#01 | AI & Emotions

Moderator

TBA

Matthew McKeever (HKU): AIs Don’t Have Emotions — We Should Attribute Emotions to AI

Abstract. I present a conceptual engineering argument suggesting that we should introduce concepts for AI emotions. My argument proceeds in two steps. First, I present reasons for thinking that current-gen LLMs don't have emotions. Second, I argue that despite this, we should deploy concepts standing for AI emotions. The reason for this pertains to our moral character: if we habitually treat human-seeming entities (as LLMs are) as without emotions, there's a risk we'll carry that behaviour over to actual humans, and that will harm our relations to them and thus our moral character. Conceptual revision will help avoid that.

Simon Goldstein (HKU) — AI Pain

Abstract. I will examine whether current or near future AI systems could experience pleasure and pain. I will begin by surveying desire satisfaction and phenomenal theories of pleasure and pain. I'll argue that desire satisfaction theories make it relatively easy for current or near AI systems to feel pain, while phenomenal theories may make it more difficult. Then I'll focus more specifically on a few different functional conditions for bodily pain. I'll argue that our concept of pain is potentially indeterminate regarding which functional condition exactly is required for bodily pain, and this creates corresponding indeterminacy about AI bodily pain.

Zoom meeting ID: 679 2010 7239

Zoom passcode: CEET-W24