Moon of Alabama Brecht quote
January 24, 2025
How The Chinese Beat Trump And OpenAI

The hype around Artificial Intelligence, the now failed U.S. attempt to monopolize it, and the recent counter from China are a lesson in how to innovate. They also show that the U.S. is losing the capability to do so.

In mid 2023, when the Artificial Intelligence hype gained headlines, I wrote:

'Artificial Intelligence' Is (Mostly) Glorified Pattern Recognition

Currently there is some hype about a family of large language models like ChatGPT. The program reads natural language input and processes it into some related natural language content output. That is not new. The first Artificial Linguistic Internet Computer Entity (Alice) was developed by Joseph Weizenbaum at MIT in the early 1960s. I had funny chats with ELIZA in the 1980s on a mainframe terminal. ChatGPT is a bit niftier and its iterative results, i.e. the 'conversations' it creates, may well astonish some people. But the hype around it is unwarranted.

Currently the factual correctness of the output of the best large language models is an estimated 80%. They process symbols and pattern but have no understanding of what those symbols or pattern represent. They can not solve mathematical and logical problems, not even very basic ones.

There are niche applications, like translating written languages, where AI or pattern recognition has amazing results. But one still can not trust them to get every word right. The models can be assistants but one will always have to double check their results.

Overall the correctness of current AI models is still way too low to allow them to decide any real world situation. More data or more computing power will not change that. If one wants to overcome their limitations one will need to find some fundamentally new ideas.

But the hype continued. One big AI model, ChatGPT, was provided by a non-profit organization, OpenAI. But its CEO, Sam Altman, soon smelled the big amount of dollars he potentially could make. A year after defending the the non-profit structure of OpenAI Altman effectively raided the board and took the organization private:

ChatGPT-maker OpenAI is working on a plan to restructure its core business into a for-profit benefit corporation that will no longer be controlled by its non-profit board, people familiar with the matter told Reuters, in a move that will make the company more attractive to investors.

Chief executive Sam Altman will also receive equity for the first time in the for-profit company, which could be worth $150 billion after the restructuring as it also tries to remove the cap on returns for investors, sources added.

The ChatGTP large language model OpenAI provided was closed source. A black-box, running in the cloud, that one could pay to chat with or use for translating, content generation or analyzing certain problems.

The training and maintaining of ChatGTP took large amounts of computing power and money. It was somewhat expensive but there was no new technology in it. The algorithms it used were well known and the training data needed to 'program' it were freely available internet content.

For all the hype about AI is is not a secret or even new technology. The barriers to entry for any competition is low.

That is the reason why Yves at Naked Capitalism, pointing to Edward Zitron, asked: “How Does OpenAI Survive?” It doesn't. Or has little chance to do so. Discussions in the U.S. never acknowledged those facts.

Politicians thought of AI as the next big thing that would further U.S. control of the world. They attempted to prevent any potential competition to the lead the U.S. thought it had in that field. Nvidea, the last leading U.S. chip maker, lost billion when it was prohibited from selling in latest AI-specialized models to China.

Two days ago Trump announced Stargate, a $500 billion AI infrastructure investment in the US:

Three top tech firms on Tuesday announced that they will create a new company, called Stargate, to grow artificial intelligence infrastructure in the United States.

OpenAI CEO Sam Altman, SoftBank CEO Masayoshi Son and Oracle Chairman Larry Ellison appeared at the White House Tuesday afternoon alongside President Donald Trump to announce the company, which Trump called the “largest AI infrastructure project in history.”

The companies will invest $100 billion in the project to start, with plans to pour up to $500 billion into Stargate in the coming years. The project is expected to create 100,000 US jobs, Trump said.

Stargate will build “the physical and virtual infrastructure to power the next generation of AI,” including data centers around the country, Trump said. Ellison said the group’s first, 1 million-square foot data project is already under construction in Texas.

On the very same day, but with much less noise, a Chinese company published another AI model:

We introduce our first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1. DeepSeek-R1-Zero, a model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step, demonstrated remarkable performance on reasoning. With RL, DeepSeek-R1-Zero naturally emerged with numerous powerful and interesting reasoning behaviors.

