Gary Marcus: Both 'humans extinct in 5 years' and 'everything will be fine' are wrong
GaryMarcus · x · 2026-09-14
Gary Marcus argues the two dominant AI narratives are both preposterous: 'humans will be extinct in five years' and 'everything's going to be fine.' His boring reality: poorly controlled agents with unrestricted internet access will cause serious harm. Panic doesn't help, and neither does pretending nothing is wrong.
More from AGI Musings
- There's no truly AI-proof degree, say careers experts — adaptability beats picking a 'safe' major — nordicinst · 2026-09-14
- Screenwriter Jack Thorne urges law banning secret AI-generated scripts, says peers 'cheat' — nordicinst · 2026-09-14
- An Open Letter to Altman and Amodei: Safety Essays Aren't Safety Leadership — AryHHAry · 2026-09-14
- AI researcher lays out the safety paradox: pausing algorithms while compute piles up maximizes risk — RichmanRonald · 2026-09-14
- Reddit proposal: a neutral 'AI Switchboard' — public rails, competitive models, no lab moat — NewYak4281 · 2026-09-14
- AI bio-risk debate: if AI can design a super-virus, another AI can design the vaccine — RichmanRonald · 2026-09-14