The ongoing competition between the U.S. and China in artificial intelligence has intensified, particularly around the concept of 'distillation,' which involves training AI models using outputs from more advanced models. U.S. companies and officials have accused Chinese labs of leveraging this technique to gain an unfair advantage, labeling it as theft.
Aidan Gomez, CEO of Cohere and a key figure in AI research, counters this narrative by asserting that while some distillation occurs, it does not fully account for the impressive capabilities of Chinese AI models, which have begun to outperform their American counterparts in certain benchmarks. This perspective challenges the prevailing view among U.S.
AI labs, such as Anthropic, which has repeatedly warned about the risks posed by Chinese AI development. The Chinese government has dismissed U.S. allegations as unfounded, and analysts suggest that restrictions on advanced semiconductor technology have inadvertently spurred innovation within China.
As the landscape evolves, there are concerns that Western policymakers may be underestimating China's technical advancements, which could have significant implications for the global AI market and competitive dynamics