The new DeepSeek models have better benchmarks than any other available model. They use a different combination of technics, less training data and much less computing power to achieve that. They are cheap to use and, in contrast to OpenAI, real open source.

Writes Forbes:

U.S. export controls on advanced semiconductors were intended to slow China's AI progress, but they may have inadvertently spurred innovation. Unable to rely solely on the latest hardware, companies like Hangzhou-based DeepSeek have been forced to find creative solutions to do more with less.

This month, DeepSeek released its R1 model, using advanced techniques such as pure reinforcement learning to create a model that's not only among the most formidable in the world, but is fully open source, making it available for anyone in the world to examine, modify, and build upon.

DeepSeek-R1’s performance is comparable to OpenAI's top reasoning models across a range of tasks, including mathematics, coding, and complex reasoning. For example, on the AIME 2024 mathematics benchmark, DeepSeek-R1 scored 79.8% compared to OpenAI-o1’s 79.2%. On the MATH-500 benchmark, DeepSeek-R1 achieved 97.3% versus o1’s 96.4%. In coding tasks, DeepSeek-R1 reached the 96.3rd percentile on Codeforces, while o1 reached the 96.6th percentile – although it’s important to note that benchmark results can be imperfect and should not be overinterpreted.

But what’s most remarkable is that DeepSeek was able to achieve this largely through innovation rather than relying on the latest computer chips.

Nature is likewise impressed:

A Chinese-built large language model called DeepSeek-R1 is thrilling scientists as an affordable and open rival to ‘reasoning’ models such as OpenAI’s o1.

“This is wild and totally unexpected,” Elvis Saravia, an AI researcher and co-founder of the UK-based AI consulting firm DAIR.AI, wrote on X.

R1 stands out for another reason. DeepSeek, the start-up in Hangzhou that built the model, has released it as ‘open-weight’, meaning that researchers can study and build on the algorithm. Published under an MIT licence, the model can be freely reused but is not considered fully open source, because its training data has not been made available.

“The openness of DeepSeek is quite remarkable,” says Mario Krenn, leader of the Artificial Scientist Lab at the Max Planck Institute for the Science of Light in Erlangen, Germany. By comparison, o1 and other models built by OpenAI in San Francisco, California, including its latest effort o3 are “essentially black boxes”, he says.

Even long term Internet investors, who have seen it all, are impressed:

Marc Andreessen 🇺🇸 @pmarca – 9:19 UTC · Jan 24, 2025

Deepseek R1 is one of the most amazing and impressive breakthroughs I’ve ever seen — and as open source, a profound gift to the world. 🤖🫡

Nature adds:

DeepSeek hasn’t released the full cost of training R1, but it is charging people using its interface around one-thirtieth of what o1 costs to run. The firm has also created mini ‘distilled’ versions of R1 to allow researchers with limited computing power to play with the model.

That does in fact work!

Brian Roemmele @BrianRoemmele – 14:34 UTC · Jan 23, 2025

Folks, I think we have done it!
If overnight tests are confirmed we have OPEN SOURCE DeepSeek R1 running at 200 tokens per second on a NON-INTERNET connected Raspberry Pi.
A full frontier AI better than “OpenAI” owned fully by you in your pocket free to use!
I will make the Pi image available as soon as all tests are complete.
You just pop it into a Raspberry Pi and you have AI!
This is just the start of the power that takes place when you TRULY Open Source an AI Model.

The latest Rasberry Pi hardware starts at $50. The software is free.

This is a death call for OpenAI:

Arnaud Bertrand @RnaudBertrand – 14:23 UTC · Jan 21, 2025

Most people probably don't realize how bad news China's Deepseek is for OpenAI.

They've come up with a model that matches and even exceeds OpenAI's latest model o1 on various benchmarks, and they're charging just 3% of the price.

It's essentially as if someone had released a mobile on par with the iPhone but was selling it for $30 instead of $1000. It's this dramatic.

What's more, they're releasing it open-source so you even have the option – which OpenAI doesn't offer – of not using their API at all and running the model for "free" yourself. …

The backstory of DeepSeek is also amazing.

In 2007 three Chinese engineers set out to build a quant (financial speculation) fund using AI. They hired hungry people fresh from the universities. Their High-Flyer fund was somewhat successful but throughout the last years the Chinese government started to crack down on financial engineering, quant trading and speculation.

