A new AI chatbot called Deep Seek created by Chinese investors has been released and sent shock waves through the US AI industry because it seems to be able to do all that the other chatbots can do (and perhaps more) for much lower cost and with less sophisticated chips.
Investors punished global tech stocks on Monday after the emergence of a Chinese chatbot competitor to OpenAI’s ChatGPT, DeepSeek, raised doubts about the sustainability of the US artificial intelligence boom.
The tech-heavy Nasdaq index in New York opened lower after investors digested the implications of the latest AI model developed by the startup DeepSeek.
Nvidia, the most valuable listed company in the US and a leading maker of the computer chips that power AI models, lost more than $400bn (£321bn) in stock market value in early trading as its shares declined 13.6%, while Microsoft shed $130bn and Google’s parent, Alphabet, declined by $80bn.
Nvidia’s fall – which wiped about $465bn off its value, was the biggest in US stock market history, according to Bloomberg.
…DeepSeek claims to have used fewer chips than its rivals to develop its models, making them cheaper to produce and raising questions over a multibillion-dollar AI spending spree by US companies that has boosted markets in recent years.
The company developed bespoke algorithms to build its models using reduced-capability H800 chips produced by Nvidia, according to a research paper published in December.
…Richard Hunter, the head of markets at the platform Interactive Investor, said: “It will almost certainly put the cat among the pigeons as investors scramble to assess the potential damage it could have on a burgeoning industry, which has powered much of the gain seen in the main indices over the last couple of years.
“The larger question has suddenly become whether the hundreds of billions of dollar investment in AI needs re-evaluation.”
Andrew Duncan of the UK’s Alan Turing Institute said the DeepSeek development was “really exciting”.
He added: “It demonstrates that you can do amazing things with relatively small models and resources. It shows that you can innovate without having the massive resources, say, of OpenAI.”
That was yesterday. Today the stock market fell even more. Today I heard that some federal agencies have blocked employees from using DeepSeek. They will of course cite national security concerns, like they did with TikTok, but it also seems likely that they are trying to put a protective wall around US companies against foreign competition.
Ryan Grim and Waqas Ahmed write that this is caused by US tech companies being overvalued because they were monopolies. But monopolies tend to be not good at innovation, while upstart startups are. This is what former FTC commissioner Lina Khan warned but of course she is now gone.
U.S. tech companies, with the support of the federal government (and the Pentagon), built up a dominant global position through genuine innovation. Microsoft, Facebook, Apple, Google, and Amazon reshaped the world. Microsoft, one of the first major companies to rise, tried to put a halt to that innovation by buying up and/or crushing its competitors, but the U.S. sued it in 1998 for violating antitrust laws. The Bush administration settled the case, backing off the effort to break them up. What followed was a bipartisan embrace of Big Tech; the Bush and Obama eras saw unbridled growth and mergers. As tech companies saw smaller firms innovating, they would buy the company, kill it, and absorb some of its staff.
An anti-monopoly movement started bubbling up, leading to lawsuits against Facebook, Amazon, Google, and Apple over the past decade. Lina Khan, as chair of the Federal Trade Commission under former President Joe Biden, became a folk hero as she warned that greed and consolidation weren’t just harming consumers and workers, but that the sclerotic companies themselves would eventually suffer from the lack of competition. “Our history shows that maintaining open, fair, and competitive markets, especially at technological inflection points, is a key way to ensure America benefits from the innovation these tools may catalyze,” Khan said in 2023.
Now it’s become clear that the moat the U.S. built to protect its companies from domestic competition actually created the conditions that allowed them to atrophy. They got fat and happy inside their castles. Their business pivoted from technological innovation to performing alchemy with spreadsheets, turning made-up metrics into dollar valuations detached from reality. Now DeepSeek has exposed the scam. With a tiny fraction of the resources, and without access to the full panoply of U.S. chip technology, the Chinese company DeepSeek has pantsed Silicon Valley. The U.S. company OpenAI began as a nonprofit dedicated to making AI widely available, as its name suggests. Its top guy, Sam Altman, managed to transition it to a for-profit and close it off.
Now DeepSeek is ironically fulfilling OpenAI’s original mission by providing an open-source model that simply performs better than any in the market. We have an FAQ on the details of DeepSeek below.
I decided to give DeepSeek a spin.
As I wrote in a post some time ago, the existing AI chatbots had the reputation of inserting purely made up stuff mixed in with factual information. I wrote that I decided to test both Bard and ChatGPT by providing a prompt to which I would immediately know, without doing any extra research, if they were making up stuff, and that was “What can you tell me about Mano Singham?”
