Accelerating Research and Engineering with AI | Theoretically Speaking

Large language models are now increasingly used to accelerate many research and engineering activities, thanks to their broad world knowledge and ability to work with arbitrary data. They have had great success in programming and math, and are being applied to biology, chemistry, and many other areas. But what are the strengths and limitations of AI for these tasks, and how can we make it even better? Matei Zaharia will describe research from UC Berkeley and other groups on using AI for engineering and research, including algorithms that use AI to search for better solutions to a problem (e.g., GEPA, AlphaEvolve, and AdaEvolve), and results researchers have been getting with them in computer science, biology, and other areas. In general, AI can be trained and applied effectively when solutions are easy to verify, but it has a much harder time when they aren’t (e.g., in chemistry and biology, where experiments take days). We are also getting better, however, at translating more problems into the easy-to-verify setting. In this Theoretically Speaking public lecture, Zaharia will discuss where this might lead, based on the trends we are seeing so far, and how science and engineering might look in the future when we have highly effective, tireless AI coworkers to explore solutions for us.

Light refreshments will be provided before the talk, starting at 4 p.m.

Theoretically Speaking is a lecture series highlighting exciting advances in theoretical computer science for a broad general audience. Events are free and open to the public, with first-come, first-served seating. No special background is assumed. Registration is required. This lecture will be viewable afterward on this page and on our YouTube channel, following captioning.

The Simons Institute regularly captures photos and video of activity around the Institute for use in publications and promotional materials. 

If you require special accommodation, please contact our access coordinator at simonsevents@berkeley.edu with as much advance notice as possible.