Philosophy of Mind Seminar (Friday - Week 4, TT23)

A controversy currently rages over whether the most advanced AI’s, particularly large language models, can be said to understand the items which they classify and generate. In this talk I show that in order to better grasp what is at stake in this controversy, it is helpful to revisit a philosophical debate that took place at Davos back in 1929, between Martin Heidegger and Ernst Cassirer. Roughly speaking, Cassirer’s account of function-concepts (first presented in his 1910 book Substanzbegriff und Funktionsbegriff) lends support to advocates of AI understanding, whereas those denying AI understanding would do better to utilise the Heideggerian notion of humanity’s being-in-the-world than the referential account of meaning put forward by Bender and Koller (2020). I will discuss how Cassirer’s position draws inspiration from Ernst Mach’s anti-metaphysical account of scientific progress, and how Heidegger’s is based on his interpretation of Kant as laying the foundations for metaphysics.