greatly. To paraphrase mister Twain, the robustness of our scientists modeling is greatly exaggerated.
willis i replied to clyde by accident because of internet oopsies.
Hi Willis
The Prof George EP Box quote “all models are wrong…” is accurate enough. Even the simplest models are missing some factor — maybe it’s the tiny change in the plastic regime of a torsion bar depending on temp and humidity in an elevator — but the expected and extreme cases accounted for may not be affected by that parameter. (Not modeled, and it’s ok.) By “wrong” one can impute “incomplete” in that a change of materials classes could very well require knowing the plastic regime parameters, at which point the model’s lack of that particular control becomes apparent.
They are also wrong in that they can only have finite representations of infinite numbers.
https://en.wikipedia.org/wiki/Floating-point_arithmetic
As a result a small error is introduced every time you do a calculation with floating point numbers. When you do an iterative calculation, such as Willis’s example above, this error compounds each time through the loop. After millions, or billions, of iterations the error can be almost as big as the number you you are examining.
There are ways to minimize this effect. For example, an iterative algorithm for a square root actually yields successively better approximations.
You can extract square roots (providing you can represent the number you are rooting) exactly. No need for approximations, except in choosing when to stop adding significant figures.
Just because some functions under some conditions will converge doesn’t mean that all functions converge under all conditions. The trick is to find a function that doesn’t diverge.
you can minimize but never eliminate. and if the scale of the error still overshadows the expected result you have garbage.
Willis,
It sounds like our paths were very similar, except I’m 15 years behind you. I was lucky in that personal computers like the TRS-80 and Apple II came out when I was in High School. I spent a lot of time up at my local Radio Shack in those days banging away on a Model I TRS-80 (4K of RAM and cassette tape storage). The manager didn’t mind as long as I unofficially helped out with the cleaning and restocking a bit. By the time I got a job there, I knew the products and where the were in the store better than the manager.
Anyways, I just wanted to add my voice to this conversation: a computer model can only be as good as our understanding of what we are modeling. On top of that, even if you do a good job with V&V, you will still have bugs and undetected loss of precision due to a variety of issues. This is a particularly nasty problem when the iteration count gets high. And if you don’t do a good job with V&V, you are doomed. In other words, you are dead on Willis.
And just in case you are wondering, I’m not just some entry level grunt slinging java-script and html, I have worked in the embedded control world for most of my 30+ year career. If you wish, you can check out U.S Patent 6584356, which is for the first use of a real time operating system in a cardiac pacemaker.
“…a computer model can only be as good as our understanding of what we are modeling.”
I agree. Another way of stating that is: If you don’t understand the problem well enough to work it on an adding maching (given enough time of course), you can’t model it.
Great trip down memory lane. My first language was FORTRAN (actually WATFOR on punch cards), but my first love was Algol, which in a way lives on as Pascal. Pascal was the great breakthrough language in that it could rapidly compile on an Apple ][ with only 64 kbytes of memory!!! Pascal is also considered “extinct” but it lives on as Delphi and Lazarus and of course it helped to make C++ a more useable language than the unnecessarily cryptic C. A well-written C++ program resembles Pascal more than C.
Here are a couple languages that I also played around with over the decades. Forth was a language written by someone that must have fallen in love with HP calculators. Either that or someone from Poland walking backwards. Then there was the bizarre Prolog, where you specified the answer and the computer figured out how to get there. It almost never worked for me.
These days, I like to use opensource when possible, so if I have to tinker with Matlab, I prefer Octave. I’ve used R, but not to the level you have mastered.
Finally, I completely agree with your comments about models. I remember creating an NFL football betting system where I had “regressed the past”, very much as climate models attempt to do, but with far fewer parameters. It predicted the past beautifully. The future, not so much.
Willis, do you have a web site where it is possible to download sample and/or simple R scripts to perform some of the magical analysis you do on things like Ceres and Argo data?
Forth! I forgot Forth on my list of languages, and also Logo.
As to a web site, there are a few scripts in my DropBox public folder.
The problem with them are:
1) Each of the CERES datasets (e.g. TOA shortwave, TOA longwave, and the like) is about 50 Mb in R “save” format and expands to about 130 Mb once opened, and there are 34 of them. That’s a total of a gig and a half …
2) My scripts are not just user-unfriendly. They’re actually user-aggressive. They’re never in order, I call on things at the top that aren’t defined until the bottom, names are clear to me but no-one else, if you program you know the drill.
3) They depend on functions I’ve written in my “CERES Functions.R” setup file
4) A whole heap of the rest depend on my personal “Willis Functions.R” setup file.
Rummaging in my dropbox folder I only find a folder with code for my Corona graphs. It does have my “Willis Functions.R”.
I suppose at some time I’ll get around to putting together a folder with all the individual stuff to run up a CERES graph … but discovering new things is so much more fun, plus ooooh, look over there, shiny!
w.
“if you program you know the drill.” very true, and if you program for a living, and do it your way, they show you the door very quickly.
I’ve been a programmer for almost 40 years, and what Willis describes happens at every company I’ve ever worked at.
Sure, the style manual may say otherwise, but when it’s crunch time and deadlines are approaching, style, shmyle, get the dam thing running.
Mr. Jackson: Your comment would be cruel, maybe funny, if he had ever been shown the door by “them”. If “they” never showed him the door, then your comment shows your ignorance. Have you got an example of a “they” who showed the door to the target of your obsession?
How could one access your Public Dropbox folder?
It’s here, but don’t abuse it. Also it contains a lot of old, meaningless, or outdated stuff. Enjoy.
w.
Thanks Willis. This will be very educational and a perfect way to get reacquainted with R.
Atari produced cartridges for Forth and one of their more interesting educational forays was a cartridge with Logo and Pilot. Pilot used recursion and produced some of the most interesting graphics I have seen outside of fractal graphics.
My Commodore Vic-20 also had a Pilot cartridge and “Compute” magazine showed a few very interesting graphics programs for it to type in & run.
“Great trip down memory lane. My first language was FORTRAN”
My first language was also FORTRAN, an old version called “PDQ FORTRAN” which ran on the first computer I ever programmed on, an IBM 1620, which had a console typewrite for I/O and also a Card Reader/Punch for I/O and something like 60KB of storage (“Core” storage like the picture at the top of the article). I also owned a APPLE ][ and learned to program PASCAL on that machine.
Most of my programming was “Business” programming but the same rules applied, if you did not understand the process you were trying to automate your program would produce lots of useless output (but do it quickly!). I was always amazed at the number of people that thought that if the output came from a computer, it must be correct.
My first FORTRAN program (1966) was written to run on a DDP-24, but it wouldn’t compile because it exceeded the 24K of memory. It made me a better programmer because I had to learn to use DO loops and subroutines. The punched paper-tape input also forced me to learn BASIC so that I could test all the subroutines before committing to the time to punch up the tape.
I had forgotten Forth – I kind of liked programming in that language. I got a taste of assembler on an HP3000 mini before that, so Forth wasn’t quite as novel as it would have otherwise been. So many languages over a lifetime, but I have to say, you can make a right mess in any of them. Just like you can write really elegant code in any of them. Some of the best designed and structured code I’ve ever written was in Z-80 assembler, I kid you not. The hardest part of writing a program is not the coding; a dunce can be taught to code. The hardest, and most important parts are the requirements and design. If you don’t know what you are supposed to be doing, and/or don’t have a good design on how to do it, then it doesn’t matter how masterful you are at churning out code, it will still be useless.
So many “developers” today startup any new program they’ve written in the debugger because they expect it to fail. I don’t. I expect it to work and I’m always a bit peeved with myself when it doesn’t. Now there’s usually something that doesn’t work quite right, but the number of initial bugs is quite low now. I once told a fellow developer that I do most of my debugging in the design phase. He rolled his eyes, but my code had the lowest number of defects reported by the verification team and I still turned out as much code per month as anybody else.
I agree with your thoughts on code design and I despair of many “modern” programs that are just bloatware and have poorer performance than some stuff I used to run on a CPM card in an Apple ][! With the exception of video and graphics, which need a LOT of memory, most programs don’t give all that much added value over their ancestral roots from the 80s and early 90s. I particularly despair of Labview programs. Although it is barely possible to write good Labview code, you need to have a background in classic coding techniques to even think about them. And don’t get me started on documentation! It’s the definition of spaghetti code, yet it’s all the rage in the instrumentation world. Try V&V with Labview, then take a stiff drink. Any language where all of the global variables and subroutines reside in separate files is an absolute joy to maintain.
My biggest complaint about C++ is how much stuff it does for you. Unless you are very disciplined and spend time thinking about everything that is going on under the hood, you will end up with slow, bloated code every time.
Because of how much it does for you, it is possible to write code faster. Which is why many managers love it. The philosophy these days seems to be that if a program is slow, they can just throw more iron at it and if it uses a lot of memory, then memory is cheap.
It’s very frustrating for someone like who learned on systems with just a few hundred thousand bytes of memory, and more memory cost something like $1000/megabyte.
Also I had to learn all kinds of techniques to make sure my code was efficient. Almost all of those techniques have been built into modern compilers.
Labview is definitely one of my least liked computer languages/environments. It is cryptic, poorly conceived, and difficult to write functional programs in. I also detest Perl, however you can write good code using it which approaches elegance. There are also a lot of good packages available to extend it. The problem with Perl is that it makes it all too easy to write nearly undecipherable crap; and in fact it could be argued that it encourages such coding practices. But I digress. The key is to use the proper tool for the job which is available to you. Unfortunately that is sometimes tools like Labview and Perl.
I agree that spending time thinking about what you want to do is critical, however, depending on how complex the problem you are working on is, it can be nearly impossible to think through all the possible variations and combinations prior to coding and testing.
A couple of years ago I was asked to write a program that could first format and display individual records from an input file, then as a second phase, add a filter that could select individual records. They wanted to be able to do things like ((Field1+Field2)>Field3) & (Field4=”D”)). They also wanted the ability to make modifications to fields, such as right and left trimming, converting to upper or lower case, etc.
First I had to develop a way to describe incoming records so the program could isolate the individual fields. Then I had to design a way to parse then resolve the equation. The parser and resolver had to be able to handle ASCII and binary data, arithmetic, logical functions as well as process the output of the various sub-functions.
I ended up designing a doubly linked list that fully described each element in the record. I went with doubly linked list because I needed to be able walk the list in both directions and a linked list made it easier to remove elements as they were resolved. The parser walked the link to find the deepest elements, then it applied standard precedence rules to figure out the order within a single level of parenthesis.
Applying all the rules, figuring out which functions don’t work with each other, figuring out how to collapse the linked list as the functions are resolved, ended up taking about 15,000 lines. It two weeks to wrote the code. Then I started feeding it simple formulas and gradually increased the complexity. It took another two weeks until I was satisfied that all of the bugs had been beaten out of it. The program has been in use for two years and other than an occasional request for enhancement, there have been no new bugs found.
Agreed. The problem domain is seldom small enough to allow you to consider every possible input, output, and processing path. However the more you can account for in the requirement and design phases, the fewer problems you will encounter later, and the less likely a costly logic redesign will be needed.
I was taught R in undergrad, but I switched to Bash/Awk after finding a good job after school. I taught myself by patience in RTFM. It was quite a challenge and no one expected that from me.
I became that much more efficient than others stuck in Excel, R or Python, I got an award for it 🙂
I haven’t been able to find an easier, terser and cleaner way to manipulate plain financial tables.
I like Linux because it’s a gui wrapped on top of a complete programming environment, rather than a programming environment you have to install on top of a black box gui.
After more than a decade of financial fortune telling (still do it), I decided to put my skills to something else. I hope some people are enjoying mommy’s amateur science hobby other than my children.
Thanks for the post, Willis. -Z
“ it’s a gui wrapped on top of a complete programming environment”
..
No, Linux is not a gui. Linux is not a programming environment. Linux is an operating system. The programming environment runs on top of Linux. The gui also runs on top of Linux. Linux can be run without either a gui or a programming environment.
Then Windows and Mac is just a kernel too. Weird way to see things. No?
I like to include basic user utils as well. Linux uses them when booting into a USEABLE environment.
Current Macs are unix based. Windows is a bowl of spaghetti masquerading as an operating system (which you correctly called a black-box gui.)
Zoe, while we disagree on many things to do with the physics of this strange universe, I totally admire both your persistence, your imagination, and your programming skills. Always good to hear from you.
w.
Been through much the same history on computing. At school I learned Algol and FORTRAN, and we had access to time on the local university’s computer (limited to 1 minute of runtime per programme). Punched out cards with a hand punch. Those doing economics were able to run a macroeconomic model at LSE via a teletype machine: today’s Sim-Econ games are probably much more sophisticated.
My introduction to modelling applications was somewhat different. I was fortunate enough to be employed in looking at some of the implications of the infamous Limits to Growth study, so the next language I learned was DYNAMO (BASIC and 8080 assembler came later, when I got my first computer – a ZX81 which I soon expanded to 16k of RAM in which I contrived to run refinery LP simulations), in which their mode was written. I spent some time picking over its entrails and writing other simultaneous differential equation models both as programming practice and as numerical testing of the tendency of solutions to blow up because of the limitations of rounding errors and Runge-Kutta fourth order integrations. I then helped design (my co-designer was a physicist by training, and insisted on things like dimensional analysis) and did the writing and punching in FORTRAN and the many runs of a resource model. We had the benefit of some expert input from mining and geological and metallurgical specialists. We tried to capture some of their ideas in the modelling. Despite 4 card trays (about 2 ft long each), in reality it was fairly rudimentary. But it was more than enough to teach me that Limits to Growth was a prisoner of its assumptions for data and modelling: there were other answers that were more plausible.
Looking back on it decades later, I am pleased to report that our modelling turned out to be much closer to reality than that implied by Limits to Growth. Perhaps we were just lucky. Perhaps because we didn’t start by assuming the answer and tweaking the modelling to fit. The other great lesson was from seeing modelling used for political purposes. It has made me naturally suspicious of models, and insistent on proper relation to real world physics and measurement ever since. Not long after, I got introduced to some of the math of chaos and catastrophe theory which provided more insight on the need for caution in following models.
who needs GCMs when anyone can model what the IPCC will decide what the models should say
Since most of the models turn into nothing more than y = mx+b equations after about 50 years all the complexity of the models is just wasted effort.
Where y = temperature and x = CO2 concentration 🙂
Actually its y = temperature and x = input forcings in W/m^2
Another big problem , is that if they are hindcasting to say, GISS,
.. they are hindcasting to something that is fabrication in the first place. !
If they reproduce that temperature fabrication,
… they are almost certainly WRONG before they even start
eg, If your elevator shaft is 100m tall, and you give it data that says its 110m tall
Things aren’t going to work very well !!
Might work fine until someone press the wrong button :<)
That’s right, the alarmists are hindcasting the bogus, bastardized instrument-era Hockey Stick temperature profile.
They should try hindcasting for the *real* global temperature profile as represented here by the U.S. regional surface temperature chart, which is also representative of regional surface temperature charts from around the world.
U.S. regional surface temperatue chart (Hansen 1999):
Hindcast this.
About a decade behind you, e.g. PLATO IV, OPM. Modelers best assume Constructal law.
It’s not just computer models.
My favorite paper was one in which a colleague had a strange result. The measured rate coefficient should have been constant but was inversely proportional to the reactant concentration. I thought of a reason why this should occur and a quantitative model that was just basic arithmetic. A computer was only used to calculate the values of three variable paramers for a best fit to the data.
