Why aren't more lab researchers leaving for METR? An AI-safety talent debate

a__tomala · x · 2026-09-08

kipperrii argues it 'feels wrong' that more people from frontier labs haven't left to join METR or similar AI-safety evaluation efforts — the optimal headcount may be low, but likely higher than actual movement. atomala follows up asking where people focused on architecture/infra (rather than data/benchmarks) could work outside the major labs. The thread highlights the ongoing talent-flow debate between frontier labs and AI-safety orgs.

Original post →

More from AGI Musings

AGI Musings channel →