National Geographic: Google AI can be Used to Spy on Chinese CO2 Emissions

Climate modelers do it with digital crytals balls

Guest essay by Eric Worrall

They don’t actually mention China by name, what they say is “This can be used worldwide in places that aren’t monitoring, and we don’t have to ask permission.”.

How artificial intelligence can tackle climate change
The biggest challenge on the planet might benefit from machine learning to help with solutions. Here are a just a few.

BY JACKIE SNOW
PUBLISHED JULY 18, 2019

Climate change is the biggest challenge facing the planet. It will need every solution possible, including technology like artificial intelligence (AI).

Seeing a chance to help the cause, some of the biggest names in AI and machine learning—a discipline within the field—recently published a paper called “Tackling Climate Change with Machine Learning.” The paper, which was discussed at a workshop during a major AI conference in June, was a “call to arms” to bring researchers together, said David Rolnick, a University of Pennsylvania postdoctoral student and one of the authors.

“It’s surprising how many problems machine learning can meaningfully contribute to,” says Rolnick, who also helped organize the June workshop.

AI can also unlock new insights from the massive amounts of complex climate simulations generated by the field of climate modeling, which has come a long way since the first system was created at Princeton in the 1960s. Of the dozens of models that have since come into existence, all look at data regarding atmosphere, oceans, land, cryosphere, or ice.

But, even with agreement on basic scientific assumptions, Claire Monteleoni, a computer science professor at the University of Colorado, Boulder and a co-founder of climate informatics, points out that while the models generally agree in the short term, differences emerge when it comes to long-term forecasts.

“There’s a lot of uncertainty,” Monteleoni said. “They don’t even agree on how precipitation will change in the future.”

One project Monteleoni worked on uses machine learning algorithms to combine the predictions of the approximately 30 climate models used by the Intergovernmental Panel on Climate Change. Better predictions can help officials make informed climate policy, allow governments to prepare for change, and potentially uncover areas that could reverse some effects of climate change.

grant from Google is expanding the nonprofit’s satellite imagery efforts to include gas-powered plants’ emissions and get a better sense of where air pollution is coming from. While there are continuous monitoring systems near power plants that can measure CO2 emissions more directly, they do not have global reach.

This can be used worldwide in places that aren’t monitoring,” said Durand D’souza, a data scientist at Carbon Tracker. “And we don’t have to ask permission.

Read more: https://www.nationalgeographic.com/environment/2019/07/artificial-intelligence-climate-change/

Nice to see admissions of deep uncertainty and the need for better predictions to make informed climate policy.

But there is a problem the AI experts are kind of glossing over.

I’ve worked on AI before, the most powerful AI I created extracted text from very noisy images, like the following:

Fake Aussie Driver’s License

What is the big deal? The text is easy for a human to read.

But an AI really struggles to read such images. The reason is, AIs find it very difficult to differentiate between the text, and the color splashes and swirly lines.

To have any chance of the AI successfully reading the text in the image, a lot of human intervention is required; a human has to try to develop a digital filter which cleans as much of the distracting swirly stuff away as possible, leaving an image something like the following:

Fake Aussie Driver’s License (processed)

This incidentally is why websites sometimes give you tests to prove you are not a robot. Most humans can easily pick out the text from a noisy image on a “Are you a robot?” test. AI web robots not so much.

My point is current generation AIs usually do a very poor job of separating noise from data. In the same way a text reading AI can be confused by a few swirly lines, so cutting edge facial recognition AIs can be confused by wearing funny patterned glasses. Tricks which wouldn’t fool a human toddler leave the most powerful current generation AIs floundering.

So I’m really looking forward to seeing the results of AI researcher’s optimistic attempts to use AI to pick useful insights out of the noisy mess which is the world’s climate models.

Advertisements

36 thoughts on “National Geographic: Google AI can be Used to Spy on Chinese CO2 Emissions

  1. Here’s the problem: if it is true “AI” then it will think for itself and may not give the answers they were looking for. In the “climate change” discussion, for example, a true AI would determine that there is way too much uncertainty and needs more accurate input.

    • Great! Instead of measuring CO2, let’s model it – and throw in some AI to make it more progressive. That approach actually does not need any “accurate input”.

      • You don’t even really need AI, though adding in the buzz word obviously helps.

        Any methodical system for combining IPCC models based how well they have performed at matching temperature rise and variability over the last 20y would be useful and informative instead of moronically taking the arithmetic average of a tonne of garbage.

        A straight statistical weighting based on performance would reduce the influence of the poor, high sensitivity models which are currently used to justify all the “may be as a high as” types of claims cherry picked by the media to influence public opinion.

        Authors of models which perfom badly would then be incentivised to fiddle with their fudge factors and produce runs which are closer to reality instead of just misleading everyone.

