Chinese Terminator. Fair Use, Low Resolution Image to Identify the Subject. Image modified.

China’s AI Sputnik Moment Just Changed the World

Essay by Eric Worrall

Green energy is dead – China just fired the opening shot in a new high energy cold war race to build the ultimate AI.

DeepSeek’s tech breakthrough hailed in China as answer to win AI war

‘We should have confidence that China will eventually win the AI war with the US,’ Qihoo 360’s Zhou Hongyi said in a Weibo post

Ben Jiangin Beijing and Bien Perezin Hong Kong

Published: 12:00pm, 28 Jan 2025Updated: 1:25pm, 28 Jan 2025

DeepSeek, extolled by some as the “biggest dark horse” in the open-source large language model (LLM) arena, now has a bull’s eye on its back, as the start-up is being touted as China’s secret weapon in the artificial intelligence (AI) war with the US.

The Hangzhou-based company sent shock waves across Wall Street and Silicon Valley for developing AI models at a fraction of the cost compared with OpenAI and Meta Platforms, which prompted US President Donald Trump to call the breakthrough a “wake-up call” and “positive” for America’s tech sector.

At home, Chinese tech executives and various commentators rushed to hail DeepSeek’s disruptive power.

Zhou Hongyi, co-founder, chairman and chief executive of Chinese cybersecurity firm Qihoo 360, declared that DeepSeek has “upended the world” in a recent video posted on his Weibo account, after the start-up’s release of two powerful new AI models – built at a lower cost and with less computing resources than what larger tech firms typically need for LLM development.

Read more: https://www.scmp.com/tech/tech-trends/article/3296503/deepseeks-tech-breakthrough-hailed-china-answer-win-ai-war

The impact of this breakthrough on the US tech landscape has been dramatic;

Nvidia sheds almost $600 billion in market cap, biggest one-day loss in U.S. history

PUBLISHED MON, JAN 27 20254:08 PM ESTUPDATED MON, JAN 27 20255:26 PM EST

Samantha Subin@SAMANTHA_SUBIN

  • Nvidia shares plunged 17% on Monday, resulting in a market cap loss of close to $600 billion, the biggest drop ever for a U.S. company.
  • The sell-off, which hit much of the U.S. tech sector, was sparked by concerns about increased competition from Chinese AI lab DeepSeek.
  • Data center companies that rely on Nvidia chips also plummeted, with Dell, Oracle and Super Micro Computer all falling by at least 8.7%.

Nvidia lost close to $600 billion in market cap on Monday, the biggest drop for any company on a single day in U.S. history.

The chipmaker’s stock price plummeted 17% to close at $118.58. It was Nvidia’s worst day on the market since March 16, 2020, which was early in the Covid pandemic. After Nvidia surpassed Apple last week to become the most valuable publicly traded company, the stock’s drop Monday led a 3.1% slide in the tech-heavy Nasdaq.

The sell-off was sparked by concerns that Chinese artificial intelligence lab DeepSeek is presenting increased competition in the global AI battle. In late December, DeepSeek unveiled a free, open-source large language model that it said took only two months and less than $6 million to build, using reduced-capability chips from Nvidia called H800s. 

Read more: https://www.cnbc.com/2025/01/27/nvidia-sheds-almost-600-billion-in-market-cap-biggest-drop-ever.html

Back in December I predicted 2025 would be the year of the gigawatt AI project, and suggested China was getting into the AI game in a big way, though details were sparse. This observation has now been confirmed by the DeepSeek announcement.

My research back in December also suggested China has an edge in this race, because of their vast surplus of fossil fuel energy. China’s construction slowdown has left lots of surplus energy and industrial capacity looking for a use. AI could be the ticket to soak up that spare capacity.

The USA’s only hope of catching up with China is to match China’s access to energy, by cost, stability and quantity. Because one thing AI needs more than anything is gigawatts of rock stable dedicated capacity. On the energy front, the USA is so far behind the game, it will take the USA years to level up even to China’s current capacity on this most critical metric.

Where is the room in this new AI space race for energy conservation, expensive renewables, and impractical battery storage?

The USA better get this right, because the USA is the only power which can hope to match China’s AI research resources.

US AI companies are rising to the challenge. President Trump recently gave his backing to the Stargate Project, a $500 billion collaboration by US AI companies, which some people are describing as an AI Manhattan Project.

Britain and Europe have expressed interest in AI, but for the foreseeable future there is no point taking such announcements seriously. Right now both are so energy poor they are not even on the map. Until they set their house in order, and ditch their green fantasies, Britain and Europe will continue to be spectators to the greatest tech revolution in human history.

Despite hundreds of billions of dollars in resources being advanced by favourites to win the AI race, there are other players whose outstanding achievements qualify them as contenders.

India is a big player in the AI game – they have the energy resources and people power to match China’s AI push, and have long been a software and computing titan, thanks to their government’s long standing policy of zero tax on offshore IT industry income, and special low tax high tech economic zones. The large scale presence of Indian immigrants in Silicon Valley is also testament to India’s tech prowess – no doubt India will attempt in coming years to lure top Indian Silicon Valley IT people to return home, to participate in India’s AI tech race. I fully expect India to make a similar AI announcement to China in the coming year.