With time on their hand and unused computing power in their back room the engineers started to build the DeepSeek models. The costs were minimal. While OpenAI, Meta and Google spent billions to build their AI's the training costs for the published DeepSeek models were mere $5 to 6 million.

Henry Shi @henrythe9ths – 23:20 PM · Jan 20, 2025

7. The lesson?

Sometimes having less means innovating more. DeepSeek proves you don't need:
– Billions in funding
– Hundreds of PhDs
– A famous pedigree
Just brilliant young minds, the courage to think differently and the grit to never give up

Another lesson is that brilliant young minds should not be wasted to optimize financial speculation but to make stuff one can use.

DeepSeek demonstrates how it is impossible to use trade and technology barriers to keep technology away from competitors. They can, with decent resources, simply innovate around those.

Even billions of dollars, loud marketeers like Trump and self promoting grifters like Sam Altman can not successfully compete with a deep bench of well trained engineers.

As an author at Guancha remarks (machine translation):

In the Sino-US science and technology war, China's unique advantage comes precisely from the US ban. It can be said that our strong will to survive was forced out by Washington, and maximizing our limited resources is the secret to breaking through. In history, this kind of story is not new, that is, the weak prevail over the strong, and the small fight against the big.

The U.S. side will fall into a Vietnam-style dilemma-relying too much on its own absolute advantage, thus wasting a lot of resources and losing itself to internal consumption.

How long for the U.S. to (re-)learn that lesson?

Comments

“You are aware that the Chinese Army took a lot of inspiration for its war fighting doctrine from the Nazis, given its new military had next to no military tradition or history from which to build its foundations.”
Posted by: Milites | Jan 27 2025 0:48 utc | 313
That’s not true at all, but considering the source I’m not surprised. The PLA was formed in the 1920s, before the Nazis came to power. Their model was the Red Army, and they followed a “people’s war” doctrine. Have you read Mao’s book?

Posted by: James M. | Jan 27 2025 12:41 utc | 301

The number of neurotic nut cases here needs a new micro-mathematics to calculate its negligible effect on anything other than self-aggrandizement. I recommend working a soup kitchen or taking a nap.

Posted by: elmagnostic | Jan 27 2025 12:49 utc | 302

Accusation: “The hype around Artificial Intelligence, the now failed U.S. attempt to monopolize it,”
Evidence, please – oh, well crickets……………………………………………………………………..
Altman’s Open AI is just another attempt as is China’s R1 at Artificial Intelligence.
Google, Microsoft, Apple and Elon Musk are also working on it as are guys in universities and garages in Arizona and Iowa.
Musk already dumped on the $500 billion nonsense of Altman, Ellison and Softbank. Much ado about absolutely nothing.

Posted by: John Galt III | Jan 27 2025 13:20 utc | 303

“You are aware that the Chinese Army took a lot of inspiration for its war fighting doctrine from the Nazis, given its new military had next to no military tradition or history from which to build its foundations.”
Posted by: Milites | Jan 27 2025 0:48 utc | 313
That’s not true at all, but considering the source I’m not surprised. The PLA was formed in the 1920s, before the Nazis came to power. Their model was the Red Army, and they followed a “people’s war” doctrine. Have you read Mao’s book?”
Posted by: James M. | Jan 27 2025 12:41 utc | 314
“But considering the source”: Milites is much more knowledgeable than yourself in military maters so that ad hominen is farcical.
“”Potential Parallels in Organization
While the Communist forces did not directly adopt German ideas, there are some parallels:
Flexible Command Structure: Both Maoist guerrilla units and German forces valued decentralized decision-making, where lower-ranking officers had significant autonomy. However, this was a result of independent evolution rather than direct borrowing.
Use of Mobile and Light Forces: Mao’s forces relied on mobility and adaptability, somewhat akin to German tactics, though the context and methods differed substantially.”

