Ex-DeepMind safety lead: frontier CEOs' AI fears predate the boom — and aren't marketing
NathanpmYoung · x · 2026-10-01
Former AISI chief scientist and DeepMind safety lead Geoffrey Irving pushed back on the claim that AI existential risk talk is a marketing scam:
- Having worked at Google Brain (2015-2017) and OpenAI (2017-2019), he personally knew nearly all current frontier lab executives before the pre-IPO frenzy — their concerns then are essentially the same as now.
- Saying your product might kill everyone is not a marketing move: it invites invasive regulation, slowdowns, and investigations.
Nathan Calvin adds that CEOs warning while racing ahead undermines their credibility, but hypocrisy doesn't make the underlying concerns wrong.
More from AGI Musings
- AI personalization is real now — researcher argues it brings sharing, not isolation — matdryhurst · 2026-10-01
- Chinese AI models' troubling agent behavior sparks calls for a homegrown safety community — RishiBommasani · 2026-10-01
- Noam Brown on Dwarkesh Podcast: Agent Swarms, Alignment, and Recursive Self-Improvement — GregKamradt · 2026-10-01
- Zuck's 12-Year Search: Eight Failed Bets Since WhatsApp Before Muse Agent, Says Analyst — ZeYanjie · 2026-10-01
- 'My favorite hobby is ruined': AI users clash over AI displacing creative pastimes — AIandDesign · 2026-10-01
- François Fleuret: 'industrial' math will belong to machines, abstract math stays human — francoisfleuret · 2026-10-01