Presenter
Rick Korzewka, AI Impacts
Rick is the Director at AI Impacts, where he started as a researcher in 2019. He splits his time between running the organization and doing research. His current research interests are analyzing AI progress in the context of human performance, learning from the past about how we can confront big problems that require foresight, and figuring out how technological progress works, especially around very large advances. Areas where he has more enthusiasm than expertise include game theory, ethics, and AI alignment. Outside of work, he enjoys competitive cycling, tinkering, and tabletop gaming. He holds a PhD in physics from the University of Texas at Austin.
Summary:
Rick Korzewka discusses technical projects related to AI development and cooperation. He emphasizes the importance of addressing the question of AI and cooperation, referencing the recent request for comment from the White House OSTP. He highlights the need for seeking marginal gains in technical projects for AI cooperation, such as seismic signal analysis and predicting AI capabilities. Finally, he also addresses the challenge of distinguishing between dangerous and safe AI systems and the value of capability predictions in AI policy concerns.
