EE Seminar: Undetectable Watermarks for Language Models

(The talk will be given in English)

Speaker:     Dr. Or Zamir

                             School of Computer Science, Tel Aviv University 

Monday, November 18^th, 2024

12:00 - 13:00

Undetectable Watermarks for Language Models

Abstract

Recent advances in the capabilities of large language models such as GPT-4 have spurred increasing concern about our ability to detect AI-generated text. Prior works have suggested methods of embedding watermarks in model outputs, by noticeably altering the output distribution. We ask: Is it possible to introduce a watermark without incurring any detectable change to the output distribution?

To this end we introduce a cryptographically-inspired notion of undetectable watermarks for language models. That is, watermarks can be detected only with the knowledge of a secret key; without the secret key, it is computationally intractable to distinguish watermarked outputs from those of the original model. In particular, it is impossible for a user to observe any degradation in the quality of the text. Crucially, watermarks should remain undetectable even when the user is allowed to adaptively query the model with arbitrarily chosen prompts. We construct undetectable watermarks based on the existence of one-way functions, a standard assumption in cryptography.

We will also cover subsequent generalizations to steganography, and robust versions.

 Short Bio

Or Zamir is an assistant professor at the school of computer science in Tel Aviv University. Prior to that, he did his postdoc at the Institute for Advanced Study and Princeton University. His research interests include algorithms, data structures, and theoretical computer science. 

השתתפות בסמינר תיתן קרדיט שמיעה = עפ"י רישום שם מלא + מספר ת.ז. בטופס הנוכחות שיועבר באולם במהלך הסמינר