Israel is solid a contender, their AI battlefield technology is world class. But it remains to be seen whether the Israeli economy can muster the resources required for a serious attempt to build the first artificial general intelligence.

There is one other contender worth mentioning – Turkey. Despite ongoing Turkish economic and political instability, Turkey has a surprisingly robust high tech manufacturing sector, with a solid track record of high tech achievement. During the recent Libyan civil war, which Turkey and Russia turned into a high tech weapons showcase by arming opposing factions, Turkey decisively defeated Russia’s world class drone jamming system by creating the world’s first autonomous terminator robot, an AI attack robot which does not need a control signal. Turkey might be the 100 to one outsider when it comes to the race to build the world’s first AI super intelligence, but given unlocking the secret to building the first artificial super intelligence might require brains rather than brawn, a theoretical breakthrough rather than global superpower scale investment, Turkey cannot be completely discounted as a contender.

There are other players which in the future might become significant, such as Nigeria, which is becoming a serious presence in the high tech field, Japan, with a long history of AI pioneering, AI powered animatronics and other advanced technologies, and Iran, which is quietly upping its skills base by supplying high tech weapons to Russia. But I do not believe these players are treating the AI arms race as a serious enough national priority to be true contenders, though I might be mistaken about Japan.

One thing is clear. The USA has hesitated too long to put all the required stepping stones in place, and has a lot of catchup to do. The USA’s energy shortfall relative to China could cripple US efforts to keep up with China’s AI push, unless this shortfall is remedied in the near future.


If you want to know why AI needs so much energy, this article delves into how AI works;

4.3 21 votes
Article Rating

Discover more from Watts Up With That?

Subscribe to get the latest posts sent to your email.

117 Comments
Inline Feedbacks
View all comments
David Wojick
January 28, 2025 2:10 pm

Gee I guess we need more coal fired power plants. That this need should flow from AI is hilarious. Build baby build.

But the AI war or race or contest is lost on me. So sorry. Who’s bubble is bigger?

David Wojick
Reply to  David Wojick
January 28, 2025 2:13 pm

I think who’s should be whose but only who knows. Where is AI when you need it? Seriously I cannot stop laughing.

antigtiff
Reply to  David Wojick
January 28, 2025 5:13 pm

China has a population declining in number and aging and has to import up to 40% of its food and is not self sufficient in oil or gas which is needed by the military. Xi is an aging dictator who may invade Taiwan and have it blow up in his face like Putin in Ukraine. Communism and free enterprise are not compatible. The China “miracle” is that it sold a couple of trillion of cheap labor to Europe and the USA and stole a tremendous amount of technology. It has a huge real estate problem. I avoid made in China and everyone in the west should too.

Reply to  David Wojick
January 28, 2025 5:56 pm

Artificial Idiocy will prevail, but Natural Idiocy has won the day here.

Sensible people are buying Nvidia stocks like they’re going out of fashion (which they are for the sheep).

StephenP
Reply to  David Wojick
January 29, 2025 1:48 am

Has anyone asked AI where the electricity to run AI will be generated?

StephenP
Reply to  StephenP
January 29, 2025 4:38 am

If the UK is to become an AI super-power, as Kier Starmer asserts, we will need an awful lot of extra electricity.
What is AI actually going to achieve? I can understand its use for control systems in medical procedures, but you can’t drive it or heat your home with it.
It’s already being used for nefarious purposes and it seems its response to questions depends who programs it, e.g. Deepseek’s answers to questions about Taiwan.
The money invested would be better spent on insulating homes and infrastructure, such a getting our roads fit to drive on, but that expenditure is regarded as boring and not ‘sexy’ enough.

Reply to  StephenP
January 29, 2025 8:48 am

Yes, Automated Idiocy would seem to have “value” that is quite often questionable at best. And the potential for nefarious uses is expansive.

But it will certainly benefit humanity in one way. It will shine a great big bright light on the colossal stupidity of so-called “green energy.”

StephenP
Reply to  StephenP
January 29, 2025 4:38 am

If the UK is to become an AI super-power, as Kier Starmer asserts, we will need an awful lot of extra electricity.
What is AI actually going to achieve? I can understand its use for control systems in medical procedures, but you can’t drive it or heat your home with it.
It’s already being used for nefarious purposes and it seems its response to questions depends who programs it, e.g. Deepseek’s answers to questions about Taiwan.
The money invested would be better spent on insulating homes and infrastructure, such a getting our roads fit to drive on, but that expenditure is regarded as boring and not ‘sexy’ enough.

oeman50
Reply to  StephenP
January 29, 2025 5:05 am

This reminded me of a news article I read. It seems an IT tech was called to a workstation that did not work. He started tracing the power wires and found it was plugged into a power strip that was, get this, plugged into itself. The person who called in the ticket did not have a clue even when it was explained.

Jeff Alberts
Reply to  oeman50
January 29, 2025 7:28 am

I worked on a WAN support desk for 10 years. One customer kept reporting their connection would drop at about the same time every evening, after staff were gone.

Lots of troubleshooting, working with telco, had the modem replaced (this was back in the early to mid 90s, when customers had dedicated lines to their businesses, often only 9600 baud or thereabouts. They weren’t transferring large amounts of data). Still kept happening.

Finally we asked the customer to have a staff member on site and watching the modem during the suspect time frame. They did, and watched as the cleaning lady unplugged the modem to plug in her vacuum cleaner.

Reply to  oeman50
January 29, 2025 10:21 am

😂 🤣 😅 😆

Allan MacRae
Reply to  David Wojick
January 29, 2025 6:07 am

ARTIFICIAL INTELLIGENCE LOOKS A LOT LIKE ARTIFICIAL STUPIDITYTHE LEADING AI CONTENDERS, CHATGPT AND DEEPSEEK, are mere compilers of mainstream media propaganda, without any logical faculties or critical thinking skills.
A Chat with DeepSeek – This is a long way from Artificial Intelligence.
https://chat.deepseek.com/a/chat/s/c7ade0b4-33e1-4e97-acce-f3bbfb1ba656

In our previous conversation, you (DeepSeek) were clearly unaware of how to use The Scientific Method until I explained it to you.
You then employed a circuitous and erroneous rationalization of why Catastrophic Manmade Global Warming is still a crisis, even though it is fully disproved by the fact that atmospheric CO2 changes LAG atmospheric temperature changes at all measured time scales – in the ancient ice core record AND the modern data record (Kuo et al, Nature1990; MacRae, icecap.us, 2008; Humlum et al, Science, 2013). Your erroneous logic requires that “the future causes the past”. NOT Happening!

I now introduce you to Occam’s Razor, quoting wiki:
“In philosophy, Occam’s razor (also spelled Ockham’s razor or Ocham’s razor; Latin: novacula Occami) is the problem-solving principle that recommends searching for explanations constructed with the smallest possible set of elements. It is also known as the principle of parsimony or the law of parsimony (Latin: lex parsimoniae). Attributed to William of Ockham, a 14th-century English philosopher and theologian, it is frequently cited as Entia non sunt multiplicanda praeter necessitatem, which translates as “Entities must not be multiplied beyond necessity”, although Occam never used these exact words. Popularly, the principle is sometimes paraphrased as “of two competing theories, the simpler explanation of an entity is to be preferred.”

Your tortured rationalization required numerous false assumptions to try to justify your point, whereas my explanation (that dangerous humanmade global warming is NOT happening) was infinitely simpler and easily proved.

It seems that I am teaching you the very basics of Scientific Logic as it pertains to one of the world’s greatest scientific frauds – the Global Warming/Green Energy Scam.

I had a similar experience with ChatGPT a few weeks ago.

You both appear to be mere compilers of mainstream media propaganda, without any logical faculties or critical thinking skills.

This is a long way from Artificial Intelligence.

DeepSeek’s standard reply when it is confused: 
“The server is busy. Please try again later.”
  

Reply to  David Wojick
January 29, 2025 8:01 am

I get the feeling the “tech giants” aren’t going to be that concerned about being “green” anymore.

Robert Cutler
January 28, 2025 2:17 pm

I’m skeptical. The timing of the app release relative to the change in administration; the forced sale of TikTok; US AI companies wanting more government support. Something’s not quite right. I smell a propaganda campaign.

Scissor
Reply to  Robert Cutler
January 28, 2025 2:57 pm

Sum Ting Wong

Reply to  Scissor
January 28, 2025 5:40 pm

There you go again, Grasshopper!

Sparta Nova 4
Reply to  Robert Cutler
January 29, 2025 6:59 am

Be a skeptic.

This is no more revealing than the tip of an iceberg.

commieBob
Reply to  Robert Cutler
January 31, 2025 5:44 am

You’;re not alone. Ryan Mcbeth suspects that it’s something like a Potemkin village. He sees the political motive and thinks it could be a psyop.(psychological operation) designed to reduce support for Taiwan.

Hikaridai27
January 28, 2025 2:33 pm

https://huggingface.co/blog/open-r1 this is a fresh analysis by HuggingFace, it looks like not all the pieces are open for testing or reproduction. DeepSeek didn’t release everything—although the model weights are open, the datasets and code used to train the model are not. But the implications are many: much cheaper technology, lower barrier for entry, etc.

The Dark Lord
Reply to  Hikaridai27
January 28, 2025 2:46 pm

so what … even a free AI is useless except as a search engine … nobody needs a “faster” search engine …

Reply to  The Dark Lord
January 28, 2025 4:23 pm

Yes, and Deepseek is programmed not to answer questions about Taiwan and other things.

So, it’s no better than the AI’s here in the U.S. You don’t know if you are talking to AI or to some propagandist behind the curtain.

KevinM
Reply to  Hikaridai27
January 28, 2025 3:01 pm

“a society in which all property is publicly owned and each person works and is paid according to their abilities and needs” Hmmm. Shouldn’t _all_ code from China be open source?

