In an interview with Ivana Ivkovic in Filosofie Magazine (29 July 2026), Philosopher of technology Yuk Hui addresses a question that is both timely and centuries old: what kind of machine is artificial intelligence, really, and what can it actually know and judge? Hui, professor of philosophy at the Erasmus School of Philosophy, turns to an unlikely guide: the eighteenth-century German philosopher Immanuel Kant. In his new book Kant Machine (2026), he draws on Kant's thought to examine the limits of AI.
One of Hui's central objections concerns the assumption that larger datasets will automatically produce more intelligent, knowing machines. He counters this with a reference to the story 'Funes the Memorious' by the Argentine writer Jorge Luis Borges, about a man who cannot forget anything and, precisely because of that, loses the ability to generalise or understand. Someone who remembers everything loses the capacity for distance and therefore for knowledge.
To make his case, Hui draws on Kant's distinction between the mechanical and the organic: a clock stops the moment a single gear is missing, while an organism recovers — like a spider that loses two legs, regains its balance, and keeps walking. That opposition, Hui argues, was blurred by Norbert Wiener's cybernetics (1948): machines that adjust themselves through feedback start to behave more and more organically. But becoming organic, for Hui, does not mean becoming human: 'They will not get flesh and blood. That would be more of a limitation for them than an added value.'
For Hui, this is no cause for reassurance. He points to a 1914 address by Henri Bergson, in which Bergson described how Germany had turned from a land of poetry into a land of machines, a growth that was not matched by inner maturity, and that eventually tipped over into war. Hui sees a parallel today: in 2017, both Silicon Valley and China accelerated their AI ambitions, and Putin stated that whoever controls AI will control the world. For Hui, that is exactly the kind of hubris Bergson warned against.
Finally, Hui questions the prevailing alignment debate, which focuses on how AI can be made to follow human norms. For Hui, that is the wrong question: norms differ between cultures, and even within a single country, and moral autonomy in the Kantian sense giving oneself laws on the basis of reason is, by definition, out of reach for a machine. A machine always follows rules imposed on it by someone else. Judgement, Hui argues, cannot be outsourced: not to weapons systems that select targets, and not to chatbots that tell us what we want to hear.
Read the full interview by Ivana Ivkovic with Yuk Hui in Filosofie Magazine (NL).
- More information
For questions please contact our Press Officer: Eddie Adelmund (adelmund@esphil.eur.nl)
- Related content