The final equation had k inversely proportional to the concentration and variable parameters that were realistic when fitted to the data. Good enough to claim that we figured out what was happening. The maths of the model was easy to check, just arithmetic that barely filled half a page. That the fit was good was easy to check. It was a paper that would have passed peer review and been accepted as the truth.
But! We could measure one variable parameter using a completely different experiment, so we did, and got a result a factor of ten different. We still published it as kind of close-but-no-cigar paper.
The point of the anecdote is that the model did not have the complexity of climate models. A chemical engineer could have used it as reality, plugging in variables and trusting the calculation. In reality, the mistake would have been picked up earlier but imagine if 97% of chemists said that they were 95% confident in it so ignore those who measured the variable parameter, and those sceptical of it should just shut up? Go to jail for racketeering, even.
Well done Willis. I could follow what you were saying even though I’m not a programmer. I fell in love with Chemistry after getting my first chemistry set in Junior High. Went on to become a high school chemistry teacher. I always included a section of what models are and aren’t so the students wouldn’t get a false impression that models are reality.
Brilliant synopsis.
Combine it with this interview of Judith Curry: https://www.youtube.com/watch?v=BOO9mafcA3s&t=2s
After understanding what you both wrote it makes me wonder how many people have to die when the electric energy system is totally revamped because the models said we had to.
Wow Willis. So cool to read your history. I followed a similar path until I didn’t. 2 10th grade students from every school in the city were chosen after candidates wrote a science/math exam and I got to be one of those in 1964. (Joe Berg Seminars). We got to listen to and meet professors from the university once a week and I found it a lot of fun. When we were invited to the university to learn a bit about computers I was already learning about Eccles-Jordan circuits(inventors of the flip-flop) and those punch cards we stole from the keypunch room were also good for making toy rockets!
I still have my 8k Commodore PET (Personal Electronic Transactor) with chiclet keys in the basement along with an aluminized paper printer and 20 virgin rolls of paper.
First commercial programming was done on Z80 s100 computer using CPM and Bill Gates’ first Fortran compiler to control a wood processing machine. I actually worked on a control system that had that magnetic donut memory so you didn’t have to reload the program if the power died. It came as part of the huge waferboard press from Germany in ’84.
The model of our manufacturing process that I wrote in VBA and Excel showed me how easy it is to fool yourself with these programs. I have never understood how they manage the cumulative errors in iterative climate models. Maybe they don’t?
I do not trust computer models because I worked with computer models going back as far as 1968.
That’s an excellent synopsis of my whole post.
w.
Indeed!
That brings back memories. My first employer in 1980 was Datapoint and the first computer I learned the internal workings of and how to repair was the 2200.
Two years later I joined the Amsterdam Stock Exchange as junior Cobol programmer. Punchcards were still used and very important for the department. They were our mail system, calendar, notepad, predecessor of the stickies, we could not function without them.
“I learned my fourth computer language, MS-DOS.”
..
MS-DOS is not a computer language.
You write commands in MS-DOS language. That’s all it recognizes.
If I entered “Brian is an asshat” as an MS-DOS command, it would not know what I meant.
(on second thoughts . . . )
You really are showing yourself to be monumentally STUPID, brianless. !
MS-DOS is a set of words or commands that is used for communication between the user and the computer.
It is a computer language..
Its main function just happens to be user interface and disc operation instructions.
Ha ha. Of course you’re both right. DOS is a computer program that is both an operating system (“cleverly” masquerading as a set of interrupt calls) and a user interface which includes the ability to run a rudimentary scripting language (bat files). Some people are overly picky, but if someone says they programmed in DOS, I’m going to assume they mean bat file scripting unless they specifically mention an interpreted language like early BASIC running under DOS or an early compiler (or even assembler). The best was Turbo Pascal. Most C compilers were agonizingly slow and they had a bad habit of having wayward pointers that could destroy the operating system, require a reset button or the infamous three fingered salute to recover.
One of the reasons I liked Pascal was that it forced a discipline that increased the chances that the program would actually do what you wanted. The fact that you had to define things before using them meant that it was a 1-pass compiler, vital when processors ran at the rate of a few MHz. And it also meant that the compiler and you had to be on the same page as to what variables meant. That confusion and the insistence on strict type checking meant that you tended to avoid the spectacular explosions caused by sloppy C code. Object Pascal’s fingerprints are all over C++ and although you can be sloppy with it, if you write programs as if they were in Pascal, you will be rewarded in the end.
One thing I just remembered. If you watch the original Terminator movie, the world from Ahnold’s point of view always has computer code scrolling by. I recognized where that code comes from because I once had to write subroutines in it to control lab equipment that was called from UCSD Pascal. It’s 6502 assembler code from an Apple ][, a computer not quite up to the job of controlling a Terminator.
Ah, the old Apple ][e. Did some early programming on them.
The “peak and poke” were fun because I could build little interface boxes to control external machines.
Made a cute little car that could follow a curvy line on a sheet of card.
Hmmm . . . core memory invented by Dr. An Wang, founder of Wang Laboratories.
You left out mention of the cert from Aames, and your stint at Sonoma,
Hey, I left out most of my life … you could start with this, more here.
w.
I thought is was a blog about climate, not the life and times of a wanna-be computer programmer.
As usual, what you think and reality have little in common.
The point of the article, which you once again go out of your way to miss, was about why climate models range from bad to useless.
The life history was to show why Willis is qualified to make such judgements.
Well put, MarkW.
You are being very reasonable. I fear you may be wasting your time though, since Brian seems intent on disrupting the conversation with his angry comments, regardless of topic.
BJ you are such a blowhard.
Yet brian is making it about brian being a wannabe something.. anything!!
… and its failing. !
You are still an abyss of empty blah !!
Your facade of egotistical bravado is hilarious. 🙂
You really do have deep-seated emotional and mental issues.
Pseudo intellectual is becoming a synonym for progressive.
Brain-dead, read the “About” link at the top of the page. This blog is about much more than just climate.
Brian, you are getting tiresome. Are you a thirteen year old that is trying to start a fight from a safe distance? Your constant rudeness is just ridiculous.
“Members of his class of models are notoriously cranky, unstable, and prone to internal oscillations and generally falling off the perch. “…..especially when their predictions tell you to bring an umbrella to work.
Do you think climate models are skillful?
Define “skillful”.
w.
1) https://dictionary.cambridge.org/dictionary/english/skillful
.
2) https://scienceblogs.com/stoat/2017/01/11/climate-models-have-proven-extremely-skillful-in-predicting-the-warming-that-has-already-been-observed
stoat ! ROFLMAO
The very bottop of the fetid abyss when it comes to anything related to science or morality.
You really know how to dig deep in to the putrid slime !.
Climate models have proven extremely skillful at reproducing the bogus, bastardized, instrument-era Hockey Stick chart profile.
That’s because the climate models were tuned to reproduce the bogus Hockey Stick chart “hotter and hotter and hotter” temperature profile.
You have a bogus Hockey Stick and you have bogus Climate models that reproduce the bogus Hockey Stick profile and it’s all dishonest computer manipulation of the data.
The real temperture profile of the Earth doesn’t look anything like the bogus Hockey Stick profile.
The real temperature profile of the Earth is based on actual temperature readings taken by human being over many years.
The Hockey Stick chart and the Global Climate Models are nothing more than the programming of computers to reach a certain outcome.
A huge fraud has been perpetrated on the world by alarmist computer data manipulators. That’s what we have with the bogus Hockey Stick chart and the Global Climate Models that aim to duplicate a false temperature history.
NO
They are not testable in practice, because Climate change is not a 30 year period phenomena, just because a bunch of weather record editors, with nothing but current data and altered record declared it to be so.
Earth itself does not work in such corrupted ways to make a crust.
In geology unambiguous climate change that we know of (and there is no other we know of, btw) is on a ~500 year time-scale resolution.
The LIA cycle was detected on a similar time scale.
All else is the product of unscrupulous corrupting of recent more detailed records.
While everything shorter in time-scale on the observational level is the net of weather cycle noise.
So the question of model predictive ‘skill’ is pure nonsense as applied to climate models, as they have no skill which they can demonstrate without recourse to a ‘time machine’ to sample a trend in 500 year future sample increments, which would negate the need for a GCM prediction in the first place.
Blah blah blah ‘skill’ blah blah blah.
Stop kidding yourself Brian, you fool no one here but yourself.
40 years worth of wrong predictions show that climate models have no skill.
One of them demonstrably does have skill, the Russian one. The problem is they take their one good proven successful model and then for some unaccountable reason mix it up with ones proven to have failed (by averaging their results). The result is they have degraded the one model that is probably fit for purpose.
If this were medicine, drug A, the equivalent would be saying we have 50 different models of the effects of large scale administration of drug A on the target population.
Some show mortality of 60%, some show cures of 70%. None come very close to predicting the results of previous real world trials. There is one model and only one which has successfully predicted the results of previous uses of A in this context, and it shows mortality of 20% and 5% cures, which tallies pretty well with field experience.
So what we do is average those results with all the ones that have failed, and tell our policy officials that this shows what is needed is large scale immediate dosing of the entire country…
Well, maybe I am missing something. Very much like to know what.
“Do you think climate models are skillful”
MOST CERTAINLY NOT !!
They are totally lacking is SO, SO MANY areas, and can’t even hit the side of a barn with their scatter gun. !!
They barely reach the level of a low-end computer game
Willis, that was a trip down memory lane for me. My main thing in college turned out to be math modeling of all sorts (an example, the classical predator prey equations done three ways: in calculus, in numerically solved Fortran, and via probabilistic Markov chain matrices). My senior/PhD thesis (separate story) input/output (I/O) dynamic model of nuclear power was written in Fortran on Hollerith cards—two big boxes holding about 1000 cards. And yes, debugging was a real PITA. Bonus was that Vassily Leontief won his Nobel prize in Economics for I/O that year, and as a prized pupil I got to celebrate with him.
You DO get around. When I was the first global sr partner head of BCG’s then new ‘Time Based Competition’ operations practice in the late 1980s, I bought Stella for each office’s TBC team. Stocks, flows, converters, connectors. Visually simple so our clients could understand our work and recommendations. Got to know creator Barry Richmond very well during his time at Dartmouth. He used our stuff (disguised) to market Stella at his High Performance Systems, and he brought us new clients. Win-win.
Thanks, Rud, your stories and insights are always welcome. Stella was an awesome language for constructing models of many things. Is it still around?
w.
I think I’ve been programming computers even longer than you, Willis. The first ones were IBM plug boards. But throughout my history with computers, one phenomenon stands out: people trust whatever they say. An example. I wanted a bank loan for a startup (involving computers, of course), so I did a simple model showing projected revenues and expenses, with the revenues exceeding the expenses (of course). I expected the bank manager to grill me about my assumptions. Never happened. He said, “This was done by a computer?” I didn’t correct him that it was done on a computer. Then he said, “It must be accurate. How much do you need?” I wish time had eroded this blind trust in electronics, but given the fealty paid to models–whether climate or COVID–I remain disappointed.
“The first ones were IBM plug boards.”
I learned how to program IBM plugboard machines in college though never actually did it for a living; IBM Accounting machines, Collators and Reproducers and of course keypunch and sorters which were not programmable but necessary to feed the beasts with cards. Those plugboard machine could be pretty cantankerous things to program but it was was fun when, with enough plug wires and patience, you got them to do a complex job correctly.
Regarding core memory: about 1965 or 66 as a kid I went on a tour of IBM-Boulder (Colo.) which included a walk through the huge room dedicated to fabricating magnetic core memory planes. I remember row-after-row of people sitting at benches and peering through magnifiers to thread the x, y, and sense wires by hand. Those hundreds of kilobytes of main memory were exorbitantly expensive.
The first thing I noticed in the core store memory photo was the lack of sense wires. How did the one shown sense if the polarity changed on a ‘read’?
FYI, in the Oz navy I maintained an Oz-designed anti-submarine system (Ikara) that used 32K core store memory to track submarines and guide a missile-mounted torpedo over same. It could process inputs from radar, sonar and external data links (other ships and/or helos).
Amazing what could be programmed into such a small amount of memory.
Booting the system required switch inputs using octal numbering for the code, then a magnetic tape. Lots of wire wrap backplanes as well.
Memories!!!!!
I was puzzled by the lack of sense wires in the photo also.
Sounds a bit like Apollo programming.
Spent about 10 years running an HP-1000 RTE-VI data acquisition system, it also had a 16-bit front panel switch on the CPU. Had the switch setting memorized that booted it up from the disk drive. Wrote many lines of Pascal for it.
Yes, the first thing I also noticed was the missing sensing wire that ran diagonally through the ferrite donuts. Maybe the photo was taken during manufacturing before the sense wires were threaded.
The sense wire is probably on the other side of the circuit board
>>
The first thing I noticed in the core store memory photo was the lack of sense wires.
<<
Core memory I studied usually had four wires–two for half-select, one for inhibit, and the sense wire.
Jim
There may have been a scheme that fished the polarity changes by, umm, “back EMF” on the X and Y wires. That could have been a big win, as people may never have gotten good at automating threading the sense wire.
DEC hired skilled seamstresses to “sew” their core planes.
Likely boring work that paid well.
Wow, great essay! I’ve wanted to get some understanding of climate models and what’s wrong with computer models in general. So now when some bozo tells me that climate science is settled because of definitive climate models- I’ll tell them to read your essay.
Enjoyed this a lot (maybe in part because I agree with it…!)
You allude in passing to something that has always troubled me about the models and particularly the spaghetti graphs.
It cannot be right, surely, to take a bunch of models many or most of which are demonstrably failing, and then average their forecasts and claiming that the result has any kind of validity.
If this were medicine, for instance. We had a drug whose function we did not understand, made a bunch of different models of its working, then generated forecasts for treatment of the target population. The results vary from 70% cured to 60% dying. So we average them and say on balance it is forecast to be significantly beneficial, lets go?
Or if we were designing a bridge. The results are all the way from, at a given sample set of girder parameters, all the way from failing under load once every five years to failing once in a hundred. We then average them to find out what girder parameters we should be using?
I would love to get an answer on this from someone who understand the subject better. It looks to me like a totally insane and unscientific procedure, completely unfit for purpose of generating forecasts to be used in policy selection.
But maybe someone knows different. I am very willing to be persuaded, but so far have not only found no satisfactory explanation of why averaging the bad with the good makes sense, I haven’t even come across any explanation of any sort.
Surely the only rational way is reject the failing ones, stick with the better ones, until finally we get to one decent one. Then use it.
Probably, from what I read, if we did that we’d end up with the Russian one, and all the alarm would evaporate.
Or if we were designing a bridge.
Idiotic things happen with “best models” 😀
On Christmas Day 2003, everyone was already about to put the shovel and concrete machine in the corner and ring in Christmas Eve, when a final routine check by the construction management took place.
They found that there was a difference of 54 centimeters between the bridge construction on the German side and the construction on the Swiss side. If one wanted to join the bridge in the middle, it would be difficult. “Laufenburger we have a problem” radioed the construction management to the headquarters.
Reference horizon
“The height difference can be corrected with minimal effort during route construction on the German side,” it was appeased. But clearly pointed out, “The fault is on the Swiss side.”
