A simple anti-doomer argument: LLMs can't even do near-OOD tasks, let alone RSI

mariofilhoml · x · 2026-09-14

The author offers a brief counter to AI doomers: recursive self-improvement to the point of extinction is arguably a far out-of-distribution task, yet LLMs are already very limited even at near-OOD tasks—so how can doomers claim they will magically pull off something so far beyond their demonstrated capabilities?

Related event: Researchers Push Back on Dario Amodei's Claim That RSI Has Begun(8 posts)→

Original post →

More from AGI Musings

AGI Musings channel →