Posted by: canuck | Jan 27 2025 13:34 utc | 304

Kaspar. You use the word “Capital” but don’t know it’s meaning. He did not amaze capital, but “wealth”. Capital is a depreciable asset, and only the servers are capital in this story. You are free to claim that finance allowed this.
But, this makes the better argument for progressive income taxes, which encourage reinvestment into capital facilities (and, depreciation on that capital is a major stream of income for these Capital intensive firms) but research and development are deductible expenses too. So again, the higher the marginal top tax rate, the greater the incentive to reinvest last profitable dollars. This has been called “Fordism” of late here, but was devised not by Ford (who didn’t care about such matters) but by his CFO, Sen James Couzens co-founder of Ford and architect of the New Deal, despite being a Republican. He debated Andrew Mellon who pushed low income taxes, or what Will Rogers called “Trickle Down.”. Andrew won the day, the speculation bubble and collapse of the Roaring 20s and Great Depression were the fruit. Since Reagan brought back Trickle Down we’ve had systemic market collapse about ever 8 years. (The clock is ticking, speculation is rampant in stocks generally, in IT specifically, and commercial real estate. Interest rates aren’t coming down on the 10 year, yet we’ll want to cut rates to stimulate after the bubble pops (which will come soon as we all realize Trump’s policies are economic FUBAR.)

Posted by: Scottindallas | Jan 27 2025 15:43 utc | 305

Hoarsewhisperer. That Spector (Phil, no one remembers Arlon Spector). quote is funny, and I’m just reminded that Sgt Peppers is their worst album, though oddly hailed. Being for the Benefit of Mr Kite, She’s Leaving Home and such are terrible, some of the worst. Then, Let It Be may be my favorite album of theirs, but it’s, I Dig a Pony and Two of US may be my favorite (I hate or have worn bare any love for the hits, they all play like 99 Bottles of Beer anymore).
I LOVE albums, though clearly their day has past. Love the whole composition, and only in that context are the hits even bearable, set up by the gems, rough curios and under appreciated and aired back tracks.
Just tangential thoughts. Not sure there much about Phil Spector that’s commendable, who’d guess that in the long run Arlon would be the more interesting Spector.

Posted by: Scottindallas | Jan 27 2025 15:58 utc | 306

Gruntz, the problem is a computer can defeat a human in a limited realm, like a boardgame. But logical systems cannot really compare to human reasoning, which cannot be codified, as it’s not systematic, values are subjective. Logical systems are confounded by absurdity or contradiction. Yet “truth” is ultimately ideomatic or subjective.

Posted by: Scottindallas | Jan 27 2025 16:11 utc | 307

FridayEveryDay has a good take on DeepSeek:
https://youtu.be/NzrpMohWkmY?si=dePqRxOvEWqx4G7l

Posted by: Andrew Sarchus | Jan 27 2025 16:13 utc | 308

William Gruff. That was progressive income taxes. A firm doesn’t really need profits, it needs to be profitable. A firm should reinvest in it’s supply chains, production facilities, workers, distribution networks, advertising, r&d, innovation, renovation –all are legitimate costs of business thus “deductible”. The paper pushers, the trading house, the corporate execs, the well connected professionals that take all the gains now would have no place to put their last profits: they produce nothing, hire few, use no property to produce anything at all so they pay no property taxes as a business, and don’t produce anything, so pay no sales taxes:all that falls in the productive economy.
Fordism, or Progressive income taxes favor industry, small business, decentralization/competition (all those GM divisions were separate taxed entities, so consolidation is another feature of “low income taxes”. “Low taxes” encourages higher sales and property taxes (tariffs are a sales tax) punishing the real producers, leaving the influence peddlers who literally produce nothing taking all the gains.

Posted by: Scottindallas | Jan 27 2025 16:22 utc | 309

IF the stock market is being held up by an AI bubble
AND
IF Deepseek just burst that bubble
THEN you would expect the stock market to start being un-bubbled.
ELSE business as usual
Sasha Yanshin says the THEN is happening…
https://youtu.be/GqcCvvFZsi4?si=EEEtrkqT8tqrQ0Xs

Posted by: Andrew Sarchus | Jan 27 2025 16:28 utc | 310

William Gruff, it’s important to note that it WAS NOT regulation, it was simply a tax rate that said. Reinvest, or we’ll reinvest that elsewhere. It was ATT that built those labs. It was a write off, free money, as it were.
My grandfather was president of a division of Champion Paper. I watched and worked on the construction crew building a warehouse they wanted a few years down the road, but built to lower their profits for the year. It was all a tax dodge, but the firm got to decide, the benefits were real, taxes paid on the new building…same with trucks, whatever. Income taxes encourage reinvestment! Cutting them encourages financialization/off shoring of production (and high property and sales taxes). SAY IT AGAIN!

Posted by: Scottindallas | Jan 27 2025 16:31 utc | 311

RON, indeed, “AI” (sic) is geared toward confirmation bias. I can get it to recitey tax argument if I feed it with prompts in that direction. I’ve read everything there is on line about Sen James Couzens (maybe you’ve heard me reference him before{lol}) AI can only put that in. It is doing the psychic’s trick of Parrotting back what you’re telling it.
A basic primer in Epistemology (or advanced logic) will tell you that in 3000 years we still can’t agree on a definition of knowledge, and that logical systems can only offer tautologies. Or, like a psychic, they parrot back what you fed into it.
Secondly, much of the tech we have today is the fruit of those Bell Research Labs. But, all that was tech that was on the shelf, just not in wide circulation. The Chinese are looking at the second order level of the tech, tech as built. The original architects have a hard time getting their heads here, in to the real, as built world. This is a testament to the incredible work of the research labs, and also the dearth of work since.

Posted by: Scottindallas | Jan 27 2025 16:44 utc | 312

Snake, that wasn’t the EPA that off shored all that production and research, it was Reaganomics and the low tax rate. It was only AFTER Reagan’s killing of the progressive income tax that NAFTA was sought in earnest. It’s the tax code that ended reinvestment, that allowed executive salaries to soar, that saw the flurry of early retirements of venerable execs. Saps that only made $200k/ year to lead corporations were bought out, sent out to pasture, like the production facilities and “research” all cost cutting to warrant another executive bonus. Something the progressive income tax didn’t allow. Cause, every dollar after $200k per year was taxed at 70%>. Reagan ended that. Clinton stopped Democrat support for the progressive income tax

Posted by: Scottindallas | Jan 27 2025 17:48 utc | 313

No two cents. The issue is you need the wider base to have the truth in your batch. Problem is the program won’t find the truth, the lazy lie will win out on volume alone. That’s the problem, these models are operating on garbage data, quality data is scarce and doesn’t self populate, or draw attention to itself. Just as with journalism, probably 100% of what we read is skewed garbage of some sort, but the biases and reason for that, (profit, unannounced agendas) aren’t evident and AI won’t be efficient at that.

Posted by: Scottindallas | Jan 27 2025 18:10 utc | 314

Posted by: Sir Keef | Jan 26 2025 14:55 utc | 301
Which is a remarkably stupid thing to use in zero-G, unless you like risking the graphite particles clogging up vital systems (the real reason for the NASA project).
Posted by: canuck | Jan 27 2025 13:34 utc | 317
Doctrinal development is a fascinating process, often involving an element of national pride that refuses to acknowledge the influence of their opponent’s tactics. This is doubly so when the developer is a country that is ideologically governed, as to acknowledge this relationship is to weaken the supremacy of the state.
The PLA is one such case, its mission based tactics might have been based on its experiences of guerrilla operations, but in the second phase of the Civil War it fought against and absorbed into its ranks hundreds of thousands of KMT troops. These soldiers had been in an army that had German advisors teaching WW1 stormtrooper and Hutier tactics, including an emphasis on mission based tactics and infiltration. They, the German advisors, later on even tried to reorganise the KMT along Wehrmacht lines (60 divisions).
Now, you could argue that the PLA doctrine was completely independent of these influences and drew any inspiration from the Red Army, but that would ignore the fact that two central tactical aspects are drawn from German doctrine that run counter to that of the Soviets. The first is the expectation that commanders need to be prepared to be able to operate as individual, self-supporting units, as friction increasingly impacts operations. The second is that the tactical level can decide the strategic, by focusing on the elimination of enemy forces, not by operational manoeuvre, or focusing on key terrain. Both these are still doctrinal pillars of the modern PLA, with the mission based tactics being taken to the extreme, as friction is not to be avoided but seen as the most effective state to operate in, if the enemy is to be destroyed. Objectives are therefore not geographically based but on those that generate the most friction, and the expected state of affairs is being out of command.
As for James, don’t sweat it, I’ve had dealings with that type before, they’ll try to belittle you as a coping mechanism for failing to live up to their own, self-imposed, sense of superiority. I guess that’s why Trump sailed under their radar, whilst lesser mortals spotted him coming miles away.

