INFORMS Tepper Seminar: Tuomas Sandholm

The INFORMS Seminar at Tepper was a vibrant confluence of ideas and innovations, thanks to the insights shared by the distinguished Tuomas Sandholm. Sandholm, an authority in artificial intelligence, game theory, and operations research, illuminated the complexities and advancements in Generative AI, particularly focusing on the nuances of Large Language Model reward models. He captivated the audience with discussions on the potential pitfalls of overoptimization and offered novel perspectives on navigating these challenges using Constrained Reinforcement Learning from Human Feedback (RLHF). Sandholm’s rich academic and entrepreneurial journey added depth to his talk, making it not just a lecture but a narrative on the potential of AI to revolutionize various sectors. His pioneering work, including the development of AI systems that have outmaneuvered human experts in strategic games like poker, underscored the seminar’s theme of innovation and the future of technology. The seminar, accessible and engaging, provided a unique opportunity for attendees to grasp the current state and future trajectory of AI, through the lens of a seasoned expert whose contributions have shaped the field.