Keller Jordan: pausing algorithms while compute piles up maximizes AI risk

kellerjordan0 · x · 2026-09-09

ML researcher Keller Jordan argues the way to maximize AI risk would be to pause lab algorithmic progress while hardware compute keeps exponentially accumulating, paired with heavy safety focus that avoids spooking the public. Fewer accidents for five years would relax public and government scrutiny — until progress restarts and the stockpiled compute enables an uncontrollable FOOM. Minimizing risk, he says, means doing roughly the opposite.

Related event: Keller Jordan's Provocative Take: Pausing Algorithmic Progress Could Maximize AI Risk(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →