Arvind Narayanan: Anthropic's Chart Doesn't Imply Intelligence Explosion — Delegation Isn't Automation

random_walker · x · 2026-09-22

Responding to speculative takes on Anthropic's capability chart, Princeton professor Arvind Narayanan argues it doesn't imply we're near intelligence explosion. His core chain: task delegation ≠ task automation ≠ process automation ≠ faster progress ≠ recursive self-improvement ≠ intelligence explosion. He cites the software engineering precedent — a year ago, widespread hopes/fears that AI writing 100% of code would explode SaaS output — as evidence that bottlenecks are common even when capability growth is fast, with a huge gap between benchmark curves and actual productivity. Afinetheorem and others endorse the argument.

Related event: Princeton Professor Pushes Back on 'Intelligence Explosion' Hype(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →