Altman: skeptics held AI back by betting against scaling laws
openai's sam altman says researchers underestimated scaling laws, slowing progress, as a new model proves llms can generate novel mathematical knowledge.
speaking at stanford, openai ceo sam altman stated that a generation of researchers hindered ai progress by underestimating the potential of scaling laws, calling bets against llms misguided. he argued that llms have surpassed human intelligence in specific domains, citing an instance where an openai model solved a mathematical conjecture that stumped humans for years, forcing mathematicians to reconsider the field's future. altman acknowledged that while llms still lack human judgment on long-horizon tasks, the data clearly supports continued scaling over alternative architectures like world models.