CAIA Speaker Event: Steven Basart CAIS, Ex-Google

Who: Steven Basart (VIRTUAL), CAIS, Ex-Google
When: February 6, 4-5 pm PT
Where: Gates Annex B122
Zoom: https://rit.zoom.us/j/94333284116

​​Talk: The Mechanics of Autonomy: Measuring AI Utility and Labor Substitution
To understand the future of AI, we need to measure two things. We need to know how models make choices and what they can actually accomplish. This talk covers recent work from the Center for AI Safety on Utility Engineering and the Remote Labor Index. We will look at how LLMs act as if they have goal-oriented utility functions and use the RLI to track how effectively they can perform real-world digital tasks. Join us to discuss what this data tells us about the internal "thinking" of these models and their growing capability to handle economic work.

​​About the speaker: Steven Basart is an AI researcher and research engineering manager at the Center for AI Safety, where he works on evaluating and mitigating risks from advanced AI systems. He holds a PhD in computer science from the University of Chicago, with a focus on machine learning and AI safety. His work centers on building and evaluating real-world systems that make advanced AI safer and more reliable with work such as WMDP, MMLU and HarmBench.

​​Everyone is welcome: No specific technical background is required. Come learn and ask questions.
And yes, we will have pizza and boba.