2026 workshop program
Measuring and forecasting AI scientific capabilities
The work
Abstract
TBA
The speaker
Biography
Siddharth’s research explores how AI and science can advance each other: using AI to tackle problems in the physical sciences, and using scientific domains to understand the contours of AI capabilities.
Siddharth is interested in how increasingly capable AI systems can participate in the scientific process, improving “traditional” tasks such as simulation, modeling, and inference, while also enabling qualitatively new modes of scientific investigation. This includes using AI to generate and test models, developing scalable inference techniques, and probing the limits of automated scientific reasoning. He views physics —- which has been the focus of much of his work —- as a particularly useful domain for this, offering a well-defined sandbox spanning theory, computation, and experiment for developing and testing methods with broader applicability across the natural sciences.
He is also interested in meta-scientific questions about how technological progress shapes what research is possible, how increasingly powerful AI capabilities are reshaping scientific practice, and the evolving relationship between academic and industry research.