From machine-learning algorithms detecting emergent diseases to a data-driven approach to addressing housing inequity, six UC Berkeley-led projects have been awarded funding from the recently created C3.ai Digital Transformation Institute to use artificial intelligence, or AI, to help fight the spread of COVID-19.
Each project was awarded a one-year grant in the range of $100,000 to $500,000, according to Tamara Straus, a spokesperson for the institute. The projects were chosen by more than 500 reviewers based on a variety of criteria, including originality of research, use of AI, multidisciplinary applications and potential to grow with further support, Straus added.
In addition to funding, researchers will gain access to the C3.ai computing platform and supercomputing resources from the Lawrence Berkeley National Laboratory and the University of Illinois at Urbana-Champaign. Teams will also have at their disposal the C3.ai COVID-19 Data Lake, which consolidates COVID-19-related data from multiple sources.
UC Berkeley electrical engineering and computer sciences, or EECS, professor Alberto Sangiovanni-Vincentelli is working on developing algorithms and modeling techniques to detect diseases such as COVID-19 before they spread, as well as methods to translate symptoms from foreign clinical data and possible treatments from other diseases.
“By the time they figured out what (COVID-19) was, the disease had already spread,” Sangiovanni-Vincentelli said. “We want to be able to advise the clinical people, there’s something new coming — watch out.”
Another awarded project, COVIDScholar, uses natural language processing to sort through about 100,000 pieces of COVID-19-related scientific literature, as well as papers dealing with similar diseases such as severe acute respiratory syndrome and Middle East respiratory syndrome.
Principal investigators and UC Berkeley professors Gerbrand Ceder and Kristin Persson plan to use the grant money to pay existing researchers and hire new ones, as well as to take advantage of the C3.ai computing platform and data archive.
“Basically (the AI) is reading all the papers at once,” said team co-lead and campus graduate student John Dagdelen. “You’re able to derive insights from the entire collection simultaneously, rather than just one at a time. And when you do it that way, a lot of patterns that you wouldn’t have seen otherwise start emerging.”
The team’s website already sees traffic of about 1,000 scientists per week, and Dagdelen is confident about the future of the project.
“In the near future, you’re going to have basically some sort of AI assistant that will help identify promising research leads for you,” Dagdelen said. “We’re not quite there yet, but we’re very close.”
The lineup of projects is rounded out by an investigation of COVID-19-related eviction risks, led by campus faculty Karen Chapple and Joshua Blumenstock; UC Berkeley chemistry professor Teresa Head-Gordon’s search for molecules that could disable the COVID-19 virus; and a project led by EECS professor Jennifer Listgarten using discrete mathematical techniques to discover new drug treatments for the disease.
“The combination of big data, cloud computing, and machine learning is creating an unprecedented opportunity for fast, precise analyses and informed decision making,” said Shankar Sastry, UC Berkeley professor and co-director of the Digital Transformation Institute, in an email. “We need this now more than ever. The COVID-19 pandemic shows us that without precise information, followed by quickly deployed resources, people suffer and die.”