Researcher Debate: Human vs AI OOD Generalization and Safety Focus

JacquesThibs · x · 2026-08-11

JacquesThibs joked about the common argument that "most humans are worse and don't generalize out-of-distribution (OOD) either."

In the replied context, he elaborated on the debate: while the logic of "people are not ready if it works" combined with "short timelines make assuming it works the best EV bet" makes sense, failing to understand the specific machinery of why models currently fail—and instead relying on vague scaling—might misdirect focus toward the wrong safety work.

Related event: Debate Sparks Over LLM Generalization and Scaling Laws(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →