Millière vs Mitchell: can the intentional stance make an AI bot a genuine believer?

raphaelmilliere · x · 2026-09-04

Raphael Millière debates Melanie Mitchell on whether Dennett's intentional stance applies to AI. He argues a bot that merely generates Homeric sentences won't fare better than ELIZA, but one that pursues persistent goals, switches means when they fail, and updates on new information could warrant belief/desire-like attributions — and with proper fine-tuning could embody a robust "Odysseus" persona with states about Ithaca (or a proxy). He adds that whether these are merely "quasi" beliefs is unresolved: he disagrees with Dennett that the stance is constitutive rather than evidential, but the bot could have them while the character Odysseus cannot.

Related event: Philosophers Debate Whether a Homer-Trained AI Could Have Beliefs and Desires(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →