Aaron Roth Slams Complexity-Based Arguments Against AI

UPenn theorist Aaron Roth argued that worst-case complexity arguments against the possibility of AI are bad, noting even linear classifier training is NP-hard yet ML thrives, and that the success of reasoning models makes such arguments untenable today.

2026-09-14 ~ 2026-09-14 · 4 related posts

• Aaron Roth: worst-case complexity is a bad argument against building real AI — Aaroth · 2026-09-14
• Even training a linear classifier is NP-hard to approximate, notes Aaron Roth — Aaroth · 2026-09-14
• Reasoning models' success makes denying computationalism ever harder — Aaroth · 2026-09-14
• Aaron Roth: complexity-based AI impossibility arguments now ignore reasoning models — Aaroth · 2026-09-14