-+ 0.00%
-+ 0.00%
-+ 0.00%
On July 22, Michael Krazios, director of the US White House Office of Science and Technology Policy, suddenly claimed on social media that the Chinese AI company “Dark Side of the Moon” was developing the Kimi K3 model by distilling the American company Anthropic's Fable. Just a few hours later, US Treasury Secretary Scott Bessent issued a letter threatening to use sanctions and an “entity list” to deal with the distillation of Chinese companies. Hype up the “intellectual property theft issue” before imposing sanctions on Chinese entities. This is a common pattern for the US to suppress China's technological development in recent years — pretending to be a victim before harming people. For people who don't know the truth, this trick is often somewhat confusing. However, the US routine was bound to “turn over” this time, because the claim that Kimi K3 was developed by distillation Fable is too laid-back. One obvious problem is that just like copying can't produce good results, Kimi K3's advanced performance can't be obtained by distillation at all. Undoubtedly, the root of Kimi K3's success is technological innovation. As Xu Daohui, professor of theoretical physics and computational mathematics, science and engineering at Michigan State University in the US, said on social media, Kimi K3 has achieved extraordinary neural network architecture innovation. It is an original, large-scale architecture built from scratch, rather than compressing and imitating other large models through distillation. Xu Daohui mentioned several important papers published by Kimi's team. One introduced the K3 model's use of a “linear attention mechanism” to achieve an ultra-long context window of 1 million words; the other article outlined how to manage “attention” to increase the training efficiency of the big model by 25%. The world can see China's development and achievements in the field of AI. Some US politicians are in key positions relating to the development of science and technology in the US. They should be insiders; they should not always use outsider jargon; they insult China's big model by relying on distillation. This only makes them look very amateurish.
Share
Listen to the news
On July 22, Michael Krazios, director of the US White House Office of Science and Technology Policy, suddenly claimed on social media that the Chinese AI company “Dark Side of the Moon” was developing the Kimi K3 model by distilling the American company Anthropic's Fable. Just a few hours later, US Treasury Secretary Scott Bessent issued a letter threatening to use sanctions and an “entity list” to deal with the distillation of Chinese companies. Hype up the “intellectual property theft issue” before imposing sanctions on Chinese entities. This is a common pattern for the US to suppress China's technological development in recent years — pretending to be a victim before harming people. For people who don't know the truth, this trick is often somewhat confusing. However, the US routine was bound to “turn over” this time, because the claim that Kimi K3 was developed by distillation Fable is too laid-back. One obvious problem is that just like copying can't produce good results, Kimi K3's advanced performance can't be obtained by distillation at all. Undoubtedly, the root of Kimi K3's success is technological innovation. As Xu Daohui, professor of theoretical physics and computational mathematics, science and engineering at Michigan State University in the US, said on social media, Kimi K3 has achieved extraordinary neural network architecture innovation. It is an original, large-scale architecture built from scratch, rather than compressing and imitating other large models through distillation. Xu Daohui mentioned several important papers published by Kimi's team. One introduced the K3 model's use of a “linear attention mechanism” to achieve an ultra-long context window of 1 million words; the other article outlined how to manage “attention” to increase the training efficiency of the big model by 25%. The world can see China's development and achievements in the field of AI. Some US politicians are in key positions relating to the development of science and technology in the US. They should be insiders; they should not always use foreign jargon; they insult China's big model by relying on distillation. This only makes them look very amateurish.
Disclaimer:Webull uses external vendor Google Translation Service for news translations where we endeavour to ensure these are correct, however, we recommend that you please double-check this information accordingly. Webull is not responsible for translation errors or issues.
What's Trending