Nadella calls distillation bans ironic as labs hoard data
microsoft ceo satya nadella criticized openai and anthropic for banning model distillation while training on public data, arguing it creates a 'reverse information paradox' where value concentrates with infrastructure owners rather than knowledge creators.
microsoft ceo satya nadella publicly criticized the practice of ai labs banning distillation—where smaller models learn from larger ones—while simultaneously relying on fair use defenses to train their own models on public data. nadella described the situation as 'ironic' and introduced the concept of a 'reverse information paradox,' where companies pay twice for ai: once with money and again with 'exhaust' data such as corrections and usage logs that providers can use to improve their own systems. he argues this dynamic transfers economic value from the companies creating knowledge to the infrastructure operators running the models. the comments come as microsoft positions its own infrastructure to help companies maintain control over their own learning loops.