Lambert says distillation is not clearly harmful to US AI competitiveness

Nathan Lambert says on X that it is "not cut and dry" whether distillation harms US AI competitiveness, replying to institutions examining the question.

He was responding to a post by Archie Hall, who argues in The Economist that efforts to ban Chinese models trained partly on American models are misguided.

Hall says distillation can help price competition and national security. He also says US developers' intellectual property claims are weak, because they defend their own use of training data as fair use.