Sampura Research [Talk+Social - Berkeley]
- WHO IT IS FOR: AI safety researchers, alignment engineers, and technical professionals interested in the intersection of human oversight and machine learning.
- WHAT IT IS ABOUT: Exploring how human-AI complementarity and advanced judging methods can improve the alignment of increasingly capable AI systems.
- FORMAT: A technical presentation followed by a Q&A session and a networking social.
Join Sampura Research in Berkeley for an insightful technical talk on the future of AI alignment. As AI capabilities expand, traditional evaluation methods are reaching their limits. This session explores why human oversight remains a critical, uncorrelated layer of defense against the shortcomings of LLM-only judges and how to build more robust evaluation frameworks.
Discover how Sampura Research is developing a leaderboard of over 50 alignment-relevant datasets and implementing novel methods to route and assist across human and AI intelligence. This event is an excellent opportunity to learn about the technical approaches being used to hill-climb AI safety through human-AI complementarity.
Following the presentation and Q&A, stay for a social mixer to connect with the Sampura team and other local researchers. Founded by Google DeepMind alumni, Sampura Research is a nonprofit dedicated to solving alignment through cutting-edge technical innovation.