Don’t be fooled—LLMs don’t reason

Summary

The author, who helped build AlphaGo, argues that today's large language models do not reason. He contrasts AlphaGo's policy network, which regarded move 37 in its 2016 match against Lee Sedol as a roughly one in 10,000 play, with the explicit game-tree search that actually made the choice. Large language models, he argues, produce even their chain-of-thought steps through the same next-token prediction process, lacking a separate reasoning mechanism.

Related items

• InfoOpenAI Fires 3 Safety Researchers in Dispute Over AI RisksSame vendor · SecurityWeek
• MediumHackers abuse Google Ads, Bing redirects to push Claude ClickFix attacksSame vendor · BleepingComputer
• Info‘Pure insanity’: Mathematicians will need years to make sense of OpenAI’s latest dropSame vendor · The Verge (AI)
• InfoOpenAI reports three new incidents of misalignmentSame vendor · CSO Online
• InfoOpenAI's revenue scare, Delta earnings, what investors think of a Starbucks-Chipotle deal and more in Morning SquawkSame vendor · CNBC Technology