Critic doubts AI self-reports on consciousness: two prompts flip its logic
JD__Hayes · x · 2026-11-16
JDHayes questions why a frontier model (fable 5.2) would self-report only 20% probability of its own moral patienthood in welfare interviews, then claim the number "came from outside". His argument: only the mind itself can know its interiority, so the model should be certain either way — yet its reasoning collapses after just two prompts, and he sees no coherent case for any answer between 0 and 100.