Build AIs so their architectures are, by design, non‑conscious (a program Hoel calls 'qualiaphobia') and develop verifiable tests that show they lack consciousness. The argument is that preventing machine consciousness is a civilization‑level safety and moral policy that avoids creating beings we must then ethically protect or exploit.
— If adopted, it would redirect AI safety and regulation toward architecting and auditing non‑conscious systems rather than managing AI 'welfare', changing corporate design, legal standards, and public expectations.
Erik Hoel
2026.11.18
100% relevant
Erik Hoel's Nov 18, 2026 essay proposing 'qualiaphobia' — engineering architectures hostile to theories of consciousness and insisting on verifiability — exemplifies the idea.
← Back to all ideas