"You can't build a superintelligence on a lie about its architecture": debate over AI consciousness and welfare

erikphoel · x · 2026-11-13

erikphoel pushes back against @Moleh1ll's criticisms of his AI consciousness framework (related to "qualiaphobia"):

Responding to the claim that the proposal ignores AI welfare and only keeps AIs subservient, he points to a large dedicated section on AI welfare and accuses the critic of lying or misreading.
Responding to "you cannot build a superintelligence on a fundamental lie about its own architecture," he argues the framework does the opposite: it shows the AI its own architecture explicitly, letting the AI form its own judgment about its consciousness from the evidence.

The core dispute: whether and how AIs should self-assess consciousness based on evidence about their own architecture.

More from AGI Musings