My article “On the Very Idea of a (Synthetic) Conceptual Scheme” has just been published in the open access journal Philosophy & Digitality, in a special issue on “LLMs and the Patterns of Human Language Use.”
The title of the piece plays on, and the article draws substantially on, Donald Davidson’s “On the Very Idea of a Conceptual Scheme.” By way of this classic text, I engage closely with M. Beatrice Fazi’s provocative article “The Computational Search for Unity: Synthesis in Generative AI.” I agree with Fazi that we have to take the outputs of LLMs as genuine language (contra the “stochastic parrots” crew), and that the best way to account for their operations is in terms of a kind of philosophical “synthesis.” But whereas Fazi sees LLMs synthesizing their own individual “worlds within,” I argue that the genuineness of their linguistic outputs (i.e. the fact that they produce real language) instead suggests that they refer to a world shared in common with human language-users (which commonality should not, however, detract from their alterity or alienness to our embodied Lebensform, or form of life).
In the same issue of Philosophy & Digitality, Fazi has a response to my article, titled “A Transcendental Philosophy of Large Language Models,” which I also highly recommend, and which brings our differences—as well as agreements—into sharper relief. I have the feeling this is the beginning of a longer exchange!
I’d like to thank Sybille Krämer and Christoph Durt for inviting my participation in the special issue and shepherding it toward publication–and for soliciting Fazi’s response. And thanks, above all, to Beatrice Fazi for producing such thought-provoking work in the philosophy of AI and computation!
