About the event
Title: "Engineering Sentience"
Abstract: There is an ongoing debate about whether an artificially intelligent agent can be sentient. Here I will spell out a definition of sentience that is useful for the purpose of designing and building such agents. I propose that for sentience to occur, the relevant sensory signals need to be both self-asserting and subjectively qualitative. These two properties can in turn be fleshed out in purely functional terms, meaning that their computational implementation is possible. We discuss the implications of this analysis, including what it means for animal sentience. I speculate that the functional origin for sentience is related to the need for maintaining the distinction between perception and reality.
If you would like to attend this event virtually by Zoom, please contact Drew Johnson at Department of Philosophy, Classics, History of Art and Ideas.