Debate: can an AI be exactly as capable yet obviously non-conscious?
brrrrrrennan · x · 2026-11-13
In a thread on "qualiaphobic" AI architectures for safety, brrrrrrennan argues there's no free lunch where an AI is equally capable yet suddenly obviously not conscious — implementing a lookup table as fast as weights is unclear, and a thought experiment would work just as well. erikphoel counters that all factors play a role in misaligned behavior and knowing your goals aren't "real" helps with issues like reward hacking.
Related event: Hoel's "Qualiaphobia" AI Safety Proposal Sparks Multi-Round Debate(10 posts)→