The overall project manager Beat von Arx also admitted this in no uncertain terms. However, he too did not know why at first.
Later, the public was enlightened. The cause of the error, he said, was the fact that in the area of road and bridge construction, the horizons on the German and Swiss sides were based on different reference horizons.
With it one did not say with a technical phrase the following: Germany refers with all these calculations to the sea level of the North Sea. Switzerland, which obviously prefers to look south, takes its reference from the Mediterranean Sea. The layman learns: Sea level is not equal to sea level!
Minus became plus
But that is not all. This difference between the two reference seas results in a difference of 27 centimeters. This was known when the bridge was planned 12 years earlier and included in the calculations accordingly.
Unfortunately, someone must have made a minus sign out of a plus. Because the values were corrected to the wrong side. And so, in the end, the two bridge sections differed by 27 times 2, which equals 54 centimeters.
Translated with http://www.DeepL.com/Translator (free version)
German source
Wonderful. If it had been climate science… what would they have done? Split the difference maybe, and had a 13.5 bump?
Climate science seems to have problems with sea levels 😀
Logically, there can only be one best simulation. If one averages that with all the others, then the result will be a degraded result.
Yes, that is exactly what I would have thought, and I am desperately though vainly seeking some explanation of why perfectly intelligent people with advanced degrees and lots of publications do not also see it that way.
I remember visiting the Smithsonian museum in D.C. decades ago, they had a Cray X-MP on display. It look like a modernist bench for people to sit on, it was comfortable…
If you are receiving a paper check from the U.S. Treasury from the recently passed stimulus bill, you’ll note it is also: “ 7 3/8 inches wide by 3 1/4 inches high ”
And $600 less than you were told you were getting?
I wonder why none of the usual trolls have shown up to declare that anyone who doesn’t trust models is a science denier?
Yes, where is Stokes when you most expect him?
Haven’t seen him for awhile. Maybe he lost his funding.
>>
We lost a Mars probe because someone didn’t convert a single number to metric from Imperial measurements … and you can bet that NASA subjects their programs to extensive and rigorous V&V.
<<
I wish you wouldn’t repeat this inaccurate view. This is from NASA’s Climate Orbiter Failure Board report released Nov 10, 1999:
The board’s report cites the following contributing factors:
The units problem–pounds force vs. newtons–is not an order-of-magnitude difference–only 4.45. And why didn’t they include the units like “good engineers?” Was the communications channel so limited that they couldn’t add a “nt” or “lbf” on the end?
Then there’s this from an IEEE Spectrum article: “. . . JPL’s process of ‘cowboy’ programming, and their insistence on using 30-year-old trajectory code that can neither be run, seen, or verified by anyone or anything external to JPL.”
Also from that Spectrum article: “. . . somebody at JPL ran the data through the 1998 Mars Pathfinder navigation code (different from the Mars probe code). It showed the spacecraft was off course by hundreds of kilometers, which turned out to be correct.”
Jim
How about the 1202 alarm on the LEM “Eagle” (apollo 11)
.
https://www.discovermagazine.com/the-sciences/apollo-11s-1202-alarm-explained
Total non-sequitur, Brainiac. The 1202 alarm had nothing to do with bad engineering, rather it was tribute to what they accomplished with the incredibly minimal computing hardware.
The 1202 alarm just indicated that the computer was being asked to do more than it could do, in the time allotted. The list of things to do was ordered by importance. The engineers determined that the things that weren’t getting done weren’t critical. Which is why they advised that the landing could continue.
I consider Willis’s view quite accurate. You only give eight reasons why the error had not been discovered in time.
>>
You only give eight reasons . . . .
<<
I’m sorry. I’ll try to do better in the future.
Jim
With all that wonderful computer experience, you failed to mention Unix, Linux, or any of the other variants.
As you may know, Brian, Linux and Unix are operating systems, not programming languages. Sort of like WE’s MacOS on which he now runs the R programming language. An attempted derogation FAIL?
Eschenbach thinks MS-DOS and CP/M are programming languages.
If you stretch the definition of “programming languages” then shell scripting, awk, etc., which all come with Linux allows one to “program” 😀 😀 😀
How come Wills didn’t mention java?
Brian, I talked about the languages I know and have used. It’s not a history of computers. I don’t like Java and don’t speak it.
And since I used to drive my computer around using CP/M, I called it a “language”. True, it’s an operating system … but it contains a series of commands that I used to make a computer do what I wanted it to do. I call that a “language”.
Don’t like it?
Sorry, don’t care. Brian, all you want to do is come along and pour meaningless vitriol on what I’ve written. I note that you don’t say a damn word about the real subject here, the problems with computer models. All you want to do is nitpick and stick your nose up into the air and tell us that you are a “real” programmer. Ask me if I care. Lead, follow, or get out of the way.
w.
” I call that a “language”.
Don’t like it?
Sorry, don’t care. Brian, all you want to do is come along and pour meaningless vitriol on what I’ve written.”
.
.
If you wish to do “science,” you need to use precise terminology. Obviously you “don’t care” if you misuse the terms. CPM is not a language.
Base on this posting, and your comments, you qualify as a computer “hack” (and not the derogatory meaning of that term.) You “think” you know about computers, but you don’t. This whole piece is a poor attempt to bolster a CV lacking in real world experience writing software.
.
Your skill set is equivalent to that of a “script kiddie.”
There are words for people who spend all their time whining about minutia while ignoring the bigger issues.
That word isn’t expert.
As usual, Brian tries to force everyone to go down the rabbit hole of his choice.
“There are words for people who spend all their time whining about minutia while ignoring the bigger issues.
That word isn’t expert.”
One of those decriptive words is “troublemaker”.
I don’t think this particular troublemaker is whining about anything, he is deliberately attacking the character of Willis with his nitpiking.
It’s a pathetic performance. It probably won’t end well for troublemakers.
And your knowledge is equivalent that of a limp lettuce leaf. !
Willis is a few dozen MAGNITUDES above anything you would ever be capable of.
Why do you have such a huge ego?
No one has an ego bigger than Willis’s
Nope. Based on this thread, you do.
There seem to be lots of big egos here today. So, if I may make bold a bit myself, I’d like to ask why a person has to buy a $700 Mitituyo digital caliper, and pay another $300 or so for the data cable and software for it, instead of being able to buy a $10 cable and maybe another $30 for software to use with his $30 digital caliper with the unsupported data port?
With so many coders here, surely someone has an explanation?
You have obviously never reloaded competition precision ammo. I have, for over 40 years.
Why “obviously”? And what does your comment have to do with my post or the questions I posed? Don’t be shy. If you have some information to share, shoot!
I don’t even understand what you mean by ‘reloaded competition precision ammo’. I’ve reloaded Lake City 30-06 brass, and lots of it. That’s competition brass. Does reloading it make it “precision” ammo? What do you mean by ‘precision’ anyway? I use ‘competion’ dies, and weigh each charge to a tenth of a grain, using a Bonanza Model “N” scale. I chronograph all my loads before taking them to the field, and have worked up some pretty hot loads without blowing up a single firearm in 35 years of reloading.
But my loads are intended to stop a charging grizzly primarily, and secondarily to cleanly take an elk, mulie, or whitetail out to 400 yards. Maybe that’s not “competition precision” in your book, but the precision I strive for can mean the difference between being dead or crippled/disfigured for life or walking away with a nothing more than a good scare.
“Competition ammo” is certainly more precise, under ideal conditions, than what I load. But I wouldn’t want to take on a bear, even a black bear, with it.
I’m sure you know all that, if only from shooting buddies who hunt dangerous game or in their vicinity. So what is your point?
Is it that non-competitive shooters shouldn’t/needn’t reload? Or if they reload they should just use a dipper for the powder and only employ once-fired brass?
Given the high praise Willis has bestowed on your pronouncements above:
I am surprised and disappointed…
Well, I have never shot a charging grizzly at 40 yards. Hope never have to.
But shot plenty of just 10x rings at 200 meter rifle multiple positions, plenty of 10x rings at 100 meters two hand pistol gun, and lots and lots of +96 trap, skeet, and sporting clays 4x rounds (each box of shells is 25, so a standard comp round is 4 boxes)—all with handloads.
Neither have I. But I’ve read and thought enough about this contingency to know better than to shoot a grizzly at either 400 yards (I had a magnificent blond one broadside in my sights once, and a tag in my pocket) or at 40 – better wait until the distance is 10 yards, for a better chance to break its shoulder. But you didn’t answer my question above – unless this post is meant to be an apology…
BTW – I would use the same strategy with a charging Black Bear, except at 40 yards it might be worth bluffing it by moving towards it aggressively. This worked for me once when I was armed with only a crossbow and a knife, and had no bear tag.
That’s not a question to ask of “coders.” However, you are more likely to get the correct answer from some of us than you will from any modern “economist.”
Simply put, not enough people want what you want to make it worthwhile for someone to go to the effort of supplying it.
So it’s worthwhile for factories in China to put a serial port on millions of cheap digital calipers over a period of 20 or 30 years although nobody in the world has any interest in using them?
My guess is that in China and other ‘poor’ countries you can buy the cable and the software to utilize these ports, and the manufacturers have marketing agreements that keep those elements out of the hands of North Americans so they’ll pay through the nose for the pricey Japanese or American products .
Some of my cheap Chinese calipers even came with a drawing of the cable in the instruction brochure. I believe Hornady sold one, and simply said “no” when I asked to buy a cable. I also tried Faxing the Chinese factory after buying my first one, but got no answer.
But that still wouldn’t explain why nobody has offered a simple program and maybe a cable schematic so the tens of thousands of these calipers sold here can be used efficiently.
20 years or so ago I read of a group of American archeology students who circumvented the high cost of the Mitituyo cable and software by coming up with their own home brew of the latter. So this is not a recent development, but it seems to be getting worse.
The software you seeK exists:
https://www.greatgages.com/collections/software-wedge
Well, the only cable they offer at $145 for their Lite software is for a Mitutoyo caliper as is their $241 RS232 to USB adaptor.
So – overkill and overpriced, and won’t work on my calipers. I think I’ll have better chance of success by contacting the guy at robotroom.
Thanks for your help.
Some people need the extra precision that can only be provided by the more expensive instrument.
It’s obvious that you have no need for that kind of precision.
For myself, I have no need for the kind of precision that the $30 caliper provides, so I haven’t bought one. However I don’t go around wondering why people feel the need to buy calipers in the first place.
I find it a bit arrogant for someone who can’t imagine why one would need a caliper to be passing judgments on their precision.
But here’s a freebie – I’ve used mine for measuring various dimensions of brass rifle and pistol cartridges, clutch flywheels, brake discs and pads and shims. I do have micrometers for more precise measurements, but they’re slower to use, cost more, and are overkill for the above purposes.
As for precision, I’ve always owned at least two calipers at a time, and usually three or more. It’s part of the joy of owning inexpensive measuring instruments that you can own several and check them against each other for accuracy.
So refusing to believe as you do proves that Willis has a big ego?
Really?
I would say that declaring that anyone who mustt accept your opinion as the last word on everything, indicates an ego big enough to swallow light.
Brian’s opinions are based on anti-knowledge,
His egotistic blah is a facade to hide his deep-seated mental insecurities and ignorance.
Classic projection on the part of Brainiac.
At least Willis has something to back it up.
You are an ABJECT FAILURE at everything you have posted. !
Your egotistic blathering is a facade based on trying to hide your IGNORANCE..
…. but its not working
There’s a big difference between you and Willis, with the conclusion that you come across as a complete jackass. But to explain; Willis appears to be a polymath, self-taught and capable of original thoughts and structured thinking. His analyses are fact based and challenge the inbuilt flaws of climate models. On the other hand , you just parrot what your gods say without showing a shred of understanding or thinking.
I think he has an anger problem.
Tell us [TINU} what your hat size is, Brainiac.
To ask for Java wouldt have been correct 😀
He’s mentioning the computer languages he’s worked on.
Why should he spend time talking about operating systems he may not have used?
Or are you just trying to show the rest of us how bright you are?
He posted: ” I learned my fourth computer language, CPM”
.
CPM is not a computer language.
It is plainly obvious to computer professionals that Willis is trying to “show“us how bright he thinks he is.
Hello pot?
I doubt Brian has sufficient self awareness to realize he has just been insulted.
Willis is demonstrating why he has the background to make the judgements that he is making regarding models.
He never claimed to be a world class expert on all things computer.
That’s been your role.
That is YOUR modus operandi..
FAKE bravado, backed by an empty mess.
And it is FAILING BADLY
You are coming across as a ignorant and clueless moron.
The difference is that Willis doesn’t feel the need to show us anything. He just lays out the facts.
Brian is the one who feels the need to demonstrate his mental superiority. To bad for Brian that he has never succeeded in demonstrating his mental adequacy, much less superiority.
Willis does not have to try to show how bright he is. It is obvious from his work. You, on the other hand are trying to dig deeper into a pit. Oh, you are successful at that. For a minute I though you were not capable of anything. Just do not wear out the shovel.
How typical of Brian, when caught in a lie, he changes the subject.
WRONG AS ALWAYS.. you base level ignorance is showing through more with your every post.
It contains a set series of instructions that you have to learn the meaning of to communicate with the computer.
If that isn’t a “language” then you have a really weird and twisted idea about what a “language” is.
Your puerile grunts and groans just won’t cut it. !!
Brainiac is an absolute expert on absolutely everything, and needs the entire world to see this.
Might have to rename him ‘Lawrence of everywhere’
Meanwhile he declares that Willis has a big ego because even after having reality explained to him twice, Willis still fails to agree with Brian.
Yet any Unix variant has stuff like c, lex, yacc, apache, etc built-in. Is it an OS or a language?
I studied computer science in high school in the early-mid 1970s. There were 6 students in my first computer programming class. We had a couple of IBM card punch machines in an oversized closet for preparing our assignments. Once a week, one of us would take a box or two of punch cards to the local bus station for shipment to a nearby university that had an IBM mainframe. They would run a batch job of our cards, and about a week later we would have a printout. Fun times! From there to a university degree in math and computer science, and work as a professional programmer starting in 1980.
What I’ve learned, forgotten, then re-learned over the years is that no computer program can do more than what it is programmed to do. This applies to database programs, graphic design applications, games and, of course, models. But people don’t seem to grasp this basic fact. As Willis points out, all too often we hear that a model proves such-and-such. Sorry, no model can prove anything. Models say what they are programmed to say. We see this with climate models and CV-19 models when a scary scenario is required. And when the scary thing fails to materialize, they just move the forecast of the scary thing further into the future.
The trouble with computers is that they do what we tell them to do, not what we want them to do.
I’ve been requesting a DWIM op code from chip manufacturers for decades now.
(Do What I Mean)
Oh, a computer can do many things other than what it is programmed to do. That’s when you call in the hardware techs to figure out what is wrong. I recall a long day trying to figure out why the paycheck printer was suddenly printing gibberish. Nothing wrong with the computer or its software. Nothing wrong with the printer. Nothing wrong with the signature box between the two. Hardware guy walked in, walked out, walked back in with a USB cable he pulled out of someplace else. Voila! (Nothing visibly wrong with the old cable, either – I did check that.)
Even in modern systems, a random cosmic ray, or an intermittently bad transistor among the many billions, can waste days.
“Everything that can go wrong will go wrong.”
Commentary: Murphy was an optimist.
