Aaron Roth: complexity-based AI impossibility arguments now ignore reasoning models

Aaroth · x · 2026-09-14

Concluding his thread, Aaron Roth argues that a decade ago complexity-based impossibility arguments about general intelligence could at least be framed as denying computationalism, but making them today requires ignoring the obvious success of AI and particularly recent reasoning models. Elegant worst-case complexity models, he says, are poor predictors of what is possible or easy on naturally occurring problems.

Related event: Aaron Roth Slams Complexity-Based Arguments Against AI(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →