9:15 - 9:30 am
Introduction
by Organizers [slides]
9:30 - 10:00 am
The promise and pitfalls of generative AI
by Kathleen Fisher (DARPA) [slides]
10:00 - 10:30 am
Can AI-Generated text be reliably detected?
by Soheil Feizi (U of Maryland) [slides]
10:30 - 11:00 am
Watermarking Image Generator Models
by Florian Kerschbaum (U of Waterloo) [slides]
11:00 - 11:30 am
Robustness and Security Challenges in Large Language Models
by Tatsu Hashimoto (Stanford) [slides]
11:30 - 12:00 pm
An overview of catastrophic risks from generative AI
by Dan Hendrycks (safe.ai) [slides]
12:00 - 1:15 pm Lunch
1:15 - 1:45 pm
Exploiting programmatic behavior of LLMs for dual-use
by Daniel Kang (UIUC) [slides]
1:45 - 2:15 pm
Baseless speculation on the future of automating attacks with generative models
by Nicholas Carlini (Google)
2:15 - 2:45 pm
Panel Question: What attacks become easy to mount due to GenAI technologies?
Panel Lead: Anupam Datta (Truera)
Panelists: Brad Chen (Google), Khawaja Shams (Google), Zuilfikar Ramzan (Aura Labs), Bradley Boyd (Stanford)
2:45 - 3:00 pm Break
3:00 - 3:30 pm
Challenges and progress towards socially responsible GenAI
by Diyi Yang (Stanford) [slides]
3:30 - 4:00 pm
Grounding for LLMs
by Ankur Taly (Google)
4:00 - 4:45 pm
Panel Question: What are some current and emerging technologies we should pay attention to for designing countermeasures?
Panel Lead: John Mitchell (Stanford)
Panelists: Eric Mitchell (Stanford), Elie Bursztein (Google), Dawn Song (UC Berkeley), Clark Barrett (Stanford)