'Poetry' as the name of a special human relation to the world -- some special kind of knowing, grasping, challenging or asking we effect through art -- came into focus in 18th century Europe alongside the first blushings of a theory of computation and a computational analysis of mind. This talk proposes that for all of their outward hostilities, the Romantic-and-on idea of poetry and computational approaches to thought, language, and meaning are deeply connected: starting from Kant's doctrine of the productive imagination, we will develop one historical thread that runs to the Romantic poets, Phenomenology, and literary theory, and one historical thread that runs to information theory, machine learning, and the science of neural network models. Comparing the two threads, I'll argue that poetics and the science of neural network models have genuinely (if partially) overlapping subject-matter.

Peli Grietzer is a researcher and writer specializing in ML, philosophy, and literary studies. Grietzer received his PhD from Harvard Comparative Literature in collaboration with the HUJI Einstein Institute of Mathematics.