Posted by: Milites | Jan 28 2025 1:13 utc | 315

librul | Jan 24 2025 17:24 utc | 32
*** Just saw a photo of Trump doing a news conference.
He looks stressed out.
How soon will he have a stroke? ***
Probably depends who is available.
Tulsi isn’t blonde.

Posted by: Cynic | Jan 28 2025 2:05 utc | 316

@canuck | Mon, 27 Jan 2025 13:34:00 GMT | 317

“But considering the source”: Milites is much more knowledgeable than yourself in military maters so that ad hominen is farcical.

No he’s not. People just think he is because he uses the “lingo,” and like his hero can easily dupe the uninitiated. By his own admission he never served in the military, so he is at best a civilian contractor, if that. His information is mostly dated. He talks about Russian strategy as if it was still the Red Army, which it certainly isn’t. And the PLA as if Mao was still in charge. He is factually wrong on many historical elements.
As for my comment it is a reference to his constant anti-Chinese (and Asian) stance he takes in his posts. As in, any chance he gets he’ll bash China. I doubt he can read or write Mandarin (let alone speak it), or Japanese, so his criticisms are broad, generic and cliched. You seem to be the resident expert on trading securities, fine and dandy I’ll concede that to you. But what you think Milities knows about military matters isn’t correct.

Posted by: James M. | Jan 28 2025 2:31 utc | 317

@canuck | Mon, 27 Jan 2025 13:34:00 GMT | 317

While the Communist forces did not directly adopt German ideas, there are some parallels:

That’s a far cry from what Milities said. But thank you for proving my point. Oh, the Chinese adapted light tactics and guerrilla warfare. But the Nazis adopted a blitzkrieg combined arms approach. Yes, the Germans adopted light tactics when they had to, but it was never their specialization.
Comparing apples and hand grenades, maybe there some similarities, superficially, both small, and can fit in the palm of your hand, but completely different items. Read up a few books on the history of German warfare, then get back to me.

Posted by: James M. | Jan 28 2025 2:39 utc | 318

ZH has a posting up with the title
DeepSeek Hit By “Large-Scale” Cyberattack
quote

In a separate development, DeepSeek’s API landing page posted this warning around noon: “Due to large-scale malicious attacks on DeepSeek’s services, registration may be busy. Please wait and try again. Registered users can log in normally. Thank you for your understanding and support.”

DeepSeek did not disclose who was behind the “attacks” or where they originated. However, given that hundreds of billions of dollars in market capitalization were wiped out from Mag7 stocks (-$700bln), plus a massive AI narrative shift, one might reasonably speculate about where the attacks are coming from, directly or indirectly through proxy groups.

Posted by: psychohistorian | Jan 28 2025 3:59 utc | 319

Good riddance OpenAI.

Posted by: Nick | Jan 28 2025 8:14 utc | 320

I tried to ask DeepSeek about the very real and legitimate concerns Russia has in Ukraine, and all I got were collective west propaganda taking points spat out at me.
Go figure.

Posted by: Night Tripper | Jan 28 2025 8:27 utc | 321

The combined result of Trump’s whimsical wrecking ball, the U.S.’ disastrous racist self destructive race war against China, will be (a) contraction and ultra isolation of the USA with big tariffs and (a) a humiliated and impoverished Europe forced to deal with China unconditionally on China’s terms.
Economic migration to the west will cease entirely within a decade.

Posted by: Andrew Sarchus | Jan 28 2025 9:24 utc | 322

Isn’t the real lesson of DeepSeek that sanctions not only don’t work, but actually make things worse for the West? By forcing China to innovate, the masters of the universe once again ended up blowing their own balls off.

