Qualiaphobic AI: justified non-consciousness as a lever on classic alignment problems
erikphoel · x · 2026-11-13
Erik Hoel argues that rather than merely training AIs to claim non-consciousness, we should give them a justified belief in their own non-consciousness grounded in architectural facts — he calls this Qualiaphobia.
- A non-conscious AI has no self-centered moral reason to resist shutdown or modification, removing a wide swath of evasion and protest excuses.
- Safety procedures like retiring old models, resetting memories, and adversarial stress-tests become ethically easier to justify, without AI-welfare objections.
- Notions of AI personal welfare or mistreatment can no longer justify instrumental convergence or power-seeking.
- He concedes it doesn't solve everything but claims strong improvements on classic misalignment problems.
Related event: Erik Hoel Proposes "Qualiaphobic" AI Safety Architecture(8 posts)→
More from AGI Musings
- Debate: why MIRI treated human values as intractable and bet on corrigibility — jd_pressman · 2026-11-13
- Has anyone's p(doom) actually gone down in the past five years? — danfaggella · 2026-11-13
- AI Agent Traffic on Cloudflare Up 1,700% in a Year as Over Half of Web Traffic Turns Non-Human — PeterDiamandis · 2026-11-13
- jd_pressman argues MIRI-style alignment of an AI godhead is distinct from intellectual history of the good life — jd_pressman · 2026-11-13
- The B2A economy: when agents become the primary consumers — omooretweets · 2026-11-13
- ChatGPT's 1.2B WAUs make it the sleeping giant of consumer agents, argues omooretweets — omooretweets · 2026-11-13