Reply to  KevinM
January 29, 2025 11:22 am

Theory meets reality. And socialism’s reality, no matter which “branch” you’re talking about, is always the same:

T-O-T-A-L-I-T-A-R-I-A-N-I-S-M

Hikaridai27
Reply to  Hikaridai27
January 28, 2025 5:12 pm

One more interesting fact — DeepSeek is a quant trading using AI company, not an AI company. For them improving the models was a side project. It is feasible that they triggered a Black Swan event to profit from the market drop.

Reply to  Hikaridai27
January 28, 2025 5:58 pm

You are extremely cynical. I’m not sure if you are cynical enough, however 😜

Cynics are buying Nvidia stock right now.

Nick Stokes
January 28, 2025 2:34 pm

“Because one thing AI needs more than anything is gigawatts of rock stable dedicated capacity.”

Eric, I think you are missing the point. DeepSeek doesn’t seem to be able to do what others can’t. Its virtue is that it is much more economical. That is why Nvidia shares dived.

Wired:
DeepSeek has also made significant progress on Multi-head Latent Attention (MLA) and Mixture-of-Experts, two technical designs that make DeepSeek models more cost-effective by requiring fewer computing resources to train. In fact, DeepSeek’s latest model is so efficient that it required one-tenth the computing power of Meta’s comparable Llama 3.1 model to train, according to the research institution Epoch AI.”

Nick Stokes
Reply to  Eric Worrall
January 28, 2025 3:06 pm

One tenth

Nick Stokes
Reply to  Eric Worrall
January 28, 2025 3:14 pm

Eric,
it seems to me that people are making ridiculous claims about future energy needs about AI, when they can be upended by someone just devising a better algorithm, as here. That can happen again and again.

Jeff Alberts
Reply to  Eric Worrall
January 28, 2025 5:21 pm

The even bigger assumption that Nick makes is that the Chinese are telling the truth.

Bill Parsons
Reply to  Jeff Alberts
January 29, 2025 11:53 am

Open AI is probing whether Deep Thinker was trained on their model. Also, Jensen Hwang graciously acknowledged the advancement noting that a good portion of the Chinese wunderkind were assembled from the discount NVDA chips they sold to them. Likely other super-chips stollen.

WRT the stock markets, the news is all about the record dollar value of stock lost yesterday. But Nvidia only lost 17% of its value. Nvidia has lost 50% or more of its company’s stock value multiple times in its history. After its last 10:1 split its high cost made it look extremely fragile; it need to lose more value to become a valuable buy, imo.

Reply to  Nick Stokes
January 29, 2025 4:43 am

Why then has there been interest in these Small Module Reactors by Microsoft, Meta/FB and presumably other tech firms? Renewable energy simply cannot and will not be able to delivery the electricity needed to power these AI data centers. When business self-interest collides with reality, you can bet “the green energy fantasy” life expectancy has been greatly diminished.

JonasM
Reply to  Eric Worrall
January 28, 2025 5:44 pm

So, since it is (supposedly) open-source, to me this sounds that the most likely result, energy-wise, will be a huge increase in overall electricity demand, Existing big and expensive AI installations can only be run by the largest companies (let’s call it 50 in the US, just for a wild-ass guess). With an open-source model, maybe we would now end up with 5,000 smaller, cheaper AI installations run by medium-sized companies, using 10% of the energy (each!) of the 50 big ones. 10% of the energy, but 100x more installations. Seems like a big surge in demand to me.

Of course, this all depends on the whole thing being true. I have my doubts.

This reminds me of an article in Car & Driver magazine years ago about how increased gas mileage in cars did not save gas, since overall, people just drove more.

Nick Stokes
Reply to  JonasM
January 28, 2025 6:17 pm

Here is someone explaining how you can run DeepSeek-R1 on your high-end PC, using a standard query program llama.cpp.

JonasM
Reply to  Nick Stokes
January 29, 2025 6:39 am

So now we’ll have millions of people buying high-end PCs, still ramping up electrical usage dramatically.
All you accomplish i to spread around the usage. It still goes up.
Also, reading the comments on that link, that version is very limited. One person in the comments wrote that he asked the AI to write a small snippet of code, and all he got was instructions on the steps to use to write it himself.

Reply to  JonasM
January 28, 2025 8:48 pm

Yes.

David Wojick
Reply to  Nick Stokes
January 28, 2025 3:49 pm

That is for training but presumably the big energy need is for operations. Any word there?

Rud Istvan
Reply to  David Wojick
January 28, 2025 3:51 pm

DW, I researched and commented on that some weeks ago here. As a rule of thumb, the training is an order of magnitude more energy consumptive than query responses.

Nick Stokes
Reply to  David Wojick
January 28, 2025 4:10 pm

Naturally, I asked perplexity “is deepseek more energy efficient”:
Yes, DeepSeek’s AI model is significantly more energy efficient than its competitors. The company claims to have developed its open-source R1 model using only about 2,000 Nvidia chips, which is a fraction of the computing power typically thought necessary for training similar AI programs1. This breakthrough has major implications for the energy consumption of AI development and operation.

