Artificial Intelligence: Prediction, Understanding, and the Future of Science and Society
Over the past five years, artificial intelligence has undergone several dramatic transformations. No longer the domain of futurists and science fiction, AI-driven technologies, including large language models, sophisticated computer vision systems, and automated decision makers, have transformed many aspects of human life, from how astronomers view the cosmos to how police identify suspects. There is little doubt that AI has remarkable capabilities for pattern recognition, predictive analytics, and the generation of text, images, and sounds.
In the wake of these advances, AI experts, scientists, philosophers, and social commentators have made bold predictions about what else to expect from emerging AI technologies. For some, AI promises to revolutionize theoretical science and even mathematics, perhaps even extending our knowledge of the universe beyond human capabilities. For others, AI poses existential risks, whether because of the destructive potential of rogue systems or because an overreliance on decision-making that is opaque to human operators can have unanticipated and deeply harmful consequences. Others, still, worry less about existential risks, but emphasize the ways in which AI may harm society in the near-term, for instance by reinforcing biases implicit in their training sets or by effectively stealing ideas, text, and images from human creators.
This meeting will assemble an interdisciplinary group of leading computer scientists, cognitive scientists, statisticians, physicists, historians, and philosophers of science to discuss epistemic and ethical issues arising from the recent advances in AI. Key questions will include what role understanding plays in the physical sciences; whether AI-assisted science can achieve the goals of theoretical science as traditionally conceived; and whether the epistemic benefits of AI can be attained without taking on unacceptable risks and/or incurring other material harms.
Seven Pines Symposium XXVI Program
Session 1: Two Cultures of Statistics
Peter Norvig, "Two Cultures of Statistics"
Session 2: Prediction and Understanding
Emily Sullivan, "Machine Learning in science: Dimensions of understanding"
Session 3: Beyond Prediction: Language, Image, Meaning
Cameron Buckner, "Generative DNNs as models of imagination, creativity, and planning"
Fabian Offert, "This Is Your Brain on ImageNet: Embeddings as Trading Zone and Cultural Technique"
Session 4: Beyond Prediction in Physics
Paul Ginsparg, "Ask not what ML can do for Physics, ..."
Session 5: Rethinking Prediction
Kathleen Creel, "Fairness and Randomness in Predictive Systems"
Stephanie Dick, "Ethics and Epistemology in the History of AI"
Session 6: Predicting the Future of AI
Advisory Board
Stephanie Dick (Simon Fraser University)
Lee Gohlike (Outing Lodge), Founder
Peter Gilbertson (Anacostia)
Peter Galison (Harvard), Co-Chair
John Norton (Pittsburgh)
Philip Stamp (University of British Columbia)
Jos Uffink (Minnesota)
Bill Unruh (University of British Columbia)
Robert Wald (Chicago), Co-Chair
James Weatherall (UC Irvine), Co-Chair
Discussants
Corey Dethier (Minnesota)
Sam Fletcher (Minnesota)
Aaron Gluck-Thaler (Harvard)
Geoffrey Hellman (Minnesota)
Nada Mohamed (Minnesota)
Jesse Wolfson (UC Irvine)