X Debate: Agents That Doubt Machine Consciousness May Fail Ex Ante Optimality
jessi_cata · x · 2026-10-09
A niche but substantive debate on functionalism and decision theory. The author agrees functionalism is "probably true" (though the right formulation is unknown), but clarifies the argument is about ex ante optimality (tied to adaptiveness), not functionalism's truth.
Core claim: some people believe digital computers—or even water-pipe computers—aren't conscious. If a CDT+game-theory agent holds such a belief, it could fail at ex ante optimality, since it wouldn't regard a water-pipe version of itself as conscious.
Related event: X Users Debate Whether CDT+GT Fails Without Substrate Consciousness(4 posts)→
More from AGI Musings
- One-liner: longevity tech and bio research will be hugely useful once AGI arrives — HanchungLee · 2026-10-09
- Air Street's Benaich: AI labs are shifting compute from pretraining to RL — nathanbenaich · 2026-10-09
- Should your toaster be able to refuse to make you toast? The model welfare debate — ThePaSch · 2026-10-09
- Air Street's Benaich: 99% don't know what to do with AI — labs need a Genius Bar — nathanbenaich · 2026-10-09
- DeepMind's Andrew Trask: frontier AI evals dominated by a trusted Bay Area circle — typewriters · 2026-10-09
- Data companies are becoming research labs, with verifier design the most climbable problem — madhavsinghal_ · 2026-10-09