We are on the verge of deploying Exascale systems, which remarkably in the US are quite similar in architecture and programming models. But what comes after these initial exascale systems? And what does that mean for scientific discovery and the languages, tools, and frameworks we need to build future applications? Are we going to see AI accelerators break out into the mainstream? Are hybrid-HPC/AI surrogate models the future of scientific computing? And if so what is the architecture and software implication? Is a hardware disaggregation model viable for HPC applications? Heterogenous workflows connecting the edge to the core are becoming more important yet our software stacks are unprepared and do not treat workflows as first-class applications. Increasing numbers of applications in science appear to require combining data analysis, simulations, and machine learning. What is the environment we need for building these new classes of applications? In this talk, I’ll try to make sense of trends and future direction and outline some important research problems for the next decade.

Rick Stevens is the Associate Laboratory Director of the Computing, Environment and Life Sciences Directorate at Argonne National Laboratory, and a Professor of Computer Science at the University of Chicago, with significant responsibility in delivering on the U.S. national initiative for Exascale computing and developing the DOE initiative in Artificial Intelligence (AI) for Science.

His research spans the computational and computer sciences from high-performance computing architecture to the building of tools and methods for bioinformatics, cancer, infectious disease, and other problems in science and engineering. A recent focus has been the development of AI methods for a variety of scientific and biomedical problems. He also specializes in collaborative visualization technology and grid computing. At Argonne, he leads the Laboratory’s AI for Science initiative and is currently focusing on high-performance computing systems which includes collaborating with Intel and Cray to launch Argonne’s first exascale computer, Aurora, as well as the National AI Accelerator Testbed which brings together leading AI scientists to provide an open and unbiased environment for the evaluation of emerging AI accelerator technologies designed to accelerate training and inference for deep learning models.

Prof. Stevens is a member of the American Association for the Advancement of Science and has received many national honors for his research, including an R&D 100 award and most recently being named a Fellow of the Association of Computer Machinery (ACM) for his continuing contributions to high-performance computing.