Scientific discovery

Armin Tavakoli discusses the growing ability of commercial AI models to contribute to scientific discovery with limited human involvement. The talk surveys current capabilities and examines their consequences for academia, organizing the discussion around grants and recruitment, scientific practice itself, and university education. It proposes possible responses and invites a wider discussion of how institutions and researchers should address these challenges. Hybrid seminar in the Rydberg Lecture Hall, Department of Physics, Professorsgatan 1B, Lund, Sweden, with public Zoom attendance. Coff

Supports agentic AI research and open-source tools, including multi-agent collaboration, scientific discovery, autonomous trajectory verification, and customization through post-training or optimization of prompts, tools and agent scaffolding. Research may address safe agent behavior, data integration, enterprise operation and human–AI collaboration. Submit a scientific proposal using the ARA template, encouraged to be no more than four pages excluding appendices. Decisions are expected in February 2027. Applications close on 4 November 2026 at 11:59pm Pacific Time.

…between learning systems and scientific discovery.

…explains his perspective on safety, scientific discovery and the longer-term role of AI in understanding ourselves.

…Wilczek about the role of beauty in scientific discovery. The episode connects empirical questions about the human response to beauty with the use of aesthetic intuitions in developing physical theories.

…of using deep learning as a method for scientific discovery and latent associations.

…severely limited, holding back both scientific discovery and policy interventions. This talk discusses how developmental, cognitive and neuroscientific approaches might provide a new and improved way of studying social media effects. It will detail new studies in support of this