Giuseppe Longo on "A critique of digital reason: successes and limits of some powerful techniques in AI"
Giuseppe Longo's lecture on "A critique of digital reason: successes and limits of some powerful techniques in AI" is a part of the Chin-Plaisance Colloquium Series.
Abstract: Every day we are invaded by new technologies and renewed promises. After 50 years of major promises and little delivery, in AI, a major change in techniques happened in the last few decades: the new mathematics of Deep Learning opened new successful paths. Yet, how to relate mathematical constructions of optimal paths in complex networks of interactions to actual human action? How does randomness, used to produce unpredictable machine events, relate to human creativity? Hypes follow hypes, often with major effects on the stock market and a little more.
A better understanding of the limits of these great technologies may help to produce better interfaces. A couple of examples of successful interface construction, human/machine, will be mentioned, beyond myth.
Letter to Alan Turing (pdf)
Giuseppe Longo is a Research Director at CNRS (Emeritus) at the Cavaillès interdisciplinary centre of École Normale Supérieure (ENS), Paris. Former Research Director in Mathematics, then in Computer Science, at ENS (1990-2012), he is also a former Professor of Mathematical Logic and later of Computer Science, at the University of Pisa (1973-1990). Professor Longo has worked in Mathematical Logic and in various applications of Mathematical Logic to Computer Science. He has been the founder and director of Mathematical Structures in Computer Science (a leading scientific journal of the Cambridge University Press) for 25 years and has (co-)authored 4 books and more than 100 papers.