Stanford's Noah Goodman Shares First-Pass AI Safety Concept Map for Lecture
noahdgoodman · x · 2026-10-19
Stanford professor Noah Goodman is preparing an AI safety lecture for his Minds and Machines course and published a first-pass conceptual map, soliciting feedback from safety researchers like Rohin Shah, Allan Dafoe, and Owl Evans. The map organizes AI risk as Source → How It Arises → How It Escalates → Outcome: from bad actors, honest error, structural incentives, and overreliance; through capability uplift, mislearned goals, race dynamics, and deference; escalating via diffusion, power-seeking, deception, delegation, and epistemic decay; toward outcomes ranging from utopia to catastrophe, loss of control, institutional upheaval, and disempowerment.
More from AGI Musings
- AI debate: offloading thinking or instilling your own thinking into the model? — eschadiol · 2026-10-19
- AI agents cost 20x more than conventional software, warns architect's enterprise breakdown — DavidLinthicum · 2026-10-19
- Amazon's 6-page memo culture 'nuked' by AI: employees now tag memos 'AI;DR' — emax · 2026-10-19
- Mathematician pushes back on Zhang critique: math is pure meritocracy, and AI may end that — AlexKontorovich · 2026-10-19
- Markets as 'the most dangerous game': everyone simulating everyone's predictions — willcb · 2026-10-19
- Economist Daniel Rock: AI circles' intelligence worship is a stumbling block for scientific progress — danielrock · 2026-10-19