Intelligence Relies on Heuristic Guessing

geoffreyirving · x · 2026-07-08

The author argues that simply saying "don't rationalize" is not feasible, because fluid intelligence in both humans and machines is built on heuristics: thinking consists of explicit reasoning and random guessing, followed by post-hoc filtering.

He views this mechanism as the common working mode of humans and current LLMs.

Related event: Geoffrey Irving: AI Safety Must Solve Post-Hoc Rationalization(8 posts)→

Original post →

More from AGI Musings

AGI Musings channel →