Mapping AI safety: technical solvability vs institutional guardrails, one researcher's two-axis chart
joshua_saxe · x · 2026-10-12
AI safety researcher Joshua Saxe proposes a two-axis framework for the field's disagreements:
- Technical axis: can we make AI systems safe at all?
- Institutional axis: will our institutions actually prevent unsafe deployments?
Placing the camps:
- Jensen Huang, LeCun, Ng, Andreessen: optimistic on both — safety is a normal engineering problem, existing legal regimes suffice.
- Yudkowsky: pessimistic on both — alignment is a generationally hard science problem and institutions are ill-equipped.
- Safety research scientists: race to solve axis 1, assuming institutions will fail axis 2.
- Altman / Amodei: pessimistic on axis 2, which justifies racing to build the AI that solves axis 1; OpenAI Foundation and Coefficient Giving act as stand-ins for the institutions we should have.
He includes a quadrant chart plotting Ng, Hinton, Bengio, Sutskever, Musk and others across "WE'VE GOT THIS / WE'LL RALLY / WE'LL BLOW IT / WE'RE IN TROUBLE", tracking how views shifted over the years. The author places himself near the center.
More from AGI Musings
- "Results Without Understanding" Is How LLMs Work—and Science Itself Is Changing — fkasummer · 2026-10-12
- Open-Source Dev giffmana: AI Extinction Talk Is 'Complete Nonsense Sci-Fi Fan Art' — giffmana · 2026-10-12
- Programming languages were just scaffolding for thinking — for LLMs and humans alike — MarcJSchmidt · 2026-10-12
- dioscuri publishes 'Bad AI Consciousness Takes Bingo' rebutting common LLM consciousness arguments — dioscuri · 2026-10-12
- Karpathy's drill analogy distilled: 5,000 people already got entire factories — _AustinCalvert_ · 2026-10-12
- Szegedy cites his May post: Terry Tao and Judit Polgar misjudge AI's trajectory — ChrSzegedy · 2026-10-12