Wednesday, February 25, 2026

Nature Deepseek Article may be Exaggerated

The 17 September 2025 Nature cover story was this article by China AI company Deepseek. Supposedly it had a hot new AI model trained at low cost.

They aRe widely accused of using pirated Nvidia chips that were illegally imported.

Now Anthropic claims that Deepseek and other China AI companies have been using thousands of bogus accounts to steal data from Anthropic models.

Deepseek has made some decent AI LLMs, but maybe not the scientific advance they claimed. Nature said that it investigated the validity of the Deepseek claims, but I don't know how it could know what was going on in China.

Anthropic could be lying. It really hates Chinese competition. So I am not necessarily drawing any conclusions here.

at
Labels:

1 comment:

  1. CFTMarch 1, 2026 at 5:10 AM

    Soooo, the Chinese AI company has been stealing data from the American AI company....which has stolen almost everything without compensation whatsoever from everyone else. Oh no. The tragedy. The irony. The utter inability to look in the mirror.

    Pot, Kettle.

    ReplyDelete

Subscribe to: Post Comments (Atom)

Has Quantum Supremacy been Achieved?

Dominik Hangleiter 2603.09901">writes in a new paper : Recently, I gave a couple of perspective talks on quantum advantage, one at...

  • Gell-Mann agrees with me about Bell
    I have occasionally argued that Bell's Theorem has been wildly misinterpreted, and that it doesn't prove nonlocality or anything in...
  • Woit dives into Bell non-locality
    Peter Woit writes : what’s all this nonsense about Bell’s theorem and supposed non-locality? If I go to the Scholarpedia entry for Bell’s...
  • Dr. Bee on the Quantum Cat
    Dr. Bee's latest video is on Schroedinger's Cat, and she concludes: What this means is that one of the following three assumptions ...