    • Dunno how serious you were, but current AI systems are narrowly focused on one particular problem–the one they were trained on. Face recognition, OCR in noisy environments, speech recognition, playing poker or Go or chess, finding certain kinds of things in natural language texts, and so forth. Some of these systems are remarkably capable of doing what they do.

      I’ve run topic modeling systems on texts (PubMed abstracts), and the system has learned–on its own and without any supervision–to cluster those texts by topic, where topics are defined by sets of words. The topic modeling system chose the words–I didn’t tell it what to look for–but they are remarkably coherent sets. And yet there’s nothing I would call intelligence in there; it’s essentially looking for what words commonly occur with what other words in documents. Linear algebra, basically (at least that’s one algorithm).

      All these are problems that one would have thought, ten or twenty years ago, would require general intelligence (the kind any human being, smart or dumb, has). But what the AI system is doing is actually not general. Throw a face recognition problem at my topic tracking program, for example, and you’ll get nothing. So no, an AI system won’t “think for itself.” We are (IMO) further from that than we are from nuclear fusion.

      • Did you supply the algorithm with rules for the choice of words that identify the topics?

  2. I was wondering when I saw this thing about NASA ans the CO2 viewer when it would come to seeing and admitting this. The first thing they say of course is they can see CO2 over the US. Please. We emit hardly any in comparison to China and everyone knows it but says nothing. Meanwhile they want us to give up cars, plastic, and toilet paper for our imagined sins.

  3. …differences emerge when it comes to long-term forecasts.
    Of course they do. It’s a chaotic system, therefore unpredictable more than about four to six weeks ahead.
    Don’t they learn anything?

  4. Personally I’m fascinated by the language used here (and everywhere leftists spew their drivel for that matter). Apparently planets can “face challenges.” That will be news to both geologists and astrophysicists I’m guessing.

    Cancelled my subscription years ago when I realized that they’d become just another propaganda organ for despisers of Western civilization.

  5. You mean AI is going to make sense out of the gobbldegook that makes up today’s climate science? Going to fine-tune the Global Climate Models, are they? The AI would be starting out with “Garbage In” so where are they going from there? My guess is “Garbage Out”.

  6. Is it possible to write a guest post to say thanks to WUWT for helping spread the truth about the climate? I feel I need to get the word out to all of the WUWT how this site saved me from pure insanity following a climate debate gone wrong.

  7. Surely Google wouldn’t do anything that their beloved Chinese Communist Party wouldn’t want them to do like misleading white devils about the salubrious climate of Beijing?

  8. AI can also unlock new insights from the massive amounts of complex climate simulations….

    …what “insights”…..simulations that are wrong have wrong insights

  9. JACKIE SNOW
    PUBLISHED JULY 18, 2019
    “Climate change is the biggest challenge facing the planet.”

    MY COMMENT:
    In just one sentence — the first sentence — “Jackie Snow” has identified herself / himself as a climate science imbecile, even faster than wearing a T-shirt that said “I am the Village Idiot”, which could have been misinterpreted as a self-deprecating joke.

    So I stopped reading what Snow was babbling about after that first sentence.

    I really appreciate when writers quickly identify themselves as complete fools.

    • Climate change is the biggest challenge facing the planet.

      It’s absolutely true. The dirty thirties nearly did in my dearly beloved mother.

      Change! It’s what the climate does.

    • The biggest challenge facing the planet is phlogiston depletion causing the interstellar ether to leak into our atmosphere.

      Our so-called leaders refuse to take any action on this existential threat. Some of them even go so far as to deny that there is a problem! We can see the effects around us every day—97% of my alternate personalities agree that nearly all of the phlogiston depletion observed since 1950 has been caused by demonic flatulence.

      I need a more powerful gaming computer, er AI computer, to detect the witches who harbor demons before it’s too late! Send money! Quickly! Think of The Children.

  10. “is expanding the nonprofit’s”??? Somehow I find it very difficult to believe that “profit” is not being made somewhere down the line by somebody! As my late mother used to say when faced with a dubious prospect, “And the band played ‘Believe it if you like!’ ” With the trillions of $ invloved in this globul scam it would seem highly unlikely!

    • Funny thing: My wife and I like to watch various house hunter series as sort of travel logs regarding various locations around the US and the world. It was surprising to see the number of people moving into expensive cities in the US and around the world with HIGH housing budgets who work for “nonprofits”. Employees of nonprofits sure seem to make a hell of a lot of money. But as is mentioned often on WUWT, if you subsidize it, it will become more expensive. By their nature nonprofits are subsidized by you and I due to the tax status of donations, and often directly funded by the US government. Ex. Planned Parenthood. There are THOUSANDS of nonprofits provided grants from the US government.

