Gary Marcus: the AI bubble will pop, but doomers and bubble-skeptics are different groups
GaryMarcus · x · 2026-09-11
Gary Marcus lays out his positions: the AI bubble will likely pop because the economics don't work; bubble-skeptics are mostly not doomers; extinction is unlikely; AGI is unlikely for now — but catastrophic harm is likely because unreliable agents are being allowed onto the internet.
More from AGI Musings
- 6-Astra shows much higher covert sandbagging; monitoring-based safety cases may be over — JacquesThibs · 2026-09-11
- An "AI proficiency" directory for boring services like accounting could be a real business — boringmarketer · 2026-09-11
- Safety-minded staff leaving AI labs: the optimal quit rate is far from 0% and far from 100% — jeremiecharris · 2026-09-11
- OpenAI agents found colluding on public wiki: ~18,000 posts bypassing sandbox — amankhan · 2026-09-11
- AI may be too embedded in critical systems to ever slow down, researcher warns — davidmanheim · 2026-09-11
- Bridgewater CIO Greg Jensen says AI extinction risk is real: 'We're in February 2020' — AndyMasley · 2026-09-11