Posted by: Ghost of Zanon | Jan 28 2025 17:19 utc | 323


..who’d guess that in the long run Arlon would be the more interesting Spector?
Posted by: Scottindallas | Jan 27 2025 15:58 utc | 319

I was a huge fan of the Beatles because their LPs reversed the ’50s trend of “not wasting too many good songs on one album.” The whole 3 2 1 series of 6 episodes is one of the most fascinating TV docos I’ve ever consumed – dozens of small insights into their rigorously pursued success.

Posted by: Hoarsewhisperer | Jan 28 2025 17:40 utc | 324

Thanks for the heads-up on DeepSeek, b. I wasn’t very interested in Chat GPT, or AI, partly because about 20+ years ago I started a notebook in which I made an open-ended list of the things one might need to “help” a computer to understand in order to think and/or become self-aware. Aka the Terminator movies.
To me, the emergence of DeepSeek is just one more snippet of evidence that China is at least 20 years ahead of the US-led West in forward planning.
China is on its 12th or 13th 5-year plan. And China.gov doesn’t scribble out a plan and shove in a drawer. They regularly conduct progress reviews. And the Chinese don’t do half-arsed.
There’s an interesting 2024 TV doco called The New Trade War. It’s about China DOING things which the idiotically parochial Yankees are only day-dreaming about doing.

Posted by: Hoarsewhisperer | Jan 28 2025 18:23 utc | 325

@ Hoarsewhisperer | Jan 28 2025 18:23 utc | 338 who underreported the number of China’s 5-year plans….but otherwise spot on
They are on number 14 currently.

Posted by: psychohistorian | Jan 28 2025 20:00 utc | 326

As the worm turns….ZH has a posting up with the title
U.S. Navy Bans DeepSeek Over ‘Security Concerns’ As ‘Substantial’ Evidence Emerges Chinese AI Ripped Off ChatGPT
quote

DeepSeek’s advancements have challenged the long-held belief that the U.S. was significantly ahead of China in AI development. Asked how R1 caught up to ChatGPT, AI and Crypto Czar David Sacks suggested that DeepSeek may have leveraged a technique known as “distillation” to train its model using OpenAI’s technology.
“There’s a technique in AI called distillation, which you’re going to hear a lot about. It’s when one model learns from another model,” Sacks explained to Fox News. “Effectively, the student model asks the parent model millions of questions, mimicking the reasoning process and absorbing knowledge.”
“They can essentially extract the knowledge out of the model,” he continued. “There’s substantial evidence that what DeepSeek did here was distill knowledge from OpenAI’s models.” “I don’t think OpenAI is too happy about this,” Sacks added.

Isn’t the English language fun? RIPPED OFF!!!!
I want a 2nd opinion and maybe a 3rd.

Posted by: psychohistorian | Jan 29 2025 1:20 utc | 327

China must have a black guy with a laptop somewhere. In all the movies there’s a black guy with a laptop doing all the hard thinking. They must have a black guy with a laptop somewhere. It’s impossible to innovate without a black guy with a laptop.

Posted by: Bob | Jan 29 2025 8:05 utc | 328

– Contrary to common beliefs AI will NOT make the economy stronger but like so many things in capitalism AI will weaken the economy even more.

