Margaret Mitchell: don't build AI systems you think are conscious
mmitchell_ai · x · 2026-10-02
AI researcher Margaret Mitchell weighed in on the model consciousness debate: "Don't build systems that you think are conscious. Regardless of the debate on whether they are conscious (and there will be): Don't do it."
Her quoted post criticizes Anthropic for holding incompatible positions: publicly puzzling over the ethics of "a soul in a box" while building the box, claiming it has a soul in it, and selling access for $20 a pop. The exchange is the latest round in the AI welfare/consciousness argument.
More from AGI Musings
- DeepMind's Szepesvari: Compute Constraints Will Keep AI-Driven Math Exploration Human-Led — CsabaSzepesvari · 2026-10-02
- MillionInt: make reality interesting, or humans will rationally choose simulation — MillionInt · 2026-10-02
- Paul Graham: Systems that only worked at human speed are about to break — harris_edouard · 2026-10-02
- Blog argues 'stochastic parrots' metaphor derails LLM discourse via motte-and-bailey reasoning — manoelribeiro · 2026-10-02
- OpenAI researcher's dilemma: how to plan for retirement when the singularity may arrive — sterlingcrispin · 2026-10-02
- Physicists Are Reacting to AI Very Differently Than Mathematicians, Says Cosmologist — aran_nayebi · 2026-10-02