Back in the early 1980s, the tech company I worked for had a hardware/software installation in the CAD department of a large manufacturer. The issue was the computer would crash at roughly the same time each morning, just as the designers were starting their day. It turned out the boss had his own coffee maker in his office on the same circuit as our hardware. When the boss started the brewer, it caused enough of a draw and/or spike on the circuit that our hardware would crash. I guess the boss had to resort to vending machine coffee after that. 😉
“A computer model is nothing more than a physical realization of the beliefs, understandings, wrong ideas, and misunderstandings of whoever wrote the model. Therefore, the results it produces are going to support, bear out, and instantiate the programmer’s beliefs, understandings, wrong ideas, and misunderstandings. All that the computer does is make those under- and misunder-standings look official and reasonable. Oh, and make mistakes really, really fast.“
Lol! Even though I am only a retired lawyer and NOT a math whiz of any stripe, I have understood the principle very well from my years of experience with clients who tried to show me their economic models, or engineer clients who thought they had the best idea since sliced bread. All I had to do was ask them about the assumptions they made in concocting the “model” or the idea. Usually, it blew the model out of the water, making it what I feared – a pipe dream. The saying, “garbage in, garbage out” was also one I was continually reminded about when I had to confront the subject of “climate change” with my environmental clients.
When you write about a subject that interests me (and this one truly does interest me, for I like you remain curious in my old age), I always learn something new in a very understandable way. So thanks very much for this particular piece.
Thanks much for your kind closing words, Larry, it’s the kind of thing that keeps me writing. You are exactly the person I write for, someone I describe as the “interested layperson”. A woman or man who is interested and curious about the world, but doesn’t have perhaps the relevant scientific training or the relevant math skills to understand what interests them. I write for those people, so they can follow their curiosity and continue to explore and understand this magical universe.
All the best to you and yours,
w.
Another interesting post from you Willis. Keep ’em coming.
I just have one disagreement with a phrase you used –
“stay safe in these curious times”
I say we are living in INCURIOUS times.
People just accept at face value everything the media spews forth.
Hi Willis (and Larry in Texas)! I respect highly intelligent people who admit to having limited formal backgrounds in physics and chemistry. I can send you, on request at my email address rtaguchi@rogers.com , pdf Attachments summarizing basic Atomic and Molecular Spectroscopy, with links, to get you up to speed in understanding the quantum mechanical models which accurately explain the observed infrared (IR) spectra from which we calculate climate sensitivity (before feedbacks).
From 1967-1971, I was a grad student at the University of Toronto under Prof. John Polanyi who deservedly won the 1986 Nobel Prize for Chemistry (see https://en.wikipedia.org/wiki/John_Polanyi ). I thought up, executed, and wrote up a short experiment on the IR emission from vibrationally excited HF molecules (HF’) formed during the inelastic collision between electronically excited mercury atoms (Hg*) and ground vibrational state (v=0) HF molecules. See https://www.osapublishing.org/ao/abstract.cfm?uri=ao-10-8-1755 (H. Heydtmann was a visiting professor from Germany).
The relevance to climate change is that 667 cm^-1 IR photons emitted from a 288 K Planck black body (the Earth’s surface) are absorbed by v=0 CO2 molecules which are boosted to the v=1 first excited bond-bending vibrational state.
If these excited CO2′ molecules simply re-emitted IR photons, there would be no net warming of the atmosphere.
However, during inelastic collisions with N2 , O2, and Ar molecules that constitute 99+% of the dry atmosphere, the excited CO2′ molecules can transfer their vibrational energy to translational and rotational motions of the departing air molecules.
Because N2 and O2 are non-polar diatomic molecules, and Ar is monatomic, they do not possess changing electric dipole moments and therefore do not emit any significant amount of IR energy. So the extra translational and rotational energy ends up after many more collisions shared among many molecules, and the troposphere will have warmed up. This is the true mechanism for the greenhouse effect, which came to me in a flash when I was a junior in undergrad Chemistry. But I thought at the time that this was so obvious I did not pursue it further.
Prof. William Happer of Princeton is correct when he says that the CO2 IR lines are all saturated from the v=0 ground state (so doubling CO2 will have no effect). However, the two small pockets between his red and black simulated spectra involve absorption from the v=1 first excited state to higher energy vibrational states. Since at 15 Celsius only around 3% of all CO2 molecules are in the v=1 first excited state, the lines from the v=1 state are not all saturated, so doubling CO2 does result in a small amount of extra net absorption which explains climate sensitivity.
.
Thanks for that, Roger. That’s the first technical critique I’ve seen of Dr. Happer’s research.
Granted it was a small point, but I’m glad to see Happer’s work being mentioned.
Dr. Happer’s research is game-changing if confirmed. If confirmed, Dr. Happer’s findings mean we don’t have to worry about CO2 and we don’t have to shut down fossil fuels and our economy to save the world. The world will be just fine without our doing all that.
Many years ago I was a young mechanic apprentice. Early 80’s. The shop I was in bought a $40K computerized diagnostic machine.
I thought they were stupid. The machine didn’t do anything that our tune up guy couldn’t do.
But it had a cool printout. When the customers were presented with said printout saying “Based on the test data and Bruce’s expert opinion we recommend……” they almost always said yes.
Turns out those flashing lights and stupid printouts paid for the machine in 3 months. No different information, but the computer said so.
Couldn”t help but notice how none of the recognizable climate alarmists that regularly post here have written in telling how they have a similar history with computing. Maybe learning about computing when you needed to know what was going on inside the computer in gory detail gives you a healthy appreciation of the fallibility of computers and computer models.
Despite being over a decade younger than Willis, I learned programming (Fortran) on an even older computer, a CDC 160A with 8K words of magnetic core memory. I had the opportunity to build a very early MITS Altair for the local community college in 1975 – one of the very first PCs. I programmed it, which consisted of setting switches to set operations and then hitting the load switch to load the instruction into memory. It beat the CDC running the same program. However, seeing how difficult it was to program, I concluded that PCs were going nowhere. Mere miles away, at almost exactly the same time, Bill Gates came to a different conclusion and began work on Altair Basic, Microsoft’s first product..
Well see my post: I’ve got years of computing experience… including systems programming, maintaining operating systems and designing editors and working on early word processing systems…
And all of it totally WORTHLESS in your hands..
An unfortunate lapse of tone. Degrades the forum to the kind of mindless abuse you find on Ars Technica when someone departs from the Party Line. Griff just thinks differently. You will not change his mind like this.
As Sai Baba said:
<i>Before you speak, ask yourself: Is it kind, is it necessary, is it true, does it improve upon the silence?</i>
yawn
Yes, griff does debase the forum.,. always.
Let me say something about griff and other folks. One of the worst parts of the climate world is the endless ad hominem attacks which, sadly, come from both sides.
In Latin “ad hominem” means “to the man”. In discussion, an “ad hominem” argument means attacking the person putting forth an idea instead of attacking the idea itself.
As a perpetual target of such attacks because of my total lack of formal qualifications and certificates, let me say that they don’t harm me, they harm the reputation of the person making the attacks.
So this is a request that everyone here, but particularly the skeptics, avoid such attacks. If griff or Brian Jackson or anyone presents ideas that you disagree with, attack the ideas. But don’t attack the person.
Note that there is a difference between making a personal observation instead of a discussion of ideas (which is an “ad hominem” argument), and making a personal observation in addition to a discussion of ideas (which is not an “ad hominem” argument. For example, in addition to Brian’s ideas being largely wrong, I find him to be unpleasantly aggressive …
So a plea: let’s treat griff and all of those who disagree with us with respect.
Finally, when I was a kid I used to think my dad was saying “Even a blind hawk will find an acorn once in a while”. I spent years wondering what a hawk would want with an acorn, and it wasn’t until I was an adult that I realized he was saying “hog” instead of “hawk” … the old adage means that even someone who comes up with lots of wrong ideas may indeed come up with a good idea.
And this is the final reason why an ad hominem argument is meaningless—this time, for a change, the person might indeed be right.
Best to all,
w.
Sorry Willis, i can’t go with you on this one. Griff is not presenting alternative ideas, he’s making drive by shootings. While we can never truly know someone’s motivations, it is very difficult to believe after years of observation that his purpose is anything but disruption.
Komeradecube March 14, 2021 8:11 pm
Responding to drive by shootings with ad hominem attacks is still a mistake.
Plus, in this thread, griff hasn’t done any drive by anythings. He’s actually just contributed to the conversation like anyone else.
And responding to that with personal attacks makes you, in this instance, worse than him.
w.
Yes, this is absolutely right.
The point is not how anyone feels about a comment.
I would say to Fred250: How you feel about Griff or anyone else is irrelevant to how you should respond. The discussion is not about you and your feelings. Its like conducting a loud conversation at a play, its just getting in the way. You may feel Griff gets in the way, and maybe he sometimes does, but that is not a reason for starting up yet another irrelevant off-topic conversation.
We are not participating, or should not be participating, in these discussions in order to express or relieve our feelings.
We should be participating in them in order to make some contribution to a pleasant and sensible if sometimes hard-argued discussion and increase our understanding.
The right thing to do with a ‘drive-by’ is simply ignore it. Because responding in kind is going to impair the whole discussion.
You can see where this ends up if you visit the Ars Technica comment pages on some controversial topic. In no time over there you have a chorus of people frothing at the mouth over Trump and heaping abuse on people they accuse of being politically incorrect, without regard to the nominal topic under discussion. Or posting ridiculous huge graphics which I have not figured out what they mean, but they are evidently part of the same pattern.
It prevents any sensible discussion of any topic. The forum becomes unreadable. This is why people should not do it. We are not interested in how people feel about Griff or anyone else. That is not what we are here for. The best advice is, people should keep their feelings to themselves. If they cannot control themselves, stop reading.
Watts is unusual in having very light moderation and allowing a wide range of views. But for this to be possible, and the forum still readable, people have to control their posts to being points of view on the subject. Otherwise readability will eventually require heavier and heavier censorship.
Willis’ advice is right. Quote the words, and direct our comments to them. And keep it topical, not personal.
Hear, hear! And I hope your friend Rud is “listening” when YOU speak.
Brings back memories.
Our punch-card programs had a thousand cards, in huge great trays – and woe betide the fool who dropped a tray. (Program destroyed.)
And the Datapoint 2200 was a wonderful machine with removable storage disks (about 35 cm across).
Were you responsible for the program which output an image of a naked lady on the printer? I also seem to remember a rudimentary ‘space invaders’ game.
RE
I was told by a friend who worked in the computer lab, that the admins decided to implement a page limit per job after one programmer submitted a program that had an infinite loop that had only one command in it. A form feed that was sent directly to the printer.
I’m told the paper arced half way across the computer room before touching down. An entire box of paper was exhausted before someone could hit the kill switch.
The offending programmer was immediately hauled into the computer room, given a trash can and instructed to clean up his mess.
Perhaps you mean this?
https://en.wikipedia.org/wiki/Star_Trek_%281971_video_game%29
I recall it finding its way onto mainframes at work not long after. Some people got through roll after roll of thermal paper on their office computer terminals. CRT screens were at least a little more discrete, but the cost in time wasted must have been monumental.
yes… seeing someone drop a stack of cards had a horrible fascination to it…
It was common to draw a diagonal line across a big card deck so that one could put it back together quickly.
One day I saw the Univac 1108 operator deal with a card jam by extracting the pieces and just throwing it out instead making a new card. It turns out that was in the data portion of the deck and was just sample points. Missing one or two was not a big problem.
A great article, Willis, with which I completely agree. I am another old physics guy who has done the hard yards in Fortran, Algol, Pascal, DOS, Basic, Visual Basic, C, C++, Python. I have even written in machine code for a vacuum tube computer (SILLIAC, Sydney Uni, 1963). I have coded fluid dynamic models of tidal and tropical cyclone forcing of ocean currents and became very skeptical of their value. The problem is that computer models are deterministic and cannot handle the stochastic phenomenon of turbulence because the Navier-Stokes equations break down at high Reynolds numbers. (I wrote a book on this, ISBN: 1-5275-3206-2). The idea that present day carbon emissions will remain in the atmosphere for millennia is false and based solely on climate models. Much of the hysterical rhetoric about emissions is based on this unwarranted assumption (Reid and Dengler, JGR Atmospheres, under review).
John,
Would SILLIACSbe the computer that we saw at Sydney when our class of 20 or so from RAAF College aPt Cook were on tour in 1969? Geoff S
Jean ?
you must remember jean !
and lunar landing
and
and
Great trip down memory lane Willis; thank you!
Re: (Continuous simulation) models. The value of modelling lies in building the model. The building process reveals what information is needed (both qualitatively and quantitatively) to close the loops and so points the modeller directly at the research that needs to be done. Parameters are admissions of failure to obtain the necessary information / relations.
Lunar Lander. On a KSR-33 teletype terminal. Good times.
Yes but what about:
Structured Expert Judgement and Calibrated Mental Models?
That I just read about in the previous Watts Up With That Sea Level Post?
Sorry, I couldn’t resist (-:
What, no PL/1? 🙂
PL/1 was the best of both (COBOL & FORTRAN) worlds. !!!
Or ADA or CORAL? (Both used in defence systems programming)
And fixpac.
Ah, PL/I. I learned that language in college but never worked for (or new of) anybody who actually used it. If I recall correctly, it was sort of a mash-up of COBOL and FORTRAN.
I learned PL/1 at Penn State in the late 1980s.
We had terminals to work on, but the compiler still operated in terms of punch cards. Leading to the dreaded “Deleted card encountered” error.
In its heyday I’d bet more business applications were written in PL-I than practically any other programming language with the possible exception of Cobal. Quoting from the wiki (yea, I know): “The PL/I optimizing compiler… was IBM’s workhorse compiler from the 1970s to the 1990s”
>>
. . . with the possible exception of Cobal.
<<
There’s a tremendous amount of legacy code written in COBOL. You can still find job offers in the help wanted ads for COBOL programmers.
Jim
Maybe one thing the computer is useful for is in a similar way to how you come to *properly* understand something yourself.
viz; Try to explain it to someone else
It will come undone even then though because the computer cannot or will not ask ‘why’
The computer thus becomes a perfect slave and works to reinforce erroneous thinking
Exactly what’s going on here inside Climate Science – computers are amplifying junk.
The Datapoint reminded me of my first desktop. The HP85
Ain’t they similar?
Mostly used as a datalogger, driving and recording the outputs of HP test equipments in the Electronics Research Lab I worked in – testing little analogue circuits over extremes of time, voltage, noise and temperature
Was it via the ‘HPIB Databus’ – that chuck sticking out the rear of the museum picture.
It did play a mean version of a card game – can’t recall now but the GF at the time, an avid Cribbage player, loved it
How times have moved on eh – things have never been better. not
“Maybe one thing the computer is useful for is in a similar way to how you come to *properly* understand something yourself.”
I use to write programs for an Accountant that could never tell me exactly what he wanted, but was very good at telling me what he didn’t want. We use to circle around the problem (kind of like finding the range for Artillery) until we got something close enough that he could use it.
“It did play a mean version of a card game”
My first program (in PDQ FORTRAN!) was a program to play “Chemin de fer” (its sort of the French version of Blackjack). Not really much of a program but is seemed like a brilliant piece of work at the time.
In my first job, on some nights I drove the van loaded with trays of punched cards and operator instructions to the computer for the overnight computer run. My career nearly ended prematurely the night I didn’t close the van’s back door properly, and two or three trays of cards fell out and spilled across the street. Fortunately, it wasn’t raining, and the cards were OK, and I shoved them back in what I hoped were the correct trays – completely out of order of course – and drove on to the computer. All the cards had a sequence number in columns 73-80, so all the operator had to do was start by sorting those trays while I kept my fingers crossed all night. In the morning – there had been a successful overnight run on the computer! Computing was an interesting exercise in those days.
Nice story, Willis. I enjoy them all.
in 1963 I had heard of those big computers, but the only one I ever held was my Uncle Bill’s Magic Brain.
Keep up the good work.
I had one of those. Cute toy. Of course, I would have done almost anything for a Curta calculator. http://www.vcalc.net/cu.htm
Willis, thought I would provide you with a fact tidbit in thanks for this most excellent post.
You noted that at climate model grid sizes they cannot ‘see’ thunderstorms despite using the Navier-Stokes equations.
After finishing his quantum chromodynamics, the ever curious Richard Feynman spent four years experimentally and mathematically exploring Navier-Stokes at CalTech. His conclusions were written up in the Feynman Lectures on Physics volume 2, chapters 40 “Flow of Dry Water” (classic Feynman joke, fluids with negligible viscosity) and 41 “Flow of Wet Water” (viscosity, N-S).
Last three paragraphs of the last section of Chapter 41 are quite famous. I hauled out my well thumbed Lectures copy and provide you an excerpt:
We have written the equations of wet water flow…
But had you not visited Earth, you would not know about thunderstorms…
The next great era of awakening of human intelligence may well produce a method of understanding the QUALITATIVE content of equations. Today we cannot.
Great quote, Rud, thanks. It points out the critical importance of emergent phenomena such as thunderstorms.
w.
With which emergent phenomena I completely agree, albeit from a different emergent perspective. But Feynman preceded us both. Genius is just that.
“The next great era of awakening of human intelligence may well produce a method of understanding the QUALITATIVE content of equations. Today we cannot.”
Truly great insight. When you and Willis write something, I pay attention. I know you have lived in the real world, unlike some of the usual detractors.
As I think about the dichotomy between skeptics and true believers, I wonder if that is what separates us.
Wow, what a good read Willis. As a humble accountant who tried vainly to use Excel to model the various what-ifs in creating a budgeting tool for our management and sales team, I cannot begin to imagine the complexities you are talking about, both within the software climate modellers are using and of course within the climate system itself.
OK Willis,
You have deliberately induced time travel, and now I need to get back home. I can remember some way points in getting here: Encounters with several languages, particularly C, using various Basics, doing 6500 machine programming, borrowing my neighbor’s Trash 80, punching Hollerith cards and getting disappointing results when trying to run my programming….
You have taken me back to 1958, and like you, I am riding a bicycle around downtown San Francisco. No blueprints, though–I work out of the Western Union office around Third and Market. One of my weekly jobs is to pick up the printing mats for the weekend comics from the Chronicle office and take them across the street to the Examiner. Why they paid Western Union to do this, I have no idea–unless it had something to do with a union contract.
It’s a heady experience for a 17 year old to be walking around the street holding in his hands all the Sunday comics for most of San Francisco. But I really need to get back home in time for dinner. The models can wait.
Willis: thanks for the wonderful trip down Memory Lane! I’m almost exactly your age. While I’ve never written programs for a living, I did learn enough Fortran to do a bit of programming. And my first wage-paying job, at age 12 or 13, was sweeping out the Okla A&M Computer Center, which consisted of a vacuum-tube Univac, obsolete even then. But kind of a plum job for a kid: air-conditioning! — a rarity then. Pay was 50c. an hour!
And the anecdote of Freeman Dyson consulting Fermi on his toy model was priceless! Dyson didn’t have much use for the climate models, either. And got Nastygrams from the Usual Suspects for his pains, when he said so.
Well, we need to keep whacking away at the Good Fight! Science will self-correct, in due time? We hope before the Climate Activists wreck our economy with useless and actively harmful ‘solutions’. Giant, floating windmills in the North Sea! Or offshore Northern California. Yeah, those will work….
“And the anecdote of Freeman Dyson consulting Fermi ….”
It was Richard Feynman.
Although Feynman likely retold it, it was actually Dyson … the full story is here.
https://www.dropbox.com/s/rrqax2qtpmqupwo/dyson.pdf?dl=0
w.
Willis, thank you for this post. Ah yes, punch cards. But I digress.
I recall this earlier post in which you recounted your interaction with Gavin Schmidt on how the models made sure that energy was conserved. “So I asked him how large that energy imbalance typically was … and to my astonishment, he said he didn’t know.”
https://wattsupwiththat.com/2020/01/18/gavins-falsifiable-science/
Last year I came across this paper, Zhao et all 2018b, which characterizes the GFDL’s AM4.0 model.
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2017MS001209
Supplement 1, in the lead paragraph (link to pdf below):
“S1 Treatment of energy conservation in dynamical core. The dissipation of kinetic energy in this model, besides the part due to explicit vertical diffusion, occurs implicitly as a consequence of the advection algorithm. As a result, the dissipative heating balancing this loss of kinetic energy cannot easily be computed locally, and is, instead returned to the flow by a spatially uniform tropospheric heating. This dissipative heating associated with the advection in the dynamical core in AM4.0 is ∼ 2 W m−2.”
https://agupubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1002%2F2017MS001209&file=jame20558-sup-0001-2017MS001209-s01.pdf
So at least in the GFDL’s latest model, about 2 W/m^2 is added back for energy conservation. My sense is that the model has to blur the handling of so much energy that discriminating between incremental outward emission to space and incremental heating of the surface from greenhouse gas forcing is simply not possible.
Am I looking at this correctly?
https://judithcurry.com/2013/06/28/open-thread-weekend-23/#comment-338257
The Common Atmosphere Model 5.0 from National Center for Atmospheric Research decreed a latent heat of water vaporization to be a constant independent of temperature. They chose a value that is 2.5% too high for tropical seas (where most of surface water evaporation on our planet happens). NCAR CAM 5 is considered “science”.
https://judithcurry.com/2012/08/30/activate-your-science/#comment-234131
“If the specific heats of condensate and vapour is assumed to be zero (which is a pretty good assumption given the small ratio of water to air, and one often made in atmospheric models) then the appropriate L is constant (=L0). (Note that all models correctly track the latent heat of condensate).”
RSX-11M-PLUS operating system.
With you there. We built a multi-terminal Level-2 cell controller on M+ with industrial terminals that had a VT-100 emulation mode.
VT-100. Yeah. 🙂
Lots of DEC courses at the Butts Centre in Reading.
Excellent essay on computer models and calculations, Willis! As a Chemistry student, I took a course in Numerical Analysis in 1966-67, and our first assignments were to solve simple arithmetic calculations using (1) pencil & paper, (2) mechanical calculators filled with springs, levers and gears (the kind that went nuts if you tried to divide by zero), and (3) simple FORTRAN IV programs (involving going to the punch card room, followed by submitting the deck of punch cards to the computer operator of the IBM 7040 or 7090). The idea was to show that the computer program did no more than arithmetic, although a lot faster when many iterations and 10 significant digits were involved. Then we could get on to solving equations, calculating definite integrals, derivatives, etc. using approximations and arithmetic.
Estimates of climate sensitivity (the amount of warming when CO2 is doubled) start with computer models of infrared (IR) spectra obtained by satellites looking down on a warm Earth (for example, see Fig. 3 at https://climateaudit.org/?p=2572 ). Models in Molecular Physics of the bond-bending vibrations in CO2, H2O (water vapour), O3 (ozone), N2O are quite good and fit the observed spectra (see the MODTRAN calculated spectra to 20 km altitude at https://en.wikipedia.org/wiki/Radiative_forcing ).
William Happer has run high resolution (HITRAN) models to altitudes beyond 70 km, with similar results (see 24:13 to 29:17 at https://www.youtube.com/watch?v=CA1zUW4uOSw .
The entire 45 minute presentation is perhaps the single best introduction to the real science of climate change, and can be understood in 26 minutes when played back at 1.75x speed).
Happer’s value of 3 W/m^2 for radiative forcing corresponds to the tiny difference in area between the red and black simulated spectra. This agrees with the 3.39 W/m^2 which corresponds to the tiny difference in area between the green and blue simulated spectra in the Wikipedia article on Radiative Forcing. The oft-quoted value of 3.7 W/m^2 appears to come from the equation “delta F” = 5.35 ln2, but in fact 3.7 W/m^2 came from another computer simulation, and the factor 5.35 is derived from 3.7/ln2 . The difference between 3.39 and 3.7 W/m^2 corresponds to about 8-9%, and may indicate the possible error expected in these calculations.
A Top Of the Atmosphere (TOA) outgoing flux of 240 W/m^2 corresponds to a Planck black body temperature of 255.07 K (see https://en.wikipedia.org/wiki/Stefan-Boltzmann_law ). If we add or subtract 3.7 W/m^2 from 240 W/m^2, we get Planck black body temperatures that differ from 255.07 K by about 1.0 degree (the difference is about 0.9 K if we use 3.39 W/m^2).
Because tropospheric temperature profiles are roughly parallel regardless of latitude (if we ignore temperature inversions during the long polar nights), the change in surface temperature on doubling CO2 will be 1.0 or 0.9 degrees (depending on whether we use 3.7 or 3.39 W/m^2), not including feedbacks. The reason is that if energy is added to all the molecules of the atmosphere, the most probable distribution results when each molecule, on average, gets an equal share. I.e. each parcel of air warms up by the same number of degrees, on average, since temperature is a measure of the average kinetic energy of the molecules.
However, the error increases when we consider water vapour feedback. The reason is that CO2 is relatively constant at around 410 ppmv regardless of altitude, whereas water vapour is generally assumed to be at 50% relative humidity (obviously not true in the muggy tropics) and saturated water vapour pressure (for 100% relative humidity) varies exponentially with temperature (see https://en.wikipedia.org/wiki/Clausius-Clapeyron_relation ).
I estimate from the IR spectra from 0 to 2400 cm^-1 that water vapour feedback after one iteration is around 32%. Therefore using r = 0.32 in an infinite geometric series, the climate sensitivity including water vapour feedback could be as high as a/(1-r) = 1.47a = 1.47 K (assuming a = 1.0K). An infinite number of iterations would boost 32% feedback to 47%.
Note: if r = 2/3, then a/(1-r) = 3a = 3 degrees of warming if a = 1.0 K. This is how positive water vapour feedback boosted climate sensitivity from 1 degree to the long-quoted 3 degrees.
Yet this 3 degrees had to be at least a factor of 2 too large, for the following reason:
0.8 degrees of warming occurred between 1850 and 2019, as CO2 increased from 285 to 410 ppmv. Since warming is proportional to the logarithm of the CO2 concentration, then the MAXIMUM expected warming on doubling CO2 is 0.8[log2/log(410/285)] = 1.5 degrees. Similar calculations could have been made anytime during the last 3 decades.
And this assumes that ALL of the 0.8 degree warming was due only to CO2 and related feedbacks. It’s no good to say that EQUILBRIUM climate sensitivity is higher due to a time constant (lag), because any lag in warming would mean that temperatures would continue to increase even if CO2 remained constant. This is not consistent with the observed hiatus in warming over the last 2 decades, even as CO2 has continued to increase (not remain constant).
In addition, all computer calculated spectra, including Happer’s, assume a cloudless troposphere. But clouds reduce the overall importance of changing CO2 to only the molecules in the path length above the cloud tops – since condensed phases like cloud particles act as miniature Planck black bodies which absorb and emit ALL long-wavelength infrared (IR), meaning that any extra absorption by CO2 below the tops of clouds is cancelled by exactly that much less absorption by the cloud particles. And the Earth has about 62% cloud cover, which is increased slightly by increased water vapour, resulting in a slightly higher albedo (just consider that as the Sun rises in the tropics, a clear morning sky turns to afternoon thunderstorm clouds).
So the bottom line is that doubling CO2 would result in about 0.7 degrees net warming, including feedbacks, not 1.5 degrees, and definitely not 3 degrees.
Computer models can then be run as CO2 changes as a function of time, but all must be pinned by the assumed value of climate sensitivity. Predicted temperature changes will all be too high by how much greater their assumed climate sensitivity is than 0.7 degrees. And some want to ruin the economies of the free world over these flawed predictions???
Friden calculator?
My father’s employer had a Marchant(?) calculator that just sat there spinning the drive shaft when told to divide by zero. It also had a stop key.
Real division was much, much neater.
First computer programing course was in Fortran about 1966. You would walk across campus to submit your cards to be run. A day or two later you would check the output usually to find out the program didn’t run and you had no idea why. A great way to kill any zeal for programing. About 10 years later used Basic and found it much more satisfying since it could be submitted remotely with quick turn around. First PC experience was about 1982 when the company bought a an IBM PC. Basically they said use it if you wish but don’t expect any help. Told this story to some junior workers a few years back and they didn’t even know what a punch card was.
A Datapoint 2200!!! My first paid programming job, in 1976, was on Datapoint 2200s, soon upgraded to 5500s, with a whopping 56K of RAM, 4K of ROM, and the other 4K was interrupt vectors in RAM, but not available otherwise … I think.
Thanks for the memories. On the darker side, an extra “build” in here, near the beginning: “They started out by having us build design and build logic circuit”
Thanks, fixed. My motto is “Perfect is good enough.”
w.
Willis, you should be careful sharing this stuff, you come off sounding like a privileged hard-core rayssist, plus you’re male. It’s not a good look.
I do love your dedication to rigorous uncertainty within “the séance”.
Fortran77 and digital watches since 1977. Although I do have a 17 year old Texas Instruments programmable calculator with a Mac processor and 1 megabyte of ROM and 512k of RAM.
I’m delighted with the predictive performance range and useful time interval of current WX models RE “life or death situations”, for amelioration. But fully share your view of climate models. They’re strong evidence of the pivotal role of mass delusion in the development of human civilization, especially one with computers and Mars impact probes.
One thing though, being a failed-geo, due to wanting to understand what’s under me, much more than work as a pro, I have a problem with this sort of remark:
You’re being much too generous, as the past is not known. We have ‘records’ stored, but always stored in media that really is not good at storing things. Look at any rock outcrop, and the leached regolith above it, and the soil that forms on top. That’s the Earth’s version of a ‘floppy disk’.
#1. Records aren’t.
Most people can’t even accept that much.
Only a small-fraction of a record is ‘complete’ (usually < 0.001% ‘complete’). And even those incomplete ones are in a constant state of edit within geology, where you get to see the skin, and a few outer dermal layers at certain points. Same with anthro-studies. Same even with modern history, which absolutely must be edited to suit the now, so that the future will not get the then all ‘wrong’.
The tares : wheat ratio in the < 0.001% editing is another area of concern.
But I would say that no, “climate model can hindcast the past climate perfectly”, not even the bit for which certain eager-beavers have suitably edited and biased the < 0.001% ‘record’ contents to enable that to seem to occur with a suitably compelling delusion-forming and reinforcing capability, that improves with each suite of reality versions, which are tuned to better emit money streams.
Which is how you secure the new abacus.
And then there’s infinity. Which complicates things a fair bit, as computers can only deal with discrete defined abstractions – ‘not-infinity’. Else logic tends to explode.
It reveals a computer is fundamentally incapable of addressing the understanding of an infinity, even if said abacus was fabricated from one, as was our brain. Which is disconcerting, and best set aside until on a death-bed, where one may finally be prepared to ponder how that can be?
And still come up with no answers.
I’ve come to the view that this is all I’ll ever really ‘know’. And there’s no logic model that can change it.
It’s abstracted figments … all the way down.
Actually, scientific papers and text books share a lot of these same characteristics.
Vive la edit!
Willis,
Thanks for the memories! I’ll be 80 in June. Performance is dropping off.
First computer was a Data General Nova with 4K of memory,1968 with no mass storage and an ASR-33 punched paper tape teletype. Next was a PDP 8e, an 8K OEM on a gamma spectrometer,1970. Did a course in machine language for the Nova, failed it and forgot it. In 1976 I flew from Sydney to San Francisco with my genius mate Albert, for the sole purpose of having a lady in a huge factory withdraw some of the wires on a memory card, to replace a single faulty ferrite core on the 4K card. Such was the value of memory then.
Was fortunate from then on to have a computer department of mathematician professionals to compute to specifications I would dream up in mineral exploration work. Never did learn to program, though could read enough in several languages to get the gist and fix easy errors. Our early corporate work was on a Hewlett-Packard 2000 series.
Our geophysics group was at world leading edge with models to simulate the size, broad shape, disposition of underground discrete bodies with magnetic properties, from surface measurements of the displacement of the earth’s magnetic field. The model was created and used in pre-computer days using a mechanical calculator, enter a value then pull the handle. It worked brilliantly.This taught me that models can work in valuable ways, but that they need iteration after iteration until (practically) all measured variables are reconciled by the model. A tiny adjustment taking days to finesse could have large effects on the final result. Same applies to climate GCMs, I presume. I gave up on those when The Establishment decided it was leading edge mathematics to average the results of dozens of diverse computer runs to arrive at a best estimate, with no mathematical analysis offered as to the validity of this kindergarten grade error. Geoff S
I believe I may be older than most here – a decade older than Willis. My memories go back to Turing – well, almost! He helped with the design of ACE for NPL – and ACE was the grandfather of DEUCE, made by English Electric, where I worked for the Guided Weapons Division in Luton in 1958. I was working with the LACE team (LACE was an analogue computer) in the Maths/Physcics dept.
As a young lad of 20, the computer section, however, was of great interest to me. It was filled with young females aged around 18 and run by a very attractive female mathematician of around 25 years. They were called computers…all armed with Fridens…
Across the way from our lab was one of the DEUCEs. It was the size of a small room with an aisle down the center. I remember being shown over it with its mercury delay line RAM and rotary drum memory. I never used it but it was used in conjunction with LACE for analysis. The MTBF was about 4 hours before one of the 1500 double triodes (12AT7’s, IIRC) bit the dust, or so one of the maintenance techs said..
At the time we were having a small problem with the Thunderbird Missile, which kept blowing up on test flights for then unknown reasons. Yet simulations said all was OK with the servos, rocket parameters, etc. I think it turned out that the aerodynamics of the control fins was incorrectly modelled, found out by wind-tunnel measurements,causing instability and fin breakup.
Around that time my room mate, a mathematician engaged in programming DEUCE told me about Algol, which fascinated me and I fiddled with it, but it was 9 years before I used Fortran in the US – they actually let engineers use the IBM 360 in the lunch hour.
Complex Fortran served me for a decade or so; I also used Basic. I avoided Pascal and C and found MATLAB and never looked back – I left all programming to experts where it belongs and concentrated on what matters – ideas. “The purpose of computing is insight, not numbers.” R.W.Hamming, Numerical Methods for Scientists and Engineers.
A comment for Brian Jackson, who for some reason has an antipathy to me and is doing his best to tear me down. He claims I’m boasting about my computer skills. Not the case.
In fact, there are plenty of people who can program rings around me. They usually have PhD’s in Computer Something, like say Gavin Schmidt.
Me, I’m self-taught in programming. I learned all of my computer languages but the first two by the RTFM method. In fact, I’m self-taught in all of science—college Intro to Physics and Intro to Chemistry are as far as I ever went.
As a result, when I propose some scientific idea or make some scientific claim, far too many people like Brian attack my credentials rather than attacking my ideas. Note that he hasn’t said one word about the subject of the post, the problems with computer models.
So for this post, I figured I’d see if I could short-circuit some of those attacks on my obvious lack of any scientific credentials by describing how I mastered computer programming—through unending personal effort.
Still, that’s not enough for Brian Jackson. I well remember using CP/M to do all kinds of things on my IBM PC, so I called it a “computer language”, which I take to be any way to communicate with a computer. Sadly Brian views this as some kind of indicator that I’m truly clueless, and not what he loudly proclaims that he is, a “real” computer programmer.
Call it what you like, Brian, a language or an operating system. I’m used to folks like you who think that ad hominem attacks and nit-picking about semantics and meaningless details are a valuable substitute for actually discussing the ideas I present.
Pass … go convince someone else that you’re a “real” computer programmer, Brian. Here, you’ve already convinced most people that you are a real pric … pri … prince.
w
Take heart Willis. I’ve seen Brian’s type, probably a “young” 30’s something with Masters in something and he thinks because he learned some climate stuff from a few books and professors who acted like gods, he knows all there is to know. One thing I learned going from Masters to PhD work, is you realize that what the books say (Master level – take it on faith, replicate some well proven fact in masters thesis-graduate) and then going to: trying to fully replicate it and then go on to a new level or direction yourself (PhD work) you realize the first guys were either incredibly lucky or liars. Getting a PhD in a hard lab science at a real institution with a real thesis committee is a humbling experience, in realizing both how much of what is in college text books, and especially peer reviewed papers, is likely wrong, and how to figure out “right” answers leads down so many rabbit holes.
As such, I know Brian. He’s like the “Karens” of the social media media, all huff and puff and outrage. But Brian and his arguments wouldn’t last 3 minutes in a debate with an informed skeptic on climate change. PhD or not. It’s why the PhD Gavin Schmidt’s runs from the PhD “Roy Spencers.” Just like Dementia Biden, they’re in hid’in. They know in a live Q&A with an informed questioner will show to world they “have no clothes on.”
Sometimes it’s easier to pull someone down rather than climb to their level.
Willis,
Do what I do and just ignore Brian Jackass.
Thanks, Paul. I’m to that point. Up until now, I figured there was a chance he was convincing some of the lurkers. At this point, he’s just taking a shotgun to his own toes …
w.
Willis,as my dad use to say: “Just consider the source. It ain’t worth worrying about.”
Ric Werme’s Three Rules for dealing with trolls:
1) Don’t reply right away.
Trolls thrive on acknowledgement. The content doesn’t matter, the attention does. The quicker you reply, the stronger the reward.
2) Don’t reply unless you are adding to the discussion.
I hate the he said, she said, he said again, she said again nature of trolling.
By the time repetition sets in, other readers (if there are any left) realize its trolling.
See point 1).
3) Let the troll have the last word.
He will take it anyway.
See point 1).
Science is all about nit-picking semantics and details. That is why you are not a a scientist. You are an amateur trying to play with the big boys. I don’t have to convince anybody about anything, but I will call you out on your BS and poke at that inflated ego of yours. Your misuse of the term “computer language” make computer professionals laugh at you. (ha ha ha ha)
.
” I mastered computer programming” not you have not. You are an amateur. You attack computer models because of your ignorance. Climate models are all wrong, and they are skillful. You miss the fact that they are skillful. I challenge you as an amateur hack, to get the source code of a modern climate model, and use your script kiddie skills to IMPROVE it. That is what a real scientist would do, instead of wasting your time impressing the dweebs that inhabit this blog. The uniformed may praise you, and their feedback inflates your enormous ego, but in the end, you have contributed noting at all to science.
.
1) Science is not about nit picking things that have absolutely nothing to do with the subject at hand. Your comments are the logical equivalent of declaring that someone can’t be a scientist because they used “their” when they should have used “they’re”. The only thing you are proving is that you yourself know nothing about science.
2) The only one displaying an inflated ego is you Brian. You just can’t accept that nobody here takes you as seriously as you take yourself.
3) Despite your repeated claims, there is not a single computer professional on this site who is laughing at Willis. On the other hand, quite a few of us have demonstrated how full of sh1t you are.
4) Because he disagrees with you regarding the dividing line between OS and language, therefore he’s can’t have mastered computer programming? Once again Brian demonstrates that he considers himself to be the standard against which all others must be measured. In other words, Brian’s enormous ego is getting in the way of actually thinking for himself.
5) Willis attacks computer models because they demonstratably are not skillful and all fail even the most basic of tests.
6) To be a “scientist” Willis has to volunteer to help others fix their broken code? What makes you think any of the programmers want any help from outsiders?
7) Proving existing theories wrong does nothing to advance science? Really? And to think you consider yourself to be an expert on “science”.
8) Once again Brian demonstrates his stellar sized ego by declaring that anyone who doesn’t agree with him is an ignorant amateur.
9) BTW, I notice that once again, you don’t actually refute anything Willis wrote, you just declare that since you disagree with him, he must be wrong. That’s not exactly a scientific attitude.
“I mastered computer programming” no you have not. You are an amateur. You attack computer models because of your ignorance.”
Now I think I’m starting to get why Brian goes so hard after Willis. Willis is bursting Brian’s climate alarmist bubble with his excellent posts.
Delighted. I had to evaluate the Datapoint for the National Provident Fund I was tasked with bringing into existence. I met the programming team of the Systems Corp, based then in Hawaii, who created the elegant Databus. They came out of the payroll team in the Vietnam War.
We didn’t use them eventually as the Govt of American Samoa allowed us graveyard shift time on their IBM System 3–a 96 hole punched card input machine with tape drives and 29Mb hard disk units.
Worked with numerous programming languages including Wang 2200T Basic. (48Kb rom which even allows access to the 8″ disk functions.)
Learnt to program on a Monroe 1620 calculator which had a portapunch. Unfortunately the holes could drop out of the program cards which made fixing the life table programs impossible.
In 1976 worked on the design and build of a Z80 based 4Mhz, 64Kb ram, S-Bus with screen , plotter, teletype, 8″ FDD with CPM 1 O/S. That was Zilab 1. It was to run house plan drafting software.Originally ran extended Basic but that morphed into a variety of HP9000 calculator basic which the original software was developed in.
I much appreciate your common sense and work. I’m a fresh 81. Have looked at R, much tempted but I’m more verbal these days. Interesting too, how musicians pop up so frequently among us numeric types. My youngest daughter, 19, has just started in the vocal school at Uni.
FREEBIE: Anyone interested can get a free copy of my Bio, “My First 80 Years” which gives a much enlarged version of my historical entanglement with computing, Just go to
lambtonpublishing.com/contact
and give me the address to send it to.
Happiness and blessings to all.
Kev.
Hi Willis, thanks for reminiscending about the past and your computer experiences…. All very interesting and now documented for our future generations. I am the same age but from a different field, from focussing onto a very wide range of science and knowledge. How, in fact, the input into computers is carried out in detail, is unrelevant for me. This is nothimg but pure skills (not bad if someone can do it) and not real creative thinking….. you admit that models are only the preconceptions of model authors, who put unproven ideas into models, which makes them ‘”official science”.
This was introduction. Now to the point: For models, input variables need to be complete. If one major climate forcing is missing/ignored/not known/excluded….. then models cannot produce the realidad and do not withstand the test of accuracy.
Concerning climate modelling: The most and major climate forcing is “orbital forcing”. Not the long-term processes such as obliquity and precession/ Those two are always pulled out of basement, when the question to orbital forcing arises. Orbital forcing, which is active every couple of weeks, consists of orbital osculation, perturbation, oscillation. The orbit is subject to ‘Gauss’ Perturbing Equations, which are also used to calculate practical osculating satellite motions.
And this osculation of the Earth orbit was/is left out of climate models, inclusive in all your climate analyses.
Therefore: The firstmost question for computer programming is: Are all major variables present?
There is no answer from you on this question. You refuse to read my papers, because what you do not know, this must be wrong for you from the beginning. You know what I mean.
In June this year, I will have the paper completed, demonstrating empirical meteorological evidence from all over the continents and oceans, how orbital forcing, without any modelling, impacts our daily weather and the climate evolution.
As I say, clinkering on keyboards does not get us closer to the understanding of the climate…the proof is the divergence of CMPI6 climate models and the differences in ECS, which are incorrct, because one major climate forcing was, knowingly or unknowingly, ignored.
In any physical system in which man tries to understand, there known unknowns and unknown unknowns. Climate modeling is no different.
Men trying to understand women exhibits and proves this point pretty well.
Joaquin, you say:
Since I don’t do “climate analyses” based on climate models, I have no idea what this means.
Nonsense. I said in this very post that I think major variables, to wit emergent climate phenomena, are not present in climate models. So your claim is falsified by this post itself, and that doesn’t count the many other times I’ve said that models leave out important variables.
Actually, Joaquin, I have no idea what you do mean. I have no memory of “refusing” to read your papers, but having published over 900 posts on this site and given my pre-doomed but still fixed intention to answer all relevant comments, I might have done so. Seems unlikely, though, not my style. I may have started reading a paper of yours and given up, but I don’t recall even that.
As to the orbital perturbations of the planetary orbit from external forces, these would presumably be from the gravitational forces of the moon and the major planets. I’ve looked at some of these, for example calculating the tidal forces from the major planets. They are trivially small. The moon perturbs the earth’s orbit, and it leads to the tides. I’ve also analyzed the tides, as well as the underlying lunar tidal force itself, to see what effect they might have on surface climate datasets of several types. I found no significant correlations.
So please, dial back on the accusations. At this point, I wouldn’t look at your paper if you paid me, although tomorrow when my blood is less angrified I likely would. But insulting a man is a very poor way to get him to pay attention to your whizbang theory of everything …
w.
Lets leave things as they stand… lets avoid polemics. The new paper will not mention Moon and tides, the so-called 3-body problem, which is only relevant to a certain small. degree. The new paper deals with the “true Earth orbit trajectory” ,which is different to the generally, in climate science, assumed elliptical Kepler line. The true trajectory impacts climate and weather and I will show massive meteorological proof.
And you write you do not know what I mean.
And to polemics:You started with the “cyclemania” some years ago – the fight is still open-… but this time there will be overwhelming evidence that the the true orbit has massive cyclic imprints on global climate, observable roughly about every 2 months at fixed dates. The “cycle fighters” will, this time, retreat into their (where they live).
Thanks, Joachim. Let us know when the new paper is ready for comment, and I will read it and see what I think.
As to “cyclemania”, it’s a real thing. Doesn’t mean it applies to you, or to any given study of cycles … only those for whom the shoe fits.
Best regards,
w.
The climate models do EXACTLY the two things they are intended to do:
1) Provide a nice paycheck to the computer engineers and climate pseudoscientists who write, debug, and run them, and then advertise the outputs in journal papers, conferences, and IPCC reports.
2) Provide an alarmist CO2-climate message for the paymasters to ensure 1) continues.
Ive only found models to be useful to scope and design an experiment. They are particularly useful in convincing the budget folks that you understand what you are doing.
All I can say is that I have worked as a programmer for a hedge fund for over 20 years. I have seen many PhDs from Ivy League schools come through the door at work and despite their pedigree never ended up making any money off their sophisticated models.
