Logic, Epistemology and Metaphysics Seminar | AI and the Metaphysics of Content

Abstract: This talk draws on the forthcoming book Making AI Intelligible (by Cappelen and Dever). I outline our metasemantic framework for attributing representational states to artificial intelligences (in particular to AI systems using certain kinds of neural networks). We show how externalist frameworks from Kripke, Evans and Milikan can be de- anthropomorphised and then applied to artificial structures. We call the process anthropocentric abstraction. The process helps us understand AI better, improves our metasemantic theories (and also our meta-metasemantic theories), and contributes to the explainable AI movement. 

This seminar will be run by video link, circulated by email.

Pre-registration will be required by midday 6th April. 

This seminar is supported by the ERC project MetCogCon (GA 681422)

The Centre for Logic and Language hosts a regular seminar series - the Logic, Epistemology and Metaphysics Forum (LEM). The forum generally meets fortnightly in term time.