Imagine that finding the cure for cancer were only a matter of combining research from 50 different sets of data. Now, imagine if those data sets were scattered across cyberspace, some parts buried within research published by an obscure graduate student, others recently uploaded from historic medical texts, others hiding in plain sight in images of cells.
If finding the answers to the world’s biggest problems is indeed a matter of combing through all digital knowledge on that topic, the solution is simple — but extremely difficult for human beings to carry out within a lifetime. Advances in artificial intelligence (AI) and machine learning are giving researchers the power to accomplish feats that would formerly take humans centuries, and Pitt’s School of Computing and Information (SCI) is positioning itself to become a global leader in capitalizing on that power.
“No human can read 200,000 journal articles, but a machine can,” said Paul Cohen, the school’s founding dean.
Cohen announced the launch of the Modeling and Managing Complicated Systems (MoMaCS) Institute in May during the school’s inaugural Modeling the World’s Systems Conference in Pittsburgh. The two-day conference, which featured panel discussions on the potential for modeling national security challenges, the aging process, the opioid epidemic and more, was designed to bring collaborators to the table and examine how the institute and modeling can help to find new solutions to complicated and age-old questions.
Cohen said the institute will serve as a hub where researchers and stakeholders from across disciplines collaborate on projects that utilize the power of computer modeling and simulation.
The idea for the collaborative nature of the School of Computing and Information started with Cohen when he was program manager at the Information Innovation Office of the Defense Advanced Research Projects Agency, part of the U.S. Department of Defense. During that time, he took the lead on a program called Big Mechanism, which was focused on millions of potential protein interactions that take place within cell signaling pathways, particularly those that lead to cancer.
Cohen’s team created an AI natural language processing system to read hundreds of thousands of articles, separate information about the interactions and run tests to determine how proteins interacting with drugs operate on the pathways.
“(The team) started with 336 relevant genes, found and read 95,000 articles, extracted nearly a million assertions, filtered them and assembled them into signaling models sufficient to simulate the cellular dynamics and explained all drug-protein interactions accurately,” explained Cohen during the conference’s opening remarks. He said that with the AI natural language processing system, along with the help of an Amazon program that groups different types of data into logical categories, the entire project took less than a day.
At Pitt, the plan is to pair AI and machine learning researchers with individuals from academia, industry, nonprofits and the government to develop algorithms designed to address their specific problems and to use modeling experiments to provide concrete solutions.
“One day, presidents and cabinet officers, C-suites and lab directors will say, ‘Don’t tell me what your gut says, tell me what the evidence says; show me your models, show me the possible futures and the best interventions,’” said Cohen.
Early collaborators include Carnegie Mellon University’s School of Computer Science, which could join Pitt researchers on projects ranging from figuring out how to create fully autonomous vehicles to determining what makes people safe and happy on an individual level, according to CMU’s Andrew Moore, who also attended the modeling conference.
The institute will also team up with the Ford Foundation, a conference sponsor, for an initiative to research and model scenarios that steer people toward socially responsible investments.
During the conference, Chancellor Patrick Gallagher noted that the institute’s plan to tackle issues such as world hunger, curing disease, climate change and data breaches requires partnerships from outside the institution but also needs a level of interdisciplinary partnership that brings all schools and fields of discipline to the table.
“These are not siloed problems. For instance, ensuring the world’s population has enough to eat and combating climate change are indeed connected,” he said during his address at the conference.
“So, in creating SCI, we ushered in a new era of collaboration. And, admittedly, it’s still in its infancy. But the silos are disappearing.”