Predicting or understanding: the AI dilemma, and its relevance to AI for good

The talk by Prof. Marco Zaffalon will provide an overview of AI and its recent successes, and discuss their relation to the question of whether AI should just predict or understand. Playing a bit with ChatGPT is an easy way to face the limits of prediction, at whose core lie correlation-based methods. So how do we get to understanding? Can we enhance AI with it? Can we use it to explain AI methods, to provide ethical AI approaches and to have AI reason more like us? Most likely we can, and this talk will discuss one possible way to reach such a goal. The discussion will eventually be exemplified with an application to palliative care.

? Mentimeter: Ask your questions