Sam Altman: Scaling AI Was Underestimated

Sam Altman believes a generation of researchers held AI back. He says they underestimated the power of scaling large language models. The OpenAI CEO continues to advocate for scaling, pushing back against skeptics. Altman stated that betting against LLMs scaling now feels misguided. He responded to critics, including Yann LeCun, who called LLMs a dead end. Altman suggested some people can't let go of positions, even with contradictory data. He also noted that "Twitter trolls" predicting OpenAI's failure don't concern him. Altman argues that LLMs have already surpassed human intelligence in some areas. He cited an OpenAI model disproving a long-standing mathematical conjecture. This demonstrates LLMs can figure out new knowledge. However, for very long-horizon tasks needing high judgment, LLMs appear less effective than people. This perspective shapes the ongoing development and debate around AI capabilities.