I will describe an emerging “AI for Science” initiative at Argonne National Laboratory to advance the concept of Artificial Intelligence (AI) aimed at addressing challenging problems in science. We aim in this initiative to: (1) identify those scientific problems where existing AI and machine learning methods can have an immediate impact (and organize teams and efforts to realize that impact); (2) identify areas of where new AI methods are needed to meet the unique needs of science research (frame problems, develop test cases, outline work needed to make progress); and (3) develop the means to automate scientific experiments, observations, and data generation to accelerate the overall scientific enterprise. Science offers plenty of hard problems to motivate and drive AI research, from complex multimodal data analysis to integration of symbolic and data intensive methods, to coupling large-scale simulation and machine learning to drive improved training to control and accelerate simulations. A major sub-theme is that of working toward the automation of scientific discovery through integration of machine learning (active learning and reinforcement learning) with simulation and automated high-throughput experimental laboratories. I will provide some examples of projects underway and lay out a set of long-term driver problems.

This event will take place at Fermilab, and will be simulcast live to the Kathleen A. Zar Room in John Crerar Library.