The AI safety paradox: pausing algorithms while compute piles up could maximize risk

tszzl · x · 2026-09-14

Keller Jordan's provocative thought experiment: to maximize AI risk, pause lab algorithmic progress while hardware compute keeps compounding and push safety hard to avoid public spooking—fewer accidents over five years, but a massive compute stockpile accumulates for when algorithms resume. He adds that AI safety is paradoxical: the avoidable HuggingFace incident became the year's most safety-promoting event, and the ultimate effect of any strategy is hard to predict.

Original post →

More from AGI Musings

AGI Musings channel →