AI Trends

Ma Dongxi NLP Questions the U.S. Classification of Models by Nationality

Ma Dongxi NLPUpdated Nov 18 23:54

AI summary

The main post compares the U.S. focus on whether models are “American or Chinese” to China’s debate half a century ago over whether cattle raised by communes were “socialist or capitalist,” questioning whether this represents a return to the past. The quoted post, meanwhile, cites cases involving Kimi, DeepSeek, and others in knowledge distillation, pretraining, and synthetic data to challenge the criteria for classifying a model as “Chinese.”

Quoted post

Nathan Lambert

@natolambert

This is stupid. What counts as a chinese model? What about Thinky's model distilled from Kimi? What about composer, continued pretraining? What about Nemotron, using tons of DeepSeek synthetic data? What about that silly nous finetune for "american" values?

View original

This is stupid. What counts as a chinese model? What about Thinky's model distilled from Kimi? What about composer, continued pretraining? What about Nemotron, using tons of DeepSeek synthetic data? What about that silly nous finetune for "american" values?

View original post

Body · AI translation

🎯🎯

Half a century ago, a backward China was preoccupied with whether the cattle raised by communes were socialist or capitalist.

Now, the developed United States is astonishingly preoccupied with whether models are American or Chinese.

Who is moving forward, and who is turning back the clock on history?

Source

Ma Dongxi NLP