Thanks, Willis. This is by far the best contribution you have made to wattsupwiththat.com over the years. Have you written a memoir?
Forrest, coming from you that is high praise indeed, and greatly appreciated. I’m working on putting my various autobiographical pieces together into a memoir … but I have about 1,200 pages of them. I’ll likely publish them as ebooks in several volumes. In the meantime, you can read the individual posts here. They’re also available in the “Autobiography” section of my 2021 index to all my WUWT posts, which is here.
My best to you, and many thanks for all of your scientific contributions over your most inventive and productive life.
w.
(For those unaware of Forrest, he was one of my heroes as a kid because after Jearl Walker left, for several months Forrest wrote the column called “The Amateur Scientist” in the once-great but now rubbish “Scientific American” magazine.)
And here I mistakenly thought that I was the only person to recognize his name. I became aware of his contributions to monitoring UV when I first got involved in researching the so-called “Ozone Hole.”
And a column in Popular Electronics. Lessee, no he didn’t write the article on Big TC and Little TC (Tesla Coil). Both I and a college housemate saved that issue – I helped him build Big TC, it looked almost exactly like the cover photo.
And of course, after reading https://journals.ametsoc.org/view/journals/bams/92/10/2011bams3215_1.xml I rushed right over to Amazon to buy a Kintrex infrared thermometer.
Too many names to thank these days. Thank you Martin Gardner. Thank you Jearl Walker. Thank you Forrest Mimms. Thank you Katalin Kariko.
Willis, for me CPM was the operating system for the Osborne and similar computers (Kaypro and a few others). I learned Fortran and the computer department had a shiny new mainframe and plenty of stations with a keyboard and CRT. No punch cards any more, although plenty of my coworkers still had boxes of them from before they switched to workstations.
Even before you get into the shortfalls of modeling I also believe as Fermi said: “One way, and this is the way I prefer, is to have a clear physical picture of the process that you are calculating” IMO Climate Science got that wrong. They picked CO2 and since 1988 about 7 molecules of water vapor have been added for each molecule of CO2. Only about 2/3 of the 7 are because it got warmer and CO2 had no significant contribution to that.
The first computer I used was a CDC 6600. Really good architecture, but CDC marketed it for the scientific and engineering fields, so it couldn’t beat the IBM 360/370 series which were marketed for the business and banking fields. Pity, since its architecture was so good.
Mild nitpick – TRS-80 was the line, with various models. I had both a portable Model 100 (which is what you had) and a desktop Model III.
Willis, thank you for the excellent post. I’m sure there are a lot of readers like myself who have much appreciation for your insight into the world of computer modeling.
Thanks for the encouragement, eyesonu. Since my youth, I’ve seen part of my purpose as being to venture to parts of the physical (and intellectual) planet where most folks don’t have the opportunity to go, and then come back and report my findings.
I’m slowly getting better at it.
w.
Think about the ‘spaghetti models” offered each time a hurricane approaches North America. They all start with the same initial data, make different assumptions, and produce wildly different results, showing a landfall from Cape Cod to Miami to Merida. Each one of them has been run and it’s forecasts compared to reality many times and been tuned and improved as a result, yet they all differ. If we evacuated Miami each time an early model forecast a hurricane ending up there, we would have foolishly spent billions. Yet all of these models are have much more “V&V” than any climate model, and deal with fewer parameters over a much shorter timespan. The climate models would have us foolishly spend trillions.
Hurricane models are getting better. The 3 day forecasts are about as good as the 2 day forecasts were 15 to 20 years ago.
The difference is, with hurricanes we have a variety of predictions from models where there is no reason to prefer one to the other. So basically this is just reflecting uncertainty about the dynamics.
We don’t have one proven accurate one and a bunch of proven crappy ones, and then average the lot of them to product nonsense.
With climate we have one decent model, consistently accurate, the Russian one. And a whole bunch of others which are inconsistent with observed outcomes.
So what do we do? We average the one good ones predictions with those of the proven failures, and proclaim the result as good enough to drive policy.
Someone tell me what I am missing, why this is not insane!
I reckon you could be just the guy to give me a hand to disprove the idea that the movement of the planets affects the short-term climate/weather on Earth.
Farmers really want to know what each season will be like, either hot or cold ; wet or dry, as experienced by plants and animals.
I have what I call my “eye of newt” weather almanac (because it is so wacky) , but I’ve used it for decades because its helpful. Even my hay and silage contractor calls up for a “reading”.
I’d love to prove that it has no predictive value i.e. falsify the null hypothesis.
Whadda y reckon?
“One of the factors that cause urban environments to reach sweltering temperatures is a lack of vegetation. Trees alone could make a big difference: One model suggests that urban temperatures in Phoenix, Arizona, the hottest city in the US, could be reduced by over four degrees Fahrenheit if more trees provided cooling canopies over the scorching hot city.”
Can planting more trees keep cities from heating up? (msn.com)
The climate changers don’t do irony well. Still there is some healthy skepticism with the model-
‘“I think that there’s typically this sort of blind faith that we place in trees, that they will provide all of these wonderful social benefits,” says V. Kelly Turner, assistant professor of urban planning at UCLA. “But the environmental benefits that trees provide are entirely context-dependent,” she says. Patchy adoption of sprawling tree planting plans may not lead to the cooling bliss we desire in urban places.’
Thanks for the word “lithosphere”, complained just yesterday we don’t have a word for the more solid part of our biosphere. So I learn, love this site!
True ‘dat … WUWT ROOLZ!
w.
Definitely a memory lane post as well as a manual for modeling.
No one has mentioned Pascal. Why not? Not that I have written any programs in say the last 14 years or so but Pascal is still my go to language.
I remember Pascal from the 80s when I was at HP. If I recall correctly, C and C+ were coming up then, those three along with Unix and Basic were my programming mediums.
A quick search reveals that in addition to my mentioning Pascal in the head post, five commenters mentioned it …
w.
I’m impressed at the number of mentions of Algol. (Even discounting mine!)
Sorry missed those 5 mentions. :))
By the 1980’s Pascal was a very good language for teaching programming. Since it became all the rage we had several minicomputer reps that told us they were soon coming out with CPU chips that would execute P-Code directly. And, we had a product manage that thought it was so great you could develop an operating system in it. That project turned out to be a waste of several years and a few million dollars. Don’t know how it is now, but at least back then, there were an awful lot of incompetent people in the industry (i.e., salesmen, development managers, etc.) chasing the latest buzzword.
I the same age and with the same computers.I remember those cards and the hardship to get points in right place and the looong time waiting for next try. I left the field as a programmer when I realized that the assumptions I made (about the soil properties) where of more importance than he math.
Regarding computer models I listened to Roy Spencer and his one dimensional model.It fits!
Best regards!
Thanks Willis, Highly interesting stuff, reminds me of the computers used in the Apollo programme to land on the moon: https://www.youtube.com/watch?v=B1J2RMorJXM
I remember being told that the Apple II had more processing power than the Apollo computers.
I don’t know if that is true or not.
Most modern calculators have more processing power than the Apollo computers had. Heck, even 5 year old cell phones have more processing power than the Apollo computers did.
Thanks for the video ref. If I remember correctly the Launch Vehicle Digital Computer (LVDC) that was the autopilot on the Saturn 5 rocket was a one-bit serial computer that executed around 12k instructions per second.
Tank you for this, it really brought back the memories. It is good to see the comments from so many people who have long experience of computers and all the programming languages and operating systems.
How many of today’s computer ‘experts’ have this depth of experience and knowledge?
One of my memories of an IBM 360 was of a very large room full of large cabinets, and how the IBM engineers seemed to take up half the morning testing the computers.
Their ‘uniform’ was a white shirt and dark trousers, and they annoyed the commissionaires by insisting on parking right outside the main entrance, such was their self-imortance. The commissionaires soon put a stop to that by sticking a large message on the drivers side of the windscreen asking them not to block access for the emergency services. It was good fun seeing them taking ages trying to scrape it off with a razor blade.
The big problem now is the blind belief in the infallibility of computer output.
“Computer says so” seems to be the buzzword, or “no” in financial matters.
a veritable blast of nostalgia there! I instantly recognised the rings from core memory and the punch card… I started my programming in IBM assembler on a knock off copy of an early IBM mainframe…
Sometimes we used to hand punch individual punched cards if it was a minor change on a huge and noisy hand punch… what great fun to ‘have’ to make such a change when your colleague was suffering from a hangover! (how thoughtless is youth!)
I think that is one of the best articles I can remember you writing.
i remember a room sized punch card computer too, and the tinybbc? macs
and they both put me off of using pcs until the 2000s
now i can follow instruction to put code into my linux terminal
and sometimes get it right
lol;-)
Thank you Willis – your story brought back a flood of memories. One of my first job’s when an engineering student at Queen’s circa 1970 was to hand-process a ton of data. After a few days sweating in a hot room, I taught myself Fortran and got a mountain of work done while my employer took two weeks vacation. He was great guy but with no computer skills, so he thought I was a genius and I did nothing to dissuade him.
I thought Algol was a huge improvement over the Fortran of that day. I’ve learned many languages over the years as needed.
My most significant work was to model the economics of the Alberta oilsands and spot some future crises that would cause the owners great financial harm. I took my proposal for new Crown Royalty terms to the Syncrude Management Committee, and we got new Royalty terms that revitalized the industry. I had previously obtained new tax terms that were highly beneficial – allowing major capital expenses to be deducted in two years rather than ~ten.
The oil sands industry became the mainstay of the Canadian economy for 15 years, with over
$250 billion in new capital investments and approximately 500,000 new jobs created. Canada became the fifth-largest oil producer in the world, the largest foreign supplier of energy to the USA and the most successful economy of the G8 countries.
Subsequent idiot federal and provincial politicians trashed these fiscal terms, and together
with the actions of paid anti-pipeline thugs, the Alberta oilsands industry is now suffering. There is nothing that good men can create that imbeciles and scoundrels cannot destroy.
Excellent. You told the story of Datapoint in one line.
“There is nothing that good men can create that imbeciles and scoundrels cannot destroy.
Great overview for a non computer persn. Now do the covid models that lead to the great lockdowns causing immediate harm to tens of millions of people for no real benefit.
Uber-Geeks used to inhabit the Computer Sciences Building 24/7, intently working on their favorite projects. Punch cards were the input medium, and they had to be kept in perfect order. A common practice was to write a diagonal line with a felt pen on the card stack, to prevent disasters. A vivid recollection was some kid with a 30-inch high stack of cards tripping and spilling his card deck down the hall. Without that diagonal line, the card-sorting job was challenging – the resulting melt-downs were not pretty.
And as students invariably discovered, hanging out in the computer center after 10pm had great advantages because the turnaround time from job submission to the output bins was at a minimum.
Hah. Except for the “standalone” time by the OS developers working on improvements. I became one of those myself on CMU’s PDP-10s. At least they were timesharing systems and we didn’t have long overnight batch jobs.
I did much of my computer work late at night and early in the morning, to avoid crowding and resulting delays. Some of the late-night geeks were truly weird – guys who could communicate with machines, but not with people. We had an IBM360, which they viewed as a lesser God, or maybe not so lesser.
Some idiot foreign activists at a Montreal university destroyed their mainframe computer in a protest, and our geeks went into mourning for days – sackcloth and ashes, wailing and moaning, rocking and rolling, tearing their hair and gnashing their teeth – the whole enchilada…
If we could gather all those who took Fortran in community college, from their assisted living facilities, and flocculate them in just one, we could hear the audio version of these comments…
in fact if you listen or read carefully they are just saying we are <i>convince</i> they our results are good/robust/useful to demonstrate that a climate catastrophe is not impossible..
Well the weather models are showing that the potential for some severe weather hitting the Texas and Oklahoma panhandles this afternoon and this time I think they have nailed it.
And of course if there is a severe tornado outbreak, and it sure is looking likely, it will be hyped as evidence of climate change by some fools.
Thanks, rah. Although I did love the idea of a “severed tornado outbreak”, I fixed the typo …
w.
Thanks!
Hi Willis,
Excellent article as usual.
Would you be able to provide a link to the original of the climate model tunable parameters table please?
Thx TS
Sorry, the image is lost in the flood, but there’s a paper about them here.
w.
Much obliged, sir!
I am in awe of your knowledge and experience with computers and programming Willis and couldnt agree with you more! (I am not a programmer however). I think probably by far the most accurate model is a standard ruler and pencil. Just extend the line forward from the top and bottom of all the sinusoidal curves from the past to work out the range and put in a median line!!
On a short timeline this will show a possible increase but over the longer timeframes the range will probably show a decrease and the upward limit of the shorter timeframe! Certainly will produce a result considerably higher in credibility than the spaghetti mess of ‘computer modeled’ prediction!
A really interesting and fascinating look into your experience with computers, programming and modeling. Bravo Willis!
As I said in a comment on previous posts, climate models are wrong, but that does not mean they aren’t useful. If only we could construct a second planet earth to do experiments on we’d never need another one. Alas.
They are actually the ABSOLUTE OPPOSITE of USEFUL (except for propaganda)
The fact that they are provably WRONG and yet they are still used….
…. is causing a whole heap of wasted money, environmental degradation, electrical supply system instability, and general human suffering etc around the world.
Stop turning a blind eye to the IMMENSE DAMAGE DONE by these provably WRONG climate models.
OMG, when you fall back on that little piece of anti-science nonsense…
you show that you really have been taking the Klimate Kool-aide intravenously.
A deep insidious infection becomes obvious!
OK, enough seriousness. Time for programmer in-jokes. I’ll start.
Back when memory was limited, a shoe company goes to a programmer to automate their inventory system. He asks them “Are there any special characters in your inventory codes”? Told no, he writes up a system that works and sorts flawlessly for just letters and numbers.
But when they bring in their inventory list, the first entry is “Boot#32$1sz_9” … he says “I thought you said there were no special characters in your inventory codes!”
“Those characters aren’t special”, the shoe guy says, “we have those in all of our inventory codes” …
w.
How about anecdotes? One day at DEC when I was working third shift to get some standalone time, I bicycled into work around midnight. I went home for breakfast and changed into business attire, drove back to DEC for the bus trip to the DECUS User Society meeting that day. Then managed to miss the bus back that evening. So I joined some coworkers who were staying late, had dinner with them, talked to customers, a friend drove me back to DEC and I drove home, getting back – – – around midnight.
It was the first (and only!) time I spent a full day at work.
One time we needed to burn a whole bunch of PROMs to ship to our customers. We were a development site, not a production site, so our one and only PROM burner could only burn one set at a time I also couldn’t get started until everyone else left for the day because of ongoing development work. I was warned the day before that I was going to have to do this so I brought about 3 books to work with me.
Insert a set of PROMS, press the start program button. 20 minutes take the PROMS out and slip them into their respective tubes. Put the next set in, press the programming button. And so on, all night. I finally finished right about the time the boss pulled in. I handed him the tubes and let him know that if anyone tried to contact me that day, they would come in to work the next day about a head shorter.
did that same kind of shit working for Toyota. It’s a funny story that i hope nobody ever repeats. Except i know they will.
Yes!!
My son had a summer job where he took calls re technical problems. A good number of people had to be guided on the phone to a certain button that caused the screen to light up.
Then there was a weekend where he went speechless after a number of calls like that and had to go home to recover.