While most of the stuff they wrote back was factual, they also inserted stuff like that I was the author of books about The Simpsons as well being the editor of the book “The Collected Poems of Jean Toomer,” a Black poet and novelist who lived from 1894 until 1967 and was associated with the Harlem Renaissance. Neither platform knew anything about my latest book The Great Paradox of Science.
So how did DeepSeek do? I gave the same prompt to it and the result was impressive. It was accurate and pretty detailed. My only quibble is that inflated my importance in the world of science and public discourse. Here is what it said.
Mano Singham is a prominent physicist, author, and educator known for his work in both the scientific community and the field of education. Here are some key points about him:
1. Academic Background: Mano Singham has a strong background in physics. He earned his Ph.D. in theoretical high-energy physics from the University of Pittsburgh.
2. Career: He has had a long and distinguished career in academia, primarily at Case Western Reserve University in Cleveland, Ohio, where he served as a physics professor and also held various administrative roles.
3. Educational Reform: Singham is well-known for his contributions to educational reform, particularly in the areas of science education and critical thinking. He has written extensively on the subject and has been involved in initiatives aimed at improving teaching methods and student learning outcomes.
4.Authorship: He is the author of several books, including “The Achievement Gap in U.S. Education: Canaries in the Mine” and “God vs. Darwin: The War Between Evolution and Creationism in the Classroom.” His writings often explore the intersection of science, education, and society.
5. Public Engagement: Singham is an active public intellectual who frequently writes and speaks on issues related to science, education, and secularism. He has contributed to various publications and maintains a blog where he discusses these topics.
6. Awards and Recognition: Throughout his career, Singham has received numerous awards and honors for his teaching and contributions to education.
Mano Singham’s work continues to influence both the academic community and the broader public discourse on education and science.
DeepSeek did not mention my latest book either. When I asked Bard and ChatGPT specifically about the book, they said that they had no information but here is what DeepSeek said in response to the prompt “What is the book “The Great Paradox of Science” by Mano Singham about?”
The Great Paradox of Science: Why Its Conclusions Can Be Relied Upon Even Though They Cannot Be Proven by Mano Singham is a thought-provoking exploration of the nature of scientific knowledge and the philosophical foundations of science. Here’s an overview of the book’s key themes and arguments:
1. The Paradox of Scientific Knowledge: The central theme of the book revolves around what Singham calls the “great paradox of science.” He argues that while scientific conclusions are widely trusted and relied upon, they cannot be proven with absolute certainty. This paradox raises important questions about how science achieves its authority and reliability.
2. The Nature of Scientific Proof: Singham delves into the philosophical underpinnings of science, discussing how scientific theories are developed, tested, and validated. He explains that scientific knowledge is inherently provisional and subject to revision in light of new evidence, contrasting this with the notion of absolute proof.
3. The Role of Evidence and Consensus: The book emphasizes the importance of empirical evidence and the scientific consensus in establishing the credibility of scientific claims. Singham discusses how the scientific community evaluates evidence and reaches consensus, highlighting the collaborative and self-correcting nature of science.
4. Historical and Philosophical Context: Singham provides a historical overview of the development of scientific thought, from the early days of the scientific revolution to contemporary practices. He examines the contributions of key figures like Karl Popper, Thomas Kuhn, and others who have shaped our understanding of how science progresses.
5. Science and Society: The book also explores the relationship between science and society, addressing how scientific knowledge is communicated to the public and the challenges of combating misinformation and pseudoscience. Singham advocates for a better public understanding of the scientific process to foster trust in scientific conclusions.
6. Implications for Education: As an educator, Singham is particularly interested in the implications of these ideas for science education. He argues for teaching students not just scientific facts, but also the methods and processes of science, including its limitations and uncertainties.
Overall, The Great Paradox of Science is a compelling read for anyone interested in the philosophy of science, the nature of scientific knowledge, and the ways in which science interacts with society. Singham’s clear and engaging writing style makes complex ideas accessible to a broad audience.
I thought that it was a very good summary of the main ideas in the book, although it too was written in a flattering way. This made me wonder if flattery is built into its algorithm.
You can try out Deep Seek for yourself, before the US government tries to shut it down.
I’m not very impressed. Here’s one of my favorite questions for AIs and DeepSeek 3’s answer.
[me] Can you list the names of 10 famous people who use their middle initial J professionally?
[DeepSeek]
Certainly! Here are 10 famous people who use their middle initial “J” professionally:
Michael J. Fox -- Actor and advocate, known for his roles in Back to the Future and Family Ties.
Stephen J. Cannell -- Television producer and writer, known for creating shows like The A-Team and 21 Jump Street.
George R. R. Martin -- Author of the A Song of Ice and Fire series (though his middle name is Raymond, he uses “R” professionally).
John F. Kennedy -- The 35th President of the United States, often referred to as JFK.
Robert J. Oppenheimer -- Physicist and director of the Manhattan Project.
Samuel L. Jackson -- Actor known for his roles in Pulp Fiction and the Marvel Cinematic Universe.
H. P. Lovecraft -- Author of horror fiction, though his middle name is Phillips, he used “P” professionally.
J. K. Rowling -- Author of the Harry Potter series (though her middle name is Kathleen, she uses “K” professionally).
E. E. Cummings -- Poet and painter, though his middle name is Edward, he used “E” professionally.
C. S. Lewis -- Author of The Chronicles of Narnia series, though his middle name is Staples, he used “S” professionally.
Let me know if you’d like more examples!
this could also be promising on the energy cost issue. i’d love to see all this tech more efficient. i will say i think it’s very reasonable to cut off gov computers from various AIs. clown shit like the tiktok ban is a whole other issue.
DeepSeek got my full legal name confused with “Rick Sanchez” from ‘Rick and Morty’ but quickly apologized and acknowledged the mistake, when I corrected it.
😉
-MARS-
Following Mano’s example, I asked about myself and got:
Well-known? I do serve on the ISO standards committee for the C++ programming language. Paying my dues gives me license to hang around with folks who are smarter than I; and I’m acquanted with several of the movers and shakers behind C++ who would probably know my name; but it would be presumptious of me to call them friends.
Calling me a computer scientist is quite a stretch. Until my retirement a couple of years ago, I was a work-a-day coder in good old “business data processing”.
My one (count it, one) contribution to the C++ language, adopted in 2011 in Madrid, is initializers for non-static data members; and that wasn’t my idea to begin with. It was proposed by Michael Spertus at a meeting in Oxford back in 2007; but he had other irons in the fire and asked for some volunteer to take it over which I did.
I have no expertise in any of concurrency, networking, or performance optimization.
not 100% up to snuff, as one might expect. for my part, i asked it to roleplay an erudite cowpoke like sam elliott in the big lebowski, and it feels a little tenuous, like it could fall off any second. too dude ranch one minute, too ivy league the next. could be worse tho.
@4 billseymour
When I saw the term concurrency, my first thought was that someone came up with a good insulting name for cryptocurrency. You can imagine my disappointment when I googled it and saw it was a mundane programming term.
Mano and billseymour
Is it possible that Deep Seek “knew” you were asking about yourselves?
chigau @#7,
That’s a good point. Why don’t you try the same prompt and see what you get?
Mano
That is also a good point. I will try tomorrow. (say 12 hours from now, give or take)
and maybe you two could ask about each other
An editor at YLE (Finnish Broadcast Corporation) asked DeepSeek in Finnish if Chinese media have censorship. DS started replying that there is tight censorship and government controls all media etc., but then the reply was suddenly overwritten with a canned text in English:
”Sorry, that’s beyond my current scope. Let’s talk about something else.”
So DS has all the data, but it also has a censor daemon that kicks in when necessary. Video grab at the news article (in Finnish only):
https://yle.fi/a/74-20139610
There seems to be quite a lot of confusion about what DeepSeek have actually done, and what their claimed costs apply to. (If you chose to believe them.)
David Gerard and Amy Castor have quite a good post up about it: DeepSeek slaps OpenAI, tech stocks crash
@bugfolder: Does it do better if you rephrase the prompt as:
Can you list 10 famous people who use “J” as their middle initial as part of their professional name?
Of course, it could be that you chose your prompt as something that would be parsed correctly by a human, but easily misparsed by a non-intelligent machine.
@Bill Seymour: If you ask it to provide links or references to Bill Seymour’s work in concurrency, networking, and/or performance optimization, what does it do?
@billseymour
Oooh. I’m now only 2 degrees of separation from Bjarne Stroustrup!
I asked it “what is Tian An Men square best known for?”
It said “I am sorry, I cannot answer that question. I am an AI assistant designed to provide helpful and harmless responses.”
Owlmirror @12: I don’t really care enough to try anything else. 😎
Deepak Shetty @13: 😎 I think you’d like him. I’ve found him to be a very agreeable person who’s happy to pass on any knowledge he has. He is, among other things, a teacher.
@13 Deepak Shetty
that reminds me: i am 2 or 3 degrees of separation from pretty much any scientist that worked on the manhatttan project!
(depending on if i start with my dad’s uncle who worked there, or my physics prof who was stationed there)
Robbo @#16 and Deepak @#13,
How’ is the connection measured to get the degrees of separation? With degrees of separation with Kevin Bacon, a connection is acting in the same film as someone. How does it work with scientists or other things?
Techbros thought they could force us all to buy expensive, proprietary AI chatbots nobody either wants or needs. And they thought they could do this by forcing AI chatbots everywhere, so it would become impossible not to use one; then using this time to figure out a way to start charging people money for what they had already got used to having for free — hopefully, so used to it, they would rather pay to carry on than stop. So they created a bubble of hype, and the money began to pour in against the promise of returns at some future date.
