By Rob Mitchum // June 13, 2018
When the Computation Institute (CI) was founded in 1999, data were still transferred on CD-Rs, the internet was mostly accessed via dial-up modems, and phones were used solely for phone calls. Nearly two decades later, the world is immersed in data, computation, and connected technologies, with dramatic ramifications for science and society.
As the first joint initiative of the University of Chicago and Argonne National Laboratory, the CI kept both institutions at the crest of this computational wave with projects in physics, astronomy, social science, climatology, the digital humanities, and more. For Future Compute, a May event celebrating the CI before its closure this summer, the focus was on the next two decades of computation and its role in knowledge, policy, and research.
“The CI has been the driving force on campus and at Argonne in computational and data science,” said Eric D. Isaacs, Executive Vice President for Research, Innovation and National Laboratories at UChicago. “For 20 years, members of the CI have pursued distinctive research and scholarship, applying computer and data science problems across academic disciplines and to some of the most important issues in our society.”
The day-long gathering brought CI fellows, faculty and alumni together for talks and panels organized around forward-looking questions such as “How will the world be different once artificial intelligence is here?” Researchers from UChicago and Argonne who combine computer science with domains such as materials engineering, medicine, agriculture, and sociology discussed how tools such as machine learning and cloud computing now shape their field and, increasingly, the world beyond.
Throughout the day, speakers credited the Computation Institute with laying the foundation for this research at UChicago and Argonne. In looking back at its 19-year run, Rick Stevens and Ian Foster — the two longest-serving directors of the CI — stretched beyond the scientific discoveries and technical contributions to single out community-building as the institute’s most significant legacy.
“What made the CI work was the number of people at Argonne and Chicago that took on really challenging projects, outside the realm of what could be done in their own department,” said Foster, Arthur Holly Compton Distinguished Service Professor of Computer Science at UChicago and Senior Scientist at Argonne.
A Partnership of Humans and Machines
Many of the presentations focused on computational approaches that defined the term “AI” as augmented, not artificial, intelligence. In “Future Knowledge,” the opening panel, Knowledge Lab director James Evansfollowed the idea of human-machine cooperation from the origin of the term “cyborg” to a recent chess competition where a team of mediocre players and off-the-shelf computers beat both grandmasters and supercomputers. Argonne scientist Marius Stan, who uses computational methods to design and understand new materials for energy technologies, said that “knowledgeable machines” are now essential to science.
“Through computation we acquire a deeper understanding, our knowledge has another dimension,” Stan said. “There is truly an avalanche of data and metadata today that we need to face, and partnering with machines provides a path forward.”
A non-academic audience for these new approaches can be found in local and national governments, who strive to repurpose administrative data to improve operations and create evidence-based policies. Members of the “Future Decisions” panel talked about their work with these bodies, through CI programs such as the Array of Things urban sensing project and the Data Science for Social Good fellowship. Because many of these projects require a unique mixture of statistics, hardware engineering, social science, and communication with decision-makers at government agencies and non-profit organizations, the multidisciplinary community of the CI made for a perfect home.
“It’s been a place for us where we can share ideas across disciplines and across these fields of understanding,” said Julia Koschinsky, executive director for the Center for Spatial Data Science.
The Toolbox for Big Science
Computation Institute research groups also developed new tools and approaches for some of the world’s largest research projects and scientific instruments, including the Large Hadron Collider (LHC) at CERN, the South Pole Telescope, and Argonne’s Advanced Photon Source. As science has become more distributed and data-heavy, projects such as Globus, Swift, and the Open Science Grid have enabled collaboration and discovery in many different areas.
But as the meaning of “big data” only gets larger and larger, a panel on “Future Tools” discussed what computational products are needed for scientists to keep up. After current LHC upgrades are finished, the instrument is expected produce 100 times as much data, said moderator Rob Gardner, director of the ATLAS Midwest Tier2 Center. Panelists agreed that more usable, scalable, and sustainable tools that unlock the potential of cloud and high-performance computing will be necessary.
“If you don’t have the right culture, the tools you produce may be useful, but they won’t be interesting,” said Salman Habib, Senior Physicist & Computational Scientist at Argonne. “There should be a culture of innovation and new thinking.”
Fitting Academia Into AI
The day’s most speculative conversation took place in the last panel, centered around the impact of artificial intelligence. Stevens, the founding co-director of the CI with Professor of Mathematics and Computer Science Ridgway Scott, made his predictions for the next 15 years, covering the rise of quantum and neuromorphic computing, “active” storage devices that constantly analyze data, and AI-driven medicine from diagnosis to treatment.
As the panel demonstrated, the seeds for these sci-fi-sounding concepts can already be found today in projects such as the Learning Health framework of former CI deputy director Jonathan Silverstein or the computer vision research by Nicola Ferrier of Argonne. And the designers of these future technologies will be educated at schools like the University of Chicago, said Michael Franklin, the Liew Family Chair of Computer Science and senior adviser to the Provost on computation and data science at UChicago.
“We believe there’s an opportunity for a Chicago School of Data Science,” Franklin said. “It’s not just a combination of computer science and statistics, you have to build into it societal implications and an understanding of the fundamental questions of data science.”
Though the Computation Institute will end in June, its legacy will endure through initiatives such as the Center for Data and Applied Computing, the UChicago Center for Advanced Science and Engineering, and the expansion of the Department of Computer Science. By establishing numerous ports to exchange ideas between institutions and departments, the CI built a collaborative model that will continue to influence and accelerate science on both campuses.
“The mission was basically to make computing happen across the university and jointly with the lab,” said Stevens, Associate Laboratory Director for Computing, Environment and Life Sciences at Argonne and Professor of Computer Science at UChicago, “but the success was really about the people and interactions.”