Aw, I was convinced it was a photo of the memory from the Apollo guidance system 🙁
For the curious amongst you, there is an excellent series over on youtube covering a restoration project of one of the AGMs: https://www.youtube.com/watch?v=2KSahAoOLdU
Well worth a watch. It contains plenty of closeup of its memory modules and I was absolutely fascinated by the repairs they did.
Kildall’s CP/M was an OS not a computer language. Ask Bill, he copied it to write MSDOS . He changed the slash to backslash, so no one would notice. 😉
No, CP/M was taken from DEC systems (OS8/RT11/whatever) that used ‘/’ for specifying options.
Then Bell Labs wrote Unix, and they used ‘/’ to separate files in path names and ‘-‘ to specify options, e.g. “kill -f /bin/laden”.
When Microsoft wanted a multi-level file system, they copied Unix’s layout but changed ‘/’ to ‘\’.
Wonderful post, thank you for doing it. I love reading about your personal experiences, which are always fascinating, and your technical exposition is both dead-on and yet highly understandable. I’m almost 67, and had many of the same kinds of computer experience as you early on; ah, the good old days!
I have wondered about another potential source of error in climate model calculations of the climate state in 100 years, and have researched it enough to think it possible but not enough to find out whether anyone has actually tried to find an example. Semiconductor memory and logic circuits are susceptible to bit errors, and though the bit error rate is extremely low in modern computers, it isn’t zero. Error detecting/correcting firmware exists to handle a corrupted bit, but the more better it is, the more additional overhead it places on every computation. A study I read found that supercomputers of the type used only limited error checking/correction to trade for speed. There are classes of bit error they cannot detect. So the authors reported a series of test runs they performed with one computer, using external monitoring to check for the occurrence of any of those type errors. They found that the rate of such errors was low, it was not trivial.
I hark back to an article posted by Judith Curry citing a set of climate runs on one computer in which the initial temperature conditions were altered by 1 trillionth of a degree, and the run results began to drift apart until eventually they were completely different. With the stupendous number of computations performed in a 100-year simulation, on a high spatial resolution grid, the types of bit errors that evade supercomputer check/correct would have a high probability of altering some of the numbers by enough to affect results, but not so much as to be noticeable when they occurred in an output review.
It would be simple enough to test, though expensive. Just do consecutive 100 year runs of a climate model on the same computer, using identical initial conditions, and see if they differ from one another. It may take several runs to get some statistics, and that, as I say, would be very expensive. Is it worth it? I don’t know.
Even cheap computers include parity bits with every byte. A singe bit being changed would trigger the guard circuits to shut down the computer.
More advanced computer and all mainframes, include error detection and correction circuitry. A singe bit error would be detected and repaired without interrupting computer operation. With the simple EDC circuits, 2 bit errors would be detected, but they couldn’t be corrected.
Perhaps I shouldn’t have tried to simplify my presentation so much, since you’ve simplified the problem out of existence – except that it still exists. Parity bits were not included in early PCs, or even in Seymour Cray’s first supercomputers. Now they are, but to what effect? The multi-billion dollar Littoral Combat Ship relied on simple parity checks in its fire control system, and wound up shutting down every few minutes. You’re right in that simple EDC circuits could detect even bit errors, but not correct them, which is kinda the point here. There are classes of errors that need higher overhead EDC, and those higher overhead EDC algorithms are not included in the kind of computers used in climate modeling.
Given the huge amount of memory in a computer used for climate modeling, and the extremely long run times for a century climate prediction, the odds of even-bit memory errors go up tremendously. If they can’t be corrected, then they propagate.
And it’s not just memory errors, but errors along “noisy” data channels – such as the buses that feed the CPUs. And the CPUs themselves are prone to bit flips.
Google has found a much higher incidence of memory errors than are predicted by EDC models. Fortunately, these are in situations where data are stored statically, and redundantly. In a marching algorithm, there is no way at present to stop things and correct even bit memory errors, and no way to even detect certain bus noise errors or CPU bit errors.
>>
Heck, we don’t even know if the Navier-Stokes fluid dynamics equations as they are used in the climate models converge to the right answer, and near as I can tell, there’s no way to determine that.
<<
The Navier-Stokes Equation is one of the Clay Mathematics Institute’s Millennium Problems:
“Navier-Stokes Equation
This is the equation which governs the flow of fluids such as water and air. However, there is no proof for the most basic questions one can ask: do solutions exist, and are they unique?”
You can win a million dollars if you find a proof.
Jim
Willis, that is a great analysis of the state of modelling with respect to climate “science”.
The top “climate scientists” and politicians should be required to read it. I know that some will peek here.
I will send it to people who should read it.
You articles are always unbiased and form conclusions based on evidence.
Meanwhile, I did not see our first computer mentioned. I waited a bit and bought the Apple IIGS which was great for graphics and video. It helped our older son get into IT. After that purchase I think the subsequent computers cost less.
I only did some Fortran post university in the 1970’s but did not get into it. I think I would like to learn R..
It was humorous at work in the 70’s. Some could talk the whole shift about Ram and Rom and were really excited when an outfit doubled capacity from 1024 to 2048 kb or something like that.
The last paragraph may have been in the 1980’s.
I’d just like to add that while modern computational fluid dynamic models are indeed useful tools for aircraft design, they do so ONLY for modeling specific regimes of flight where air flow is relatively well behaved (i.e. they work well for modeling how typical commercial passenger jets are flown). As soon as you try to model high angles of attack (turning and burning in jet fighter parlance), all bets are off as turbulent flow becomes dominant and we have no good ways to accurately model turbulent flow. So, if you think airflow and ocean currents are important to climate modeling, I think it is pretty obvious we have a problem here.
A great example I remember reading about some years back was a buoy study run to map currents in the Atlantic ocean. They deployed a number of buoys that sank to target depths to cruise along, then popped to the surface to get a GPS fix, radio that in, and then drop back to depth on a regular basis to map their routes. The expectation was that they would all flow from the arctic where they were deployed down the US Eastern seaboard along the much lauded “conveyor belt.” But in reality they ended up scattered all over the Atlantic in an almost random pattern that surprised everyone. So I have always considered climate modeling a fools errand, and my only surprise has been the number of fools we have purporting to be scientists.
Just imagine a model of a working internal combustion engine where it shows: the inputs of fuel, air & electricity; a rotating shaft representing the output; and a black box with a simplified mathematical equation to approximate the output based on the inputs; no explanation of the pistons; no explanation of actual combustion & expansion; no inner workings sufficient for replication.
A climate model is like that, more like a black box full of complexity but approximates some larger functions and glosses over the smaller intricacies with parameters based on averages. No explanation of the small scale interactions which lead to weather, no inner workings of the climate ‘motor’. The daily output could then be averaged to accurately model long term climate. But all we get are guesses multiplied by more guesses and parameters adjusted until the output matches their assumptions.
I think in 2019 it was said that: If you used every supercomputer in the world (top500) combined to accurately model the climate it would take about 100 years to model 100 years.
Willis, great article, thanks.
You say “Next, all of these models are “tuned” to represent the past slow warming trend.”
Isn’t it the case that this slow warming trend has itself been “tuned” to match CO2 changes in the atmosphere through the temperature adjustments that have been made by those responsible for the official records.
Hence the models assume that CO2 is the main driver and they are tuned to a temp curve that reflects this assumption. They are truly GIGO, except that the results are treated as Gospel!
IMHO the IBM PC wrecked the microcomputer market – there were dozens of companies making micros but ‘business’ wouldn’t buy them until they had the ‘badge’ even though the first PC was very plain vanilla (until the XT and a hard drive). Taught me a life lesson about restricted corporate thinking causing mediocrity. My road less traveled:: Intel 4004, Assembly, Fortran, Basic, Pascal. A hunt and pecker typer at the punch card machine -> a long engineering formula with a lot of parens would take multiple tries for me. Really enjoyed this essay by Willis – much thanks.
Thanks for another great article. As one other commenter noted, you took a very complex topic and distilled it in a way non-technical people like me can understand. That is a rare talent. I have a modest technical background with some minor programming experience, but nothing remotely close to virtually everyone else posting here. The one similarity with many is my age – my first experience with computers was with a Data General with (if I remember correctly) 512 k memory and 2.5 MB (memorex?) disk drives. When the large washer machine size 25 MB multi-platter disk drives came out, we thought no one would ever run out of disk space.
Anyway, based on what I have read about models, my description of climate models has been reduced to a simple statement. Climate models are used to manipulate data to achieve a predetermined desired outcome and present it as fact. That, IMO is not science, it is fraud.
I look forward to your next article.
Great nostalgia……I am 77, so just a bit younger than you……so I did the punch card drill in the mid-60’s doing my zoology Masters, did my PhD stats on a Programma101 and an HP9100 desktop in the late 60’s…..as I returned to work in an Ottawa government lab in 1971, I talked my director into buying a Wang shared desktop calculator system (4 terminals on one central processor all hard wired together….way better than the standalone calculators then available)……a few weeks later though the local HP sales guy walked in with the first HP pocket calculator (~$1000)…..and right after that Xerox Data Systems (XDS) rep sold our lab a time share subscription via telephone modem to Xerox PARC in Calif….this worked really well, but very slow compared to today, and the phone bills were scary…..and we could only print out the data on a tractor feed terminal (with a paper punch tape for memory)…..no video display, but we could run our stats programs in near real time. And by the mid 70’s our Ag Department ran a similar time share facility based on an Ottawa minicomputer for its labs across the Canada; this was overtaken by the rise of “micro” computers like Apple ][e and then IBM PCs, and the rest is history.
With my kids at home we had a friend with a TRS-80 that induced me to get them a VIC-20, and very soon after a Commodore64….I was never much of a programmer, but have always been close to the industry and its developments for use in my field work research on soil fauna.
One of our research necessities was taking soil and air temps on a continuous basis for which we used Campbell Scientific data loggers with wired temp sensors; initially we set their sampling frequency way too high (say every 10 seconds) because we thought all that data could now be readily managed on our PCs….no problem. Except we quickly discovered that driving our vehicle too close to the air temp sensors in the Stevenson Screen to collect the data every couple of weeks caused a spike from the hot vehicle engine, and even from the body temp of the person collecting the data. The simplest fix was to increase the read frequency to about 10 minutes rather than trying to mess around “cleaning” up the high frequency data.
Which leads me to a couple of observations about scientists from different generations: When I started my research career as a grad student in the mid-60’s there were still a lot of old science guys who had begun their work in the 1930’s still around, and me being a big history buff, I was always happy to listen to their stories (Canadians, Americans, Brits, French, Aussies)…..and I was always most impressed with their attention to detail and their meticulous record keeping (diaries, log books, field notes). So I am just appalled when I see, for example, all those wonderful old temp records prior to about 1970 “adjusted” by some computer algorithm to fit some pre-conceived notion of what is surmised to be going on now.
My other observation about the current generation of scientists is that many of them are far too happy doing their research “on the computer”. This was something that I, then almost a grey beard in the early 1990’s, noticed and remarked upon at the time along with my peer colleagues. Whether doing lab or field research, actually collecting data from well designed experiments is critical to the integrity of science. Subsequent analysis and interpretation are important, but early experimental steps define science integrity.
Willis…..about 4 lines from the end of your excellent essay, should not the “be” precede “evidence”?…….cheers……..Alan T
Thanks, Alan, typo fixed. Of course, Brian will take that typo as evidence that I’m not a “real” writer …
w.
Excellent article, Willis.
I see there are 456 comments below for me to read.
Upon re-reading this, I realized I’d left part of my computer history out. So I added this to the head post …
Onwards, ever onwards,
w.
Willis, I enjoy reading your articles. They always provide clear insight into complicated problems. You have about a decade more experience than me (I was limited to Algol, Fortran, and Basic – I never really got into C or C++ although I managed a large software effort where the programmers used C). I will add a comment to the Verification & Validation (V&V) portion from the Systems Engineering perspective which I think is applicable to any engineering endeavor. We would start with a customer need for which we would create a “system” to satisfy that need. We then wrote a set of cascading requirements (Mission, System, Major Subsystems, and lower level subsystems). As part of this we would review the work to ensure that the Requirements we defined will satisfy the customer’s need and then we reviewed our work to ensure what was implemented (actually created) covered all the Requirements we defined. A lot of models were used in this effort but all the models had to eventually be checked against the real world. Real world checks were not ‘optional’.
Thank you for your contribution to the enlightenment of future generations.
complete drivel – the giveaway is this – So if I think CO2 is the secret control knob for the global temperature, the output of any model I create will reflect and verify that assumption.
Unfortunately for the author, Arrhenius had this sorted out a long time ago. Engineers should try to learn some science before pontificating about things that they dont understand.
>>
Engineers should try to learn some science before pontificating about things that they dont understand.
<<
Back in Roman days, Roman engineers were considered part of the common working class, while Roman scientists considered themselves part of the aristocracy. Because of the class differences, Roman scientists wouldn’t be caught dead talking to Roman engineers and vice versa. That was lucky for the rest of the world, because Roman scientific knowledge was sufficiently advanced that had they talked to their engineers, Roman weaponry would have been far more formidable than it was.
Many years ago, I was told by my engineering professors that Roman snobbery was no longer the case. In modern times, engineers and scientists freely talk to each other, and they recognized each other’s strengths. Modern society has benefited. In recent years, I have noticed an increase of anti-engineering bias by so-called scientists. In fact, it’s so bad now that I consider things like “climate science” to be an oxymoron. And my original assumption is confirmed by each passing day.
Jim
Arrhenius had the relationship between computer models and computer modelers “sorted out a long time ago”?
Who knew?
w.
I guess I’d say: Great minds rarely follow the same rut. I did follow a simiar rut, though.
Regarding fluid dynamics- I’m sure Boeing’s procedures work, at least for them.
i have a friend,a pilot in the military and last notice flying for Braniff(he’as old as we are).
A friend of his, probably 20 years ago, decided he was going to try for a speed record in a homebuilt plane. Jim was supposed to check the aerodynamics with the current “best” cfd program they could get. The design turned out to be a very small canard with the pilot crammed in in front of the wing, the motor behind, as light as possible. It didn’t even have rollbar protection for the pilot.
Jim gave him his results and it pointed to the canard being heavily loaded but within a range of speeds, weights, and wind it should be fine. The pilot/builder took it and ran. The plane was built and ready for test. They were running a 2 week flight test program which verified the speed performance and it handled pretty much as predicted.
Near the end of the second week of testing a thunderstorm popped up some miles away. The pilot wanted one last flight that day, so up he went. It was again successful and he brought the plane in. As he slowed down to land a heavy gust, estimated at 25mph or more, hit the plane at about 45deg on the left. The left canard lifted and started to roll the plane(the right canard was in the lee of the fuselage and had virtually no lift). The pilot simply couldn’t keep the left wing and canard down. In less than a couple seconds the plane flipped over and crashed.
The pilot died instantly due to the lack of a rollbar.
There are a bunch of mistakes there, primarily on the part of the builder/pilot. At the time canard homebuilts had some popularity. But starting out with a completely new design without an adequate design review by a bunch of knowledgeable eyes a CFD program alone just wouldn’t do the trick.
The hubris of the builder/pilot is matched, I think, by many of the scientists involved in climate models. Computers make it way to easy to make people with limited experience and skills think they can do miracles. Especially when lots of cohorts, $$$, and fame are deathly attractors. (pun intended).
Sorry for the length. Hope all your(as in everybody) has better luck!!! and results.