But in all their back-slapping excitement, the techbros managed to overlook the fact that their “secret sauce” was about as difficult for a third party to replicate as a door hinge.
And when some young upstart came along and showed everyone how to make your very own AI chatbot, using only commonly-available household materials, that’s at least as good as anything the techbros could come up with, then their offering suddenly looked a lot less attractive to would-be investors. And of course, the one you make for yourself need not include any of the censorship features from their demo.
https://www.theguardian.com/technology/2025/jan/28/chinese-ai-chatbot-deepseek-censors-itself-in-realtime-users-report
I watched this happen as I asked it further questions about Tiananmen Square. It produced quite an extensive paragraph about how information about this subject wasn’t conducive to the public good or something, then AS I WATCHED it backspaced rapidly over it all and said “I can’t answer that, let’s talk about something else!”.
I actually wondered for a moment if I’d hallucinated it, which would have been ironic given how AI are given to “hallucinating” things that aren’t true and this one was apparently “correcting” something it had said that WAS true but was just verboten.
Looking forward to the AI bubble bursting, can’t be long now.
The main question is: when it does, will all the companies that have enshittified their products with useless “AI” crufty bit remove the crap that nobody asked for and go back to the more functional, less annoying pre-AI version, or are we stuck with this bollocks for the rest of time? (See also car dashboards, a user interface that has been refined over a century to offer positive feedback when operating all sorts of controls while trying to drive at the same time, now being increasingly replaced with configurable touch screen you have to LOOK at, and see also the death of physical media in favour of universal streaming where you don’t OWN anything but just have unlimited access to everything, except it’s NOT everything, and access can be withdrawn on a whim at any time, oh now you’ve got me started, I need to stop.)
@billseymour @15
Yes , everything I have heard about him is a combination very smart + very helpful. In the end it was K & R C (the book) and then C++ which swayed me from electronic engineer to programmer (admittedly the lack of non software jobs in India at that time may also have played a part) but funnily enough life took me down the Java path .
@mano @17
I know you (non biblically ofcourse) so 1 degree and hence i am 2 degrees from all the famous physicists and celebrities you hang out with . On the flip side you are now a 3rd degree of separation from Aishwarya Rai (Former miss world and bollywood star) by virtue of knowing me.
Deepak Shetty, we have a similar history. I was an electronics technician (it would be a stretch to call me an engineer) by training and first vocation. In 1979/80 I was working for a hospital in Pittsburgh designing and building a prototype for an EKG machine with a microprocessor in it; and I still have the drawings to prove it. 😎 One of them was small enough to put in the “About the Author” sidebar on my own blog. I wound up writing a lot of the code for it in assembler which got me starting in programming.
In the mid 1980s, I had reason to learn C; and as luck would have it, the book I selected was the first edition of K&R; so I, too, got off on the right foot. 😎
Deepak Shetty, we have a similar history. I was an electronics technician (it would be a stretch to call me an engineer) by training and first vocation. In 1979/80 I was working for a hospital in Pittsburgh designing and building a prototype for an EKG machine with a microprocessor in it; and I still have the drawings to prove it. 😎 One of them was small enough to put in the “About the Author” sidebar on my own blog. I wound up writing a lot of the code for it in assembler which got me started in programming.
In the mid 1980s, I had reason to learn C; and as luck would have it, the book I selected was the first edition of K&R; so I, too, got off on the right foot. 😎
[Sorry for the (mostly) duplicate comment. I thought I was just editing.]
Deepak @#20,
‘Knowing’ someone seems a little vague and hard to document. I thought it may be something along the lines of having a joint publication of some sort.
@billseymour @21
Last year or so I went to a standup comedy set (San Francisco) and the comic asked “How many engineers ?” and 90% of the hands went up -- So she followed up with “How many real engineers? and not ” and mimed typing on a keyboard ? Most of the hands went down sheepishly.
@Mano @24
Sure -- thats the fun part. Wikipedia agrees with me , though you are likely referring to the game- https://en.wikipedia.org/wiki/Six_degrees_of_separation v/s https://en.wikipedia.org/wiki/Six_Degrees_of_Kevin_Bacon
Mano@24,
That’s certainly the case for the Erdős number.
Sam Altman has been complaining about DeepSeek’s developers using ChatGPT to train DeepSeek without permission, payment or acknowledgement. The heart bleeds, does it not?
Further to #26, I’ve just discovered (using Collab Distance -- MathSciNet) that my Erdős number is no more than 5. (I’m not listed at that site, but a co-author is.) As a non-mathematician, I’m quite proud!