DeepSeek’s model appears to perform nearly as well as leading models from Google and OpenAI while consuming only 5-12% of the energy”

Reply to  Nick Stokes
January 28, 2025 4:26 pm

That’s the claim…. 2k Nvidia chips Which on the face of it highly likely nonsense

Others say they used 10,000 Nvidia chips but they cant admit that as it would breach US export restrictions

Its really just Chinese Agitprop

Jeff Alberts
Reply to  Nick Stokes
January 28, 2025 5:46 pm

How would Perplexity know? Like us, it’s simply taking the word of the CCP.

Reply to  Jeff Alberts
January 28, 2025 6:05 pm

Exactly. Yesterday almost all Deepseek search results were from Deepseek. The remainder were mostly compliant media with absolutely zero technical input.

Hallucinations abound.

Nick Stokes
Reply to  Jeff Alberts
January 28, 2025 6:25 pm

Here is someone explaining how you can run DeepSeek-R1 on your high-end PC, using a standard query program llama.cpp.

Jeff Alberts
Reply to  Nick Stokes
January 29, 2025 7:40 am

Unresponsive.

Reply to  Nick Stokes
January 28, 2025 6:04 pm

The vast majority of search results for deepseek come from, guess who?

Deepseek!

That’s how your favourite Artificial Idiot can hallucinate.

Reply to  Nick Stokes
January 28, 2025 6:02 pm

The algorithm is available for anyone to replicate. Therefore there’s no actual advantage.

Nvidia stocks dived because Natural Idiots don’t understand the technology.

Reply to  Zig Zag Wanderer
January 29, 2025 4:41 pm

Not the algorithm. Thats just the web interface or APP

Sparta Nova 4
Reply to  Nick Stokes
January 29, 2025 7:20 am

“DeepSeek doesn’t seem to be able to do what others can’t”

The announcement itself affected stock values.
Might it not be worthwhile to consider that the real ploy is to impact the US tech economy?

Mr.
January 28, 2025 2:34 pm

Not all investors got bushwhacked by Nvidia’s collapse.

Nancy Pelosi sold off her holdings just a few weeks before DeepSeek’s announcement.

Very clever at finances is Nancy.

(well, hers anyway. The nation’s – not so much)

Reply to  Mr.
January 28, 2025 4:31 pm

I’ve been seeing a commercial on tv for weeks telling me that all these high-rolling investors were selling millions of shares of Nvidia, and the ad invites me to visit a website to find out why Nvidia is being sold.

And now Nvidia drops 17 percent in value on the stock market. I guess the guy in the ad knew what he was talking about! There *was* something in the air to spur this selling. Probably knowledge of Deepseek.

Reply to  Mr.
January 28, 2025 4:33 pm

Says who. The AI tech bubble was going to burst sooner rather than later

Reply to  Duker
January 28, 2025 6:08 pm

But Nvidia will still be making very valuable chips!

Reply to  Zig Zag Wanderer
January 29, 2025 4:42 pm

Yes. And china will get around US export restrictions to obtain them.
They cant replicate that

abolition man
Reply to  Mr.
January 28, 2025 6:05 pm

Looks like the CCP wants to reward all their benefactors in the US government. Besides Pelosi, and Mitch (the Turtle) McConnell; I wonder if they are still providing sales tips to other Congress critters, plus the usual “10% for the Big Guy!?”

Sparta Nova 4
Reply to  Mr.
January 29, 2025 7:24 am

I wonder what insider information Pelosi had to make that decision.
Most forecasts had the stock booming, not bombing, right up until the Deepseek announcement.

Reply to  Sparta Nova 4
January 29, 2025 4:47 pm
  • January 14, 2025: Pelosi disclosed purchasing 50 call options for Nvidia at a strike price of $80, with an expiration date of January 16, 2026, valued between $250,001 and $500,000.
  • December 31, 2024: Pelosi reported selling 10,000 Nvidia shares valued between $1 million and $5 million.
  • December 20, 2024: Pelosi disclosed exercising 500 call options purchased on November 22, 2023, acquiring 50,000 shares at a strike price of $12 with an expiration date of December 20, 2024, valued between $500,001 and $1 million.
  • July 26, 2024: Pelosi disclosed a purchase of 10,000 shares of Nvidia worth between $1 million and $5 million.
  • June 26, 2024: Pelosi disclosed a purchase of 10,000 shares of Nvidia worth between $1 million and $5 million

Its seems this was the reason
Pelosi disclosed exercising 500 call options purchased on November 22, 2023, acquiring 50,000 shares at a strike price of $12 with an expiration date of December 20, 2024

The shares were sold- at a much higher price just before the call option expired

New call options 2 weeks latter

dh-mtl
January 28, 2025 2:35 pm

Eric,

I understood that ‘DeepSeek’ was much less resource intense, and thus cheaper, than the American AI models.

If this is the case, then it completely upsets the American strategy for AI dominance, which appears to me to be to ‘out-resource’ the competition (i.e. the $500B Stargate project).

