Community Event | Café Scientifique

Community Event - Café Scientifique

Door opens at 18h30 - Last drinks round at 18h45 - the venue doesn't serve food

Invited Speaker - Prof. Florian Tramèr

Modern generative models seem to generalize extremely well! At the same time, we find that they are capable of memorizing large amounts of training data. In his talk, Prof. Tramèr will describe some of the challenges in defining and measuring memorization in machine learning, and showcase substantial data memorization in state-of-the-art models for language generation, including ChatGPT! He'll also discuss some of the implications that such large memorization capabilities have for data privacy, copyright protection, and more, and how we might mitigate these risks.

Florian Tramèr is an assistant professor of computer science at ETH Zurich.
His research interests lie in Computer Security, Cryptography and Machine Learning security. In his current work, he studies the worst-case behavior of Deep Learning systems from an adversarial perspective, to understand and mitigate long-term threats to the safety and privacy of users.

‍