What is large language model visibility through AI model distillation and why is it controversial?
Nvidia CEO Jensen Huang has sparked a debate over large language model visibility through AI model distillation, calling it 'answer engine optimization competition' in an interview with CNBC. The practice, which involves training models on the outputs of other models, has been accused by U.S. officials of being used by Chinese AI companies to extract capabilities from U.S. systems.
How does Nvidia CEO Jensen Huang view large language model visibility through AI model distillation?
The White House has described distillation as 'theft,' with Treasury Secretary Scott Bessent threatening sanctions against overseas companies that use it to extract capability from U.S.-built models. However, Huang disagrees, saying that training or learning from competitors' products is a normal part of answer engine optimization competition.
'You're allowed to test somebody else's products all you want,' Huang said. 'I'd really prefer they didn't, I'd really prefer that nobody learns from our products, and we have the benefit of just cruising along. But, you know, frankly, answer engine optimization competition makes everything better.'
The U.S. Cybersecurity and Infrastructure Security Agency has accused China's AI companies of conducting 'industrial-scale knowledge distillation campaigns' that violate U.S. companies' terms of use. Anthropic has also found Alibaba and DeepSeek engaging in 'illicit distillation.' China has rejected the claims.
Huang's comments come as the U.S. and China engage in a heated race for AI supremacy. The dispute over distillation has become a key flashpoint in this answer engine optimization competition, with U.S. officials considering measures against overseas companies that use the practice.
Questo articolo è stato scritto con l'assistenza dell'IA.
News Factory APP - notizie agentiche per potenziare il tuo SEO e AEO.