OpenAI CEO Sam Altman says AI is far from matching human performance, ‘but in terms of…’

OpenAI CEO Sam Altman says that artificial intelligence (AI), even in its most advanced form today, remains far from matching human performance when it comes to sustained, long-horizon problem-solving. The tech CEO recently appeared in the podcast hosted by Zerodha co-founder Nikhil Kamath, Altman acknowledged that systems like GPT-5 are “incredibly smart in a lot of domains” and “very superhuman at knowledge and pattern recognition, recall on these shorter-term tasks.” However, he stressed that the technology is still limited. “In terms of figuring out what questions to ask or to work on something over a very long period of time, we are definitely not close to human performance,” Sam Altman said.He illustrated this with examples from AI research. “A couple of years ago, we could solve math problems that would take, like, an expert human to solve. Recently, we got gold level performance (hinting at OpenAI's ChatGPT o3 model win over Elon Musk's xAI model Grok 4 in the final of a Kaggle-hosted tournament) on the international stage… each of those problems takes about an hour and a half. So we’ve gone from a thinking horizon of a few minutes to an hour and a half. To prove a new and important new mathematical theorem may take like a thousand hours… in the world today, we cannot at all.”While acknowledging the excitement around artificial general intelligence (AGI), Altman’s remarks underline the gap between today’s AI and the human mind’s ability to work on problems requiring months or years of sustained reasoning — a challenge he believes is still far from solved.“I think one of the things that is gonna feel most AGI-like is seeing robots just watch,” Altman said, adding that robots can “certainly have non-humanoid forms, but the world is really built for humans”. During the podcast, Sam Altman said that OpenAI is “interested in robots” and is a “new skill” for the company to learn.