Gary Marcus: Not Even Anthropic Has a Theory for Making AI Safe
GaryMarcus · x · 2026-09-20
Gary Marcus amplifies a critique arguing that even Anthropic has no plan for making AI safe in advance — there is no theory, outside evaluators don't change that, and nobody at these companies has workable ideas for capturing frontier agentic AI benefits while maintaining human oversight. Marcus's own comment: "They don't have a fucking clue. Not even a theory about how make their stuff safe."
It is a public argument over whether AI safety has any theoretical foundation, i.e. a debate about alignment and governance approaches.
More from AGI Musings
- Luiza Jarovsky: AI won't kill all humans by 2030, but could destabilize global systems — AlexTensor · 2026-09-20
- Investor: personal AI assistants have arrived, generative streaming takes off in 2027 — annbordetsky · 2026-09-20
- Every AI debate boils down to one question: who holds the bag for this bubble? — AlexTensor · 2026-09-20
- Causal AI researcher calls out Hinton for swinging from blind optimism to existential panic — AlexTensor · 2026-09-20
- Ben Bajarin: Agentic AI will spawn an 'agentic native' CPU tier in datacenters — BenBajarin · 2026-09-20
- Jaron Lanier: AI consciousness is a question of faith, not science — and that taboo is a problem — iamtrask · 2026-09-20