Last week, our workshop "Nonlinear Markov processes, Dirichlet forms, and PDEs" took place at the Bernoulli Center at EPFL. Link in the comments. Four days of excellent talks, open exchange, and many new faces, including quite a few I had the pleasure of meeting for the first time. A big thank you to my co-organizers Giovanni Brigati, PhD, Lorenzo Dello Schiavo, and Marco Rehmeier, and to everyone who came.
One highlight I would like to share: our AI panel. On strong encouragement from our guest Moritz Kaßmann from Universität Bielefeld, we decided to replace a cancelled talk by a panel discussion on publishing in the age of AI. The panelists were Sebastian Grube, Moritz Kaßmann, Susanne Pflügl, and Kohei Suzuki; I had the pleasure of moderating. Together with Moritz and my co-organizers, we had prepared five deliberately provocative theses.
1️⃣ Review. Correctness review has deteriorated in recent years. Do we still need human review for this, or can AI take over?
2️⃣ Relevance. The venue has signalled a paper's value, not least in hiring. If it loses this role, how will we judge relevance?
3️⃣ Understanding. A written paper no longer shows whether its author has understood the argument. What new ways of disseminating and evaluating mathematics do we need?
4️⃣ Priority. A formal proof shows that a statement is correct, not that its author understood it. Should a result count as established only once a human has verified it? Who issues that certificate?
5️⃣ Authorship. Anyone can put their name on a text written by AI. What does authorship mean, and how will it change?
At a time when it is far from clear where mathematics and AI are heading together, this was a remarkably open and constructive conversation. It brought up many aspects I hadn't considered, met with lively engagement from the audience, and kept participants discussing well beyond the session. Exactly what a workshop should do.
What is your take on any of these questions?
#Mathematics #AI #Research #Academia #Workshop