SN 1029: The Illusion of Thinking - Meta Apps and JavaScript Collusion
Security Now (Video)
00:00
Limitations of Large Reasoning Models
This chapter explores the challenges faced by large reasoning models (LRMs) in executing exact computations, especially through complex puzzles like the Towers of Hanoi. It highlights the significant performance disparities among non-thinking and thinking models across varying complexity levels, while also questioning the validity of puzzle-based evaluations for AI. The conversation underscores the need for a clearer understanding of AI's capabilities, cautioning against anthropomorphism and emphasizing ongoing research into model reasoning.
Transcript
Play full episode