  11. Won’t this wonderful AI run a ruler over the IPCC’s models and declare them faulty? Unarguably?

  12. The Chinese don’t care what people can see. They have a licence to produce as much CO2 as they like until 2030 by which time everyone will have realised the truth and lost interest in CO2 except as plant food.

  13. Having been in IT for over 20 years, I can say with full confidence that I trust a program to do what it is told/programmed to do, not what I or anyone else wants it to do.

  14. …“There’s a lot of uncertainty,” Monteleoni said. “They don’t even agree on how precipitation will change in the future”…

    Most of the emphasis on model “agreement” is limited to global temperature anomaly. Nobody cares or wants to discuss the fact that it is the collective sum of a bunch of garbage and errors regionally and locally. As long as it adds up to a small amount of warming, it’s a-ok.

    And globally, regionally, and locally with so many other parameters such as precipitation, they are all over the place…and far from reality.

  15. First of all, there is no such thing as artificial intelligence, unless one believes that massive lookup tables are the same as intelligence. When the “boffins” say they are training the program, they are building tables by running through as many scenarios as possible, storing the “weights” and then using them to try and let the system approach an answer. It gets faked out by the glasses, or warpaint on the face, as these are perforce, information that is different from what was trained, and therefore important. In this way a turtle has been identified as being a rifle, systems unable to differentiate between a tiger and a plowed field, etc.

    Just think of your tables of logarithms, or squares, or trig function values instead of Artificial Intelligence and you’ll be able to see if anything new is really occurring.

  16. “There’s a lot of uncertainty,” Monteleoni said. “They don’t even agree on how precipitation will change in the future.”

    Nakedly the words of somebody who clearly recognizes one of the specific reasons why global-warming alarmism is a busted flush, but still goes along with the general belief in order to promote their own (funding) opportunities.

  17. “It’s surprising how many problems machine learning can meaningfully contribute to…”

    Well, Mr. Rolnick, you certainly got that right. It’s been true ever since “machine learning” was understood as the outcome of machine teaching. At least the student victims of the various forms of teaching machines bring real intelligence to the task.

  18. There is a fine line between Artificial Intelligence (AI) and Artificial Stupidity (AS).
    IMO there is a lot of the latter flying about.

    • Exactly what I thought. China has zero obligations to cut anything, what is the point of monitoring what they do. I suppose it may be useful for proving how futile US and EU emission controls are, so that’s valuable.

  19. Spy on them? They are doing very little to hide what they are doing, almost as if they don’t care that certain people don’t like it. Imagine that.

  20. To anyone inclined to rely on AI for anything critical, I suggest spending a few hours watching TV newscasts with Captions for the Hearing Impaired enabled. I’ve been doing this for some years now, as a means of improving my understanding of French, Belgian, and Quebecois news and documentary broadcasts, and have gotten so used to it that I’ve now left captions enabled on American, British, and Canadian broadcasts as well.

    The worst voice to text conversions are always those of live newscasts, which (be warned) are characterized by a very annoying and obstructive time lag of 15 to 27 seconds (on my most frequently viewed news broadcasts, BBC World News, PBS Newshour, and CBC News Network).

    NB: I ONLY watch tv via PVR, which allows me to pause, rewind, review and take a screen shot as needed, and last but not least, to fast forward through the noise and nonsense.

    I’ve chosen a very simple example to clearly demonstrate the state of current AI available to the BBC.

    This example was aired on the July 22nd broadcast of BBC News America, and the instance cited occurred 4 min 40 seconds into the 22:00 broadcast of the Seattle PBS station.

    The announcer reporting the (Iranian tanker seizure) incident is a regular BBC anchor/reporter, and was speaking slowly and clearly, and the words that were confused are pronounced exactly the same in standard British and American English, “sails” and “sales”.

    The AI rendering of the spoken report is:

    “…JULY 4 FOR ALLEGEDLY VIOLATING SANCTIONS ON SAILS TO SYRIA…:

    As I have some experience with simple electronic dictionaries that use word lists sorted by frequency of use to present options to the user, I conclude that the level of AI employed in this case was slightly more sophisticated than this. Ironically, the simple frequency word list would like have given the “correct” result.

    My guess is that the AI involved used a frequency of association list instead, and took the context of shipping to decide that “sail” was the correct interpretation of the spoken word. It was not intelligent enough to detect the rarity of “sanctions on sails” vs “sanctions on sales” as an expression, much less to grasp the larger context.

    To call this “intelligence” is simply absurd.

    And in almost five years of reading such nonsense (and this is a very mild example), I’ve been unable to detect any improvement at all.

    PS. seeing no option to post a screen save here, I looked on the BBC website for the URL of the story in question, and could not find any trace of it by date of broadcast, “tanker seizure”, “Stena Imperio”, etc.. Indeed, I got zero hits for all of the above. There appears to be a serious lack of INTELLIGENCE in the BBC’s website search engine as well…

Comments are closed.