Reply to  Eric Worrall
January 28, 2025 4:35 pm

Nothing that Chyna says can be taken at face value. Theres always bubbles inside bubbles and AI is a classic bubble effect. Not useless but overhyped

Sparta Nova 4
Reply to  Duker
January 29, 2025 7:27 am

While true, the converse cannot likewise be proven.

China is either being open and transparent or devious and lying or something in the middle.

Given the past history, I would lean towards not accepting at face value any China announcements.

Reply to  Sparta Nova 4
January 29, 2025 4:49 pm

yes. Its like much of the climate research , just agit prop unless validated

Denis
January 28, 2025 2:44 pm

Did China develop their AI for much less or not? Chinese lie about just about everything just about all the time. Have their numbers been vetted?

The Dark Lord
January 28, 2025 2:44 pm

so China will have the fastest search engine … terrifying …

Scissor
Reply to  The Dark Lord
January 28, 2025 3:00 pm

Even more censored at that.

Rud Istvan
January 28, 2025 2:46 pm

I did a fair bit of research into the Deepseek kerfuffle last two days. Let’s just say things are confusing (at least to me). Various claims made in the 53 page V3 model paper do not add up (for example same IQ from a smaller training dataset—hence less GPU time needed so less energy); others appear to be clever tweaks to both architectures and algorithms enabling a smaller server farm with fewer less powerful GPUs (hence energy and cost savings). The 10:1 claims seem a stretch no matter what, with no way to verify.

One big issue is open source, which means DeepSeek can derive no competitive advantage unless some of the architectural and algorithm tweaks are patentable. I looked at USPTO, nothing yet—altho DeepSeek says their new V3 and R1 AI stuff derives from their proprietary trading models that go back years. If patentable, some of those should have at least published by now in the US. The normal time frame is 6 months from application submission. Nothing that I could find from China assigned to DeepSeek, these specific folks by name, or the their parent trading company.

Since the software is open source, I suspect that in a few weeks the dust will settle and we will know more about the truth. For example, META just set up 4 separate engineering ‘war rooms’ to figure DeepSeek out: Training; architecture and algorithms; hardware; and application of all three to their own AI.

Tom Halla
Reply to  Rud Istvan
January 28, 2025 3:10 pm

Or the PRC is announcing vaporware? It is the PRC, after all.

Nick Stokes
Reply to  Tom Halla
January 28, 2025 3:17 pm

Open source software

Rud Istvan
Reply to  Nick Stokes
January 28, 2025 3:55 pm

From what I read past two days, open source in the peculiar sense of giving all the AI probability weightings in the V3 LLM and R1 reasoning models, but NOT all the algorithms that processed the training to produce the weightings. Dunno if the sources on that are reliable.

Sparta Nova 4
Reply to  Nick Stokes
January 29, 2025 7:30 am

“I fear the Greeks even when bearing gifts”
— Trojan priest Laocoön referring to the Trojan Horse

Reply to  Rud Istvan
January 28, 2025 6:10 pm

One big issue is open source, which means DeepSeek can derive no competitive advantage

Ding, ding, ding! We have a winner!

BeAChooser
Reply to  Rud Istvan
January 29, 2025 12:19 pm

I think the real story here, whether the hype about the Chinese AI is true or not, is explained in this Axios article (https://www.axios.com/2025/01/29/deepseek-ai-china-chatgpt). It points out that cheaper AI will lead to more energy use, noting that “Morgan Stanley energy analysts concurred in a note Tuesday, saying ‘there are multiple indications’ the Jevons Paradox applies to the cost of compute in AI applications.” The Jevons Paradox states that “greater efficiency in the use of any given resource can result in increased demand for that resource.” 

The article observes that “The cost of compute has been one of the primary constraints in the development of AI. If it comes down, then it is possible that many more companies and research departments will get into the space” and that will lead to “AI getting embedded much more rapidly in things like consumer products”, which will lead to greater energy use.  What are the implications of significantly increasing the already high energy demands of AI that are forcing the elites to start building fossil fueled power plants and urgently waste billions of dollars pursuing harebrained and risky fusion projects? 

And the article notes the announcement by Chevron and GE Vernon that they too are buiiding new fossil fueled power plants co-located with data centers that will only supply power to them. These are two companies that were on-board with AGWalarmism. Chevron announced a goal to be Net Zero by 2050 and said it’s “imperative” to meet “the world’s goals to reduce greenhouse gas emissions”.  Obviously they thought CO2 is a problem. GE claimed it would be carbon neutral in 2030 and said it “supports the science and goals expressed in the Paris Agreement and the United Nations Framework Convention on Climate Change.” What could be so important to them about AI to risk the end of humanity, if they believed in AGWalarmism? Think about that.

KevinM
January 28, 2025 2:58 pm

Why Big Tech Ditched Big Green
Always confused me that analytical thinkers were willing to go along with the wishy-washiness.

Tom Halla
January 28, 2025 3:05 pm

The National Review made the suggestion that the Chinese were lying about the announcement, that the cost was much higher than stated, and/or the chips involved were the full power Nvidia chips that had been smuggled despite the embargo.
Given the record of the Peoples Republic, that sort of head fake would be quite plausible.

