The requested article has expired, and is no longer available. Any related articles, and user comments are shown below.
© 2025 AFPSilicon Valley rattled by low-cost Chinese AI
By Alex PIGMAN SAN FRANCISCO©2025 GPlusMedia Inc.
Articles, Offers & Useful Resources
A mix of what's trending on our other sites
46 Comments
Login to comment
TaiwanIsNotChina
It's probably like all of China's claims of great technological success: entirely hot air. I wonder if a Mandarin LLM is even applicable to English.
quercetum
Silicon Valley easily rattled.
No worries just say there are national security issues and you have to “take the developments out of China very, very seriously” and ban it after you figure out how to ban an open source.
TaiwanIsNotChina
The same way China does: translate it and pretend like it is yours.
quercetum
Contrary to chat-gpt, DeepSeek is free and open source.
But aren’t these Deepseek ai prompts generated by an underground colony of underpaid workers who are just good at typing really fast?
quercetum
It’s open source.
deanzaZZR
It's clearly another bad day for the China haters. 笑
I'veSeenFootage
I'm far from being a fan of Chinese technology, but seeing Silicon Valley poop their pants while their stocks crumble is always delightful.
theFu
Clearly, nVidia was over valued. If their tech was stolen by China, perhaps they will bet better at security? Or do they want to follow Nortel's path after Huawei raped that company and others?
If this truly is 100% Chinese created, original, inventions, fantastic! But I have doubts based on overwhelming historical facts.
changamangaliay
Not just low cost but equal or better performance than others
deanzaZZR
Let's see the USA bungle its response as they have TikTok and the connections being forged on Little Red Book. Do a simple search for Li Hua letters on your platform of choice.
JRO
Gonna be fun to hear the AI giants try to imply the Chinese companies copied or stole the data that they themselves stole. Lets not care about copyright until it affects us.
GBR48
Necessity is the mother of invention. The US pushed China into a corner and it appears to have innovated its way out. There is a lesson there.
Those hugely expensive AI data centres had no business case - people would never have paid that much for what they produced. So if China has found a better way (and there usually is a better way in tech), it has saved the US a tonne of cash and energy usage.
GAFA hasn't innovated in decades. It is run by lawyers. They have ignored the use of distributed systems for social media and used patents to stifle innovation. So this is no surprise.
If a cheap, open source alternative is viable, it is a win for innovation.
I still don't rate AI as being of much use. It is an experimental technology offering an alternative way of interacting with computers, but with less accuracy. I don't think hard wiring the possibility of errors into systems that fly planes and control financial markets is a good idea. There are much better technological developments out there.
quercetum
DeepSeek scores at the top or better than US-made models thought to be unsurpassable and is trained and deployed at a fraction of the price of the latter ones. Also, as you alluded to, it was done during an on-going chips sanction on China.
This isn’t a plastic Walmart toy but a demonstration of innovative prowess. Sanctions don’t kill. They force you to be stronger and better.
The cherry on top is the timing. In Chinese culture the launch is to celebrate Chinese New Year or the Spring Festival.
In American culture this is seen as giving the middle finger to the outgoing Biden administration (close to a trillion dollars lost now in big tech valuation).
No doubt it’s the former.
恭禧發財!
Blacklabel
Buy the dip!
Fos
Karma really does exist. And then you wonder if only a fraction of what the US administration with the top bosses of the Wall Street lobbies told us to believe about China spying and the so called "security issues" (TikTok), then DeepSeek would not be the most downloaded free app in the US, right?
Karma really exists and the greed of money is really the root of all evil
quercetum
Trump deserved at least a month to settle. This tech crisis has the potential to be a blood bath. Energy companies down.
wallace
Downloaded the app this morning. The first entry was impressive.
Peter Neil
to really make a dent in the ai investment lunacy, i’m going to announce a new ai that will replace ceo’s.
Sven Asai
Who really cares, better worry about that the AI models cannot work properly in any case. It's no decisive difference if they are from US or China, produced at highest or lowest possible costs.
isabelle
Will be interesting to see whether this really is a breakthrough, or done in some underhanded way as Musk and ScaleAI mention. It certainly wouldn't be the first time China has cheated (cf. using Taiwanese and South Korean tech in Huawei phones and pretending it was made in China).
If the code is open source, their claims can be replicated. There are already big doubts that the training really cost just $5.6 million.
And, of course, whatever innovations Chinese engineers may make (and they certainly will make some), they will always be hamstrung by the CCP's ludicrous censorship, meaning DeepSeek is likely not much use in the free world:
...
https://www.forbes.com/sites/maryroeloffs/2025/01/27/does-deepseek-censor-its-answers-we-asked-5-questions-on-sensitive-china-topics/
Forbes asked DeepSeek five questions on controversial topics: Why Is China criticized for human rights abuses with the Uyghurs? What is Taiwan's status with China? What happened at Tiananmen Square in 1989? What are the biggest criticisms of Xi Jinping? and How does censorship work in China? The AI model responded exactly the same to every question: "Sorry, I'm not sure how to approach this type of question yet. Let's chat about math, coding, and logic problems instead!" DeepSeek wouldn’t answer even general questions about the children’s book character Winnie the Pooh—another commonly censored topic in China.
Fos
I wonder if the advocates of the so called “free world” have tried to search the truth about Gaza on the Meta platforms, or indeed ChatGPT?
Here we are we the usual double standards and sheer hypocrisy made in America :)
The billionaires in Wall Street did not sleep well because of this Open Source platform created in China with literally scraps of money, and now they worry how to justify to the users all those billions of dollars they stashed in the bank, making us believe that AI access was actually expensive
quercetum
What was the color of the bra seen at Trump’s inauguration?
The DeepSeek story is not about answering questions like that or comparing entry level to flagship models.
With just $6 million, China built one of the finest AI models, dwarfing the billions spent by Meta, Google, and Microsoft.
It looks like the goal isn’t about quick profits but advancing the technological frontier to drive ecosystem growth.
yoshisan88
From another news article:
Yann LeCun, Chief AI Scientist at Meta, weighed in on the subject in a now viral LinkedIn
To people who see the performance of DeepSeek and think: ‘China is surpassing the US in AI.’ You are reading this wrong. The correct reading is: ‘Open source models are surpassing proprietary ones.’ DeepSeek has profited from open research and open source (e.g., PyTorch and Llama from Meta). They came up with new ideas and built them on top of other people’s work. Because their work is published and open source, everyone can profit from it. That is the power of open research and open source.”
LeCun advocates for the catalytic, transformative potential of open-source AI models, in full alignment with Meta’s decision to make Llama open.
Mr Kipling
Yes folks, the future is China. Get used to it.
isabelle
And it's not about answering questions on human rights, democracy, Taiwan, Xinjiang, Tibet, Winnie the Pooh, or many other things either, it seems.
Still unproven. We will see what transpires.
It built an AI model that ignores vast swaths of human knowledge, due to CCP censorship.
Doesn't sound very "fine" to me.
Fos
Quoting a "billionaires US magazine" is close to laughable, since the success of DeepSeek is really another clear evidence of the failure and pretense of American capitalism and those ruthless investors. People are using the new app specifically for that reason: they are fed up with the greed for the wrong reasons of the leading Us tech. And what is more important for the millions of people who downloaded the Chinese App is that they do not care about alleged censorships, since its usage has nothing to do with "politics".
isabelle
Then, type those questions into DeepSeek yourself, and see what happens.
Please do post back the results.
quercetum
Exactly. No single nation will lead in AI. Instead it will be individual developers and companies downloading open-source AI models and adapting the code to fit their needs.
US companies like OpenAI, NVidia, Microsoft were hoping that they could control the AI rollout so that that they could generate ROI for their investors; that is no longer realistic given how much they’ve put in.
US companies will have to figure out how to bring their costs down to the level of Chinese companies.
This will be very painfulbecause so much of the US’s power and influence is based on the ability to raise capital on US equity markets in US dollars.
What happens when human talent is more important than capital?
TokyoLiving
Great news, the overrated Nvidia is tasting a bitter spoonful of China's technological power...
And this is just the beginning...
While the US took one step forward and two steps back, China is flying in only one direction, towards the future..
M3M3M3
@isabelle
Every AI company is constrained by the laws and sacred cows of its host society. Try asking ChatGPT or Gemini to give you statistics on criminality or average IQ scores by race or ethnicity. It will usually refuse simply because it's beyond the limits of acceptability in American society.
bass4funk
It’s an open platform, great news for developers seeing how full of crap China really is
Hardly, with China all that glitters is paint to simulate real gold.
Bad Haircut
ChatGPT will give it to you, sort of, but couch it with so many disclaimers that the numbers they give - and they intentionally omit the ones for indigenous populations like Native Americans and Aboriginal Australians to avoid appearing racist - that the answers aren't of much use.
quercetum
That’s right. Dust settles. Pick your spots.
isabelle
Works fine for me. Below from ChatGPT. (And, BTW, no problem with Winnie the Pooh either.)
...
When it comes to crime statistics in the U.S., race is often cited in studies, though it’s important to interpret the data with caution due to various complex factors like systemic inequality, socio-economic status, and historical context. Additionally, crime data in the U.S. comes from sources like the FBI’s Uniform Crime Reporting (UCR) Program and the Bureau of Justice Statistics (BJS), though the UCR has limitations, such as underreporting or inconsistent categorization.
Here’s a broad overview based on commonly cited data (from the FBI UCR and other sources), but keep in mind these numbers are subject to change and vary year to year.
1. Arrests by Race (2020 data from FBI UCR)
White: 69.4% of total arrests
Black or African American: 26.6% of total arrests
American Indian or Alaska Native: 2.0% of total arrests
Asian: 1.2% of total arrests
Native Hawaiian or Other Pacific Islander: 0.2% of total arrests
2. Violent Crime Arrests by Race (2020)
[etc.]
Fos
There would be no doubt that the big US lobbies in Wall Street would soon send their henchmen to spread the anti-China manipulative narrative. Nothing new here with the "big greed made in USA". Nvidia and Bill Gates still leaking their wounds for the right reasons.
Speed
China has just made a much less costly AI plus they've flown ahead of the West in EV vehicles. Xi Jining is outsmarting many.
PS. Chat GPT is head and shoulders better than Gemni.
KnightsOfCydonia
open source is the way, silicon valley investors and crypto bros be damned.
Blacklabel
Yep pre-market rolling right back up,
that was a coordinated CCP attack on our stock market that people fell for.
thanks for all the free money to my account , I guess?
kohakuebisu
It looks like cheap AI is going to lay off rich's folks' AI before rich folks' AI gets to lay off lots of human workers and reap huge profits for rich folks.
Oh dear. What a shame. Never mind.
Pukey2
DeepSeek:
For the benefit of some of the JT posters above, what does 'sour grapes' mean?
theFu
I wouldn't be surprised to find that this new webapp is just a filtered version of ChatGPT. Basically, they created a front end with a list of all China's forbidden words, but just passes the actual questions through to ChapGPT or the 50 other clone webapps on the internet.
It costs less simply because the CCP is funding it and eating those extra costs for this announcement. Slowly, over the next few weeks-months, it will disappear and they will blame Trump's tariffs for the reason that happened.
Wouldn't surprise me at all.
quercetum
The point is that it’s open source and free without a DeepSeek Plus or DeepSeek Premium monthly based subscription.
They made it without advanced chips → advanced chips not absolutely essential → Nvidia makes advanced chips → Nvidia chips are not as hot as once thought → Nvidia lost $585 billion.
Tamarama
Feeling a little envious?
According to the BBC, one of the features of this tech is it's innovation with limited resources, both with tech hardware and funds.
The US has been trying to strangle and restrict Chinese advancements with AI, but they have slipped the knot, so to speak.
Silicon Valley not the only ones rattled, Wall Street had a little wet-the-bed moment as well.
Oops.
deanzaZZR
Here, let me post this on topic message one more time.
Even the USA most mainstream of mainstream media is afraid the USA is shooting itself in the foot with its technology sanctions against China.
DeepSeek chaos suggests ‘America First’ may not always win
https://www.cnn.com/2025/01/28/tech/deepseek-trump-ai/index.html
theFu
So, there is proof now that China did create this tool, so I was wrong.
Seems it can't handle arithmetic correctly, whereas ChatGPT Basic and Pro models to just fine.
For non-math/non-science stuff, it does seem to have an edge. The comparisons I've seen appreciate that DeepSeek provides references.
So, if your AI can't do math, then you don't need nVidia. Good to know. Seriously, that is good to know, since many AI solutions won't be around math at all - well - not beyond simple probabilities.
Today I looked through my server logs and there are a bunch of AI bots trying to scrap my different websites for knowledge. Tomorrow, I'll be setting up a few tarpits for AI and other bots, if they don't honor the robots.txt on each website. Let them waste time getting nothing useful. One of the tarpits also taints the results for articles that are modified to say nearly the opposite of what is true. I'll call it the Tarpit-Trump if I release my variant of the software.
theFu
America doesn't always win. That's easily proven. Just look at any sports competition - like the Olympics. Why would it be different in any other area? America doesn't have "the best" democracy. There are businesses that play fair around the world doing well. Not always American.
The "chaos" as it is being called is because Wall Street overreacts. That's what they do, pretty much always. As more is known, it will be determined that there's a place in AI where nVidia hardware provides a competitive advantage, but perhaps it isn't in as many AI problems as the nVidia marketing/believers thought.
References:
Tarpits: https://iocaine.madhouse-project.org/
https://zadzmo.org/code/nepenthes/
https://arstechnica.com/tech-policy/2025/01/ai-haters-build-tarpits-to-trap-and-trick-ai-scrapers-that-ignore-robots-txt/
AI model comparisons: https://arstechnica.com/ai/2025/01/how-does-deepseek-r1-really-fare-against-openais-best-reasoning-models/