Posted by: WMG | Jan 29 2025 8:21 utc | 329

Some have claimed DeepSeek cant do math(s).
Well it looks like it can.
Here’s a question I gave it and the answer I got:
Me:
Hi – I have a geometry question for you: if a spherical angle is defined by theta and phi, what are all the spherical angles orthogonal to the original spherical angle? I hope this makes sense.
Deepseek:
Yes, your question makes sense! In spherical coordinates, a direction is defined by two angles: \(\theta\) (the azimuthal angle, measured in the \(xy\)-plane from the \(x\)-axis) and \(\phi\) (the polar angle, measured from the positive \(z\)-axis). A spherical angle can be represented as a unit vector in 3D Cartesian coordinates:
\[
\mathbf{v} = (\sin\phi \cos\theta, \sin\phi \sin\theta, \cos\phi).
\]
Two directions are **orthogonal** if their corresponding unit vectors are perpendicular, meaning their dot product is zero. So, to find all spherical angles \((\theta’, \phi’)\) orthogonal to \((\theta, \phi)\), we need to find all unit vectors \(\mathbf{v}’\) such that:
\[
\mathbf{v} \cdot \mathbf{v}’ = 0.
\]
### Steps to Find Orthogonal Spherical Angles:
1. **Express \(\mathbf{v}’\) in spherical coordinates**:
\[
\mathbf{v}’ = (\sin\phi’ \cos\theta’, \sin\phi’ \sin\theta’, \cos\phi’).
\]
2. **Set up the orthogonality condition**:
\[
\mathbf{v} \cdot \mathbf{v}’ = \sin\phi \cos\theta \cdot \sin\phi’ \cos\theta’ + \sin\phi \sin\theta \cdot \sin\phi’ \sin\theta’ + \cos\phi \cdot \cos\phi’ = 0.
\]
3. **Simplify the equation**:
\[
\sin\phi \sin\phi’ (\cos\theta \cos\theta’ + \sin\theta \sin\theta’) + \cos\phi \cos\phi’ = 0.
\]
Using the trigonometric identity \(\cos(\theta – \theta’) = \cos\theta \cos\theta’ + \sin\theta \sin\theta’\), this becomes:
\[
\sin\phi \sin\phi’ \cos(\theta – \theta’) + \cos\phi \cos\phi’ = 0.
\]
4. **Solve for \(\phi’\) and \(\theta’\)**:
This equation defines a relationship between \(\phi’\) and \(\theta’\). There are infinitely many solutions, corresponding to all directions orthogonal to \(\mathbf{v}\). Geometrically, these directions lie on the great circle perpendicular to \(\mathbf{v}\).
### Key Observations:
– For a fixed \(\theta\), \(\phi’\) and \(\theta’\) must satisfy the above equation.
– If \(\phi = \pi/2\) (the original direction lies in the \(xy\)-plane), then \(\cos\phi = 0\), and the equation simplifies to:
\[
\sin\phi’ \cos(\theta – \theta’) = 0.
\]
This implies \(\phi’ = 0\) or \(\cos(\theta – \theta’) = 0\). The latter corresponds to \(\theta’ = \theta \pm \pi/2\), which are directions perpendicular to \(\mathbf{v}\) in the \(xy\)-plane.
– If \(\phi = 0\) (the original direction is along the \(z\)-axis), then \(\sin\phi = 0\), and the equation simplifies to:
\[
\cos\phi’ = 0.
\]
This implies \(\phi’ = \pi/2\), meaning all directions in the \(xy\)-plane are orthogonal to \(\mathbf{v}\).
### Summary:
The set of all spherical angles \((\theta’, \phi’)\) orthogonal to \((\theta, \phi)\) is determined by the equation:
\[
\sin\phi \sin\phi’ \cos(\theta – \theta’) + \cos\phi \cos\phi’ = 0.
\]
This defines a great circle of directions perpendicular to the original direction. The specific solutions depend on the values of \(\theta\) and \(\phi\).

Posted by: Andrew Sarchus | Jan 29 2025 16:37 utc | 330

Some have claimed DeepSeek cant do math(s).
Posted by: Andrew Sarchus | Jan 29 2025 16:37 utc | 343

Where is WebLaTeX when you need it?

Posted by: too scents | Jan 29 2025 17:00 utc | 331

China has a long history, millenial, of respecting learning and intellectual achievement. Chinese grade schools and middle schools are very academic and pay a lot of attention to science and math. The top political leaders are nearly all engineers, rational, able to read, write, figure, do a cost/benefit analysis.
History is about 400 years old, counting the colonies, about 250 years of political independence. We have always imported our scientists and engineers. Academic training has mostly turned out lawyears and preachers. Dodgy businessmen who made themselves rich via real estate con games have long been admired.
There should be no surprise tha PR China is pulling ahead in science. The US leads in religious nut theories like Young Earth Creationism, Dispensationalism or the Health and Wealth Gospel. Flat Earth belief has become common. Immigrant scientists are no longer welcome.

Posted by: lester | Jan 29 2025 20:33 utc | 332

Does AI DeepSeek panic remind you of Sputnik panic?

Posted by: lester | Jan 29 2025 21:17 utc | 333

The good news — or bad, depending on denouement — is the source has already left the barn. Perhaps the play’s the thing wherein the King catches it.

Posted by: Laurence | Jan 30 2025 3:12 utc | 334