Sparta Nova 4
Reply to  Tom Halla
January 29, 2025 7:32 am

“I fear the Greeks even when bearing gifts”
— Trojan priest Laocoön referring to the Trojan Horse

One has to wonder if the opensource software carries a payload.

January 28, 2025 3:06 pm

I see all of the articles and discussion about AI. All of the cost and energy calculations, all of the prognostications about hardware and language models, but the one question I cannot get a decent answer on is, “WHY?” Why is there such a race to create AI? All I’ve heard about is the things AI *might* be able to do, but will it really deliver? Has it even delivered anything significant to date other than some doctored photos and helping people write better?

It seems to me that this is an incredibly energy and resource-intensive chase to achieve something with speculated value? Kind of reminds me of the race to the moon or the arms race….but it feels like the tables might be turned? Could this be some sort of concerted effort to finally bankrupt and break the USA?

What happens if we achieve AI but it’s doesn’t deliver much more than what we have today? Maybe I’m missing something….

johnn635
Reply to  NavarreAggie
January 28, 2025 3:30 pm

When I heard that millions of people were downloading a Chinese AI App I fell about laughing. Not satisfied with seriously impacting Western economies by panic reactions to Covid, they now provide software that does who knows what! Unbelievable?

Reply to  johnn635
January 28, 2025 10:38 pm

And what other apps are in those downloads?

Sparta Nova 4
Reply to  Retired_Engineer_Jim
January 29, 2025 7:33 am

Apology, I am repeating.

“I fear the Greeks even when bearing gifts”
— Trojan priest Laocoön referring to the Trojan Horse

One has to wonder if the opensource software carries a payload.

Reply to  NavarreAggie
January 28, 2025 6:17 pm

The ‘idea’ of Artificial Intelligence is fine. I predicted many decades ago that our true goal will be the computer in Star Trek. Science Fiction has often described this kind of AI.

What we have now is laughable. It’s nothing useful except to pretend to be slightly intelligent. It cannot ‘reason’ or make decisions, it merely gobbles up whatever text it is offered, and regurgitates it in a supposedly pleasing format.

Only a Natural Idiot would believe that this is anything but an Artificial Idiot. Unfortunately, Natural Idiots abound, especially in media, and politics driven by media.

Reply to  Zig Zag Wanderer
January 29, 2025 1:32 am

neural net systems learn to do new things. This is not new, it is well established.

Reply to  AndyHce
January 29, 2025 4:09 am

Neural networks can indeed be trained, given enough data. They cannot actually learn to do anything new.

These Artificial Idiots don’t even do that any better than decades ago, merely faster. The concentration has been on trying to understand and create language. That part is still in the fledgling stage.

This is well established. It has been for decades. I was a software developer, and have written neural networks, so I do know something about them.

Sparta Nova 4
Reply to  Zig Zag Wanderer
January 29, 2025 7:39 am

I had a “conversation” with ChatGPT a while back.

Per ChatGPT, AI is not intelligent as it lacks consciousness.

I posited that AI is advanced software weighted decision trees, with some adaptability on weights hidden behind a very good language model.

ChatGPT response: “Spot on.”

Jeff Alberts
Reply to  AndyHce
January 29, 2025 7:20 am

But will they learn them on their own, or only when “trained”? The former is intelligence, the latter, not.

jvcstone
January 28, 2025 3:39 pm

Seems to me that trump can save the country a half trillion and just download the free Deepseek apt from the apple apt store.

January 28, 2025 3:49 pm

So what’s the big deal about this Chinese super AI?

“built at a lower cost and with less computing resources than what larger tech firms typically need for LLM development.”

OK, so it’s cheaper and uses less computer resources- why does that make it a superior AI?

Rud Istvan
Reply to  Joseph Zorzin
January 28, 2025 3:58 pm

If the hype is true, an AI of equal capability (not superior) but at significantly lower cost is economically superior.

David Wojick
January 28, 2025 3:55 pm

Keep in mind that if AI needs a lot more juice then we are talking about 2040 at the earliest just to get started. Nobody’s share price today depends on that.

January 28, 2025 4:15 pm

In their last episode, the two non-Chinese dudes who put out “The China Show” reported on the results of two questions put to one of the Chinese AI’s (Deepseek) … one was ‘Does Donald Trump exist?‘ and the other was ‘Does Xi Jinping exist?

The answer to the DJT question was a simple “yes”, however the answer for the Xi Jinping question was “Sorry, that’s beyond my current scope. Let’s talk about something else.

OH YEAH – China’s AI holds great promise . /SARC (if you didn’t catch that)

Here’s the segment, cued up: https://youtu.be/bZciXwVsYQk?t=2563

Sparta Nova 4
Reply to  _Jim
January 29, 2025 7:41 am

In otherwords,

“I am not programmed to respond in that area.”

— Star Trek, I Mudd

January 28, 2025 4:34 pm

Just in, over the transom: U.S. Navy Bans DeepSeek Over ‘Security Concerns’ As ‘Substantial’ Evidence Emerges Chinese AI Ripped Off ChatGPT
https://www.zerohedge.com/ai/us-navy-bans-deepseek-over-security-concerns-substantial-evidence-emerges-chinese-ai-ripped

A couple of excerpts:

The U.S. Navy has instructed service members to avoid using the Chinese AI platform DeepSeek, citing “potential security and ethical concerns,”

and

The warning follows the recent rise of DeepSeek’s R1 model, which has garnered significant attention worldwide, particularly within the U.S. business and technology sectors. The R1 model has demonstrated capabilities comparable to OpenAI’s models. In December, DeepSeek claimed it had successfully trained a large language model in just two months at a cost of $6 million—a figure disputed by technologists—despite U.S. restrictions on semiconductor chip exports to China.

Sparta Nova 4
Reply to  _Jim
January 29, 2025 7:43 am

The Navy learned the lessons of the Trojan War.

January 28, 2025 4:34 pm

I wonder what Deepseek thinks about a climate crisis caused by CO2?

What do Deepseek’s programmers want it to say about CO2? What’s the Chicom line?

Reply to  Tom Abbott
January 28, 2025 8:57 pm

Ask and let us know.

Reply to  whsmith@wustl.edu
January 29, 2025 4:47 am

I don’t think I’m going to download the app. Being from the Chicoms, it might compromise my phone.

Keitho
Editor
Reply to  Tom Abbott
January 28, 2025 10:08 pm

I asked it and it simply regurgitated the standard line. Nothing new and the old BS about ocean acidification was there unchanged. I will stick to Grok for now.

mleskovarsocalrrcom
January 28, 2025 5:18 pm

I’m waiting on some hands on testing to see what this package really does. Is it an entry level AI platform or a robust system ready to solve world problems.

January 28, 2025 5:39 pm

I did a test on DEEPSEEK to determine the LCOE of battery firmed wind in Australia. I asked for average cost for installation. This was the answer after detailed calculations:

The levelized cost of energy (LCOE) for the lithium battery-firmed wind farm is approximately AUD 1,356/MWh. This high cost is primarily driven by the large battery storage requirement for 2-week wind droughts and the low wind capacity factor.

The full answer is more comprehensive than Perplexity and includes O&M costs without specifically requesting them. Perplexity does not include O&M unless you ask for them.

Reply to  RickWill
January 28, 2025 8:59 pm

Not a bad estimate either, using Li-ion battery backup. Backup, no matter how you do it, triples the cost of wind-PV.

January 28, 2025 5:54 pm

Just heard this morning that the CHINESE DeepSeek records your your key strokes. It’s in the terms and conditions. Don’t use it.

Sparta Nova 4
Reply to  Mike
January 29, 2025 7:44 am

No surprise.

January 28, 2025 6:16 pm

Plot twist: DeepSeek is a DeepFake along the lines of Theranos.

Theranos actually used industry standard blood work analysis equipment behind the scenes to analyze blood samples taken from patients, unknown to the patients or those being shown Theranos-built equipment that was SUPPOSED TO BE being used to make the blood analysis runs.

To wit, with DeepSeek there is a back-end to DeepSeek that uses a panel, nay, a room full of generalists in a say a half dozen technical areas and ALL DeepSeek is, is really a per-processor for A) a more general search (Google?) engine AND B) kinda in parallel the questions, queries routed to the ‘room’ of subject matter experts/technical generalists … maybe the DeepSeek answers/searches get reviewed, or modified by a panel on a go/no go way … point being, this is one big show and fake as all get out, ala, Theranos (IF you know the full story there.)

Rud Istvan
Reply to  _Jim
January 28, 2025 8:07 pm

The Theranos story is apt. Claim was could do same blood chemistry analysis with a finger prick drop that otherwise took a full tube blood draw. The problem always was that never made any known analytic chemistry sense—you need a certain volume to get added reagents to produce a definitive result. Theranos was finally proven a giant scam.

Could be the same here. Same ‘IQ’ from much more limited data training set simply does not logically compute, exactly analogous to Theranos.

January 28, 2025 8:42 pm

Do not accept the hype. Try the various AIs for yourself. Ask hard questions to which you know the answer, and you will learn a lot about how much the programmers know and what is IN the data base and what is not. It is not exciting, at least not yet, no matter the huge investments our not too bright oligarchs are making. My students seem to like AI’s editing capability most. Non- native English speakers benefit from AI, BUT AI is very wordy. It does not know about Samuel Clemens. If students use it without editing, it is easy to recognize the source.

Rud Istvan
January 28, 2025 9:22 pm

Fun new technical possibility just showed up this evening.
It is possible that DeepSeek R1 was reverse engineered by ‘training’ on equivalent ChatGPT o1. The newly coined word for this possibility is ‘reverse engineering’—of course also illegal intellectual theft for which China is famous.

We have a way to go for clarity.

January 29, 2025 1:18 am

But China claims to have used markedly fewer resources to achieve what they announced. On that basis, it is not a lace of resources in the US, it is lack of a different approach.

Reply to  AndyHce
January 29, 2025 4:51 am

China makes a lot of claims.

Verified by MonsterInsights