Thinking Outside the Brain
By Amanda S.F. Hartle
When Carnegie Mellon University alumnus and Director of Neuroscience Research at DeepMind Matthew Botvinick (DC 2001) walked into the first biology class of his life as a teenager, he was confident all the questions in his brain — about his brain — were about to be answered.
“When I got the textbook, the first thing I did was turn to the brain section. I thought finally I was going to learn something about the brain,” Matthew recalls. “Both because it was the 1980s and this was a junior high textbook, there wasn’t much in there. I thought there’s nothing here about the mind, attention, memory, personality or awareness.”
Unable to satisfy his interest about neural networks, he pivoted toward a career in medicine. During medical school at Cornell University, he told the physician advising his psychiatry rotation his conclusions learned in that first biology class.
In response, he handed Matthew a two-volume set of books — “Parallel Distributed Processing: Psychological and biological models” by then CMU University Professor Jay McClelland.
“I was blown away. I stayed up all night reading them,” Matthew remembers. “It was an exciting discovery that I could actually pursue those things that I’d been interested in my whole life and here were a set of researchers who were doing that in a scientifically meaningful way.”
That sleepless night put him on his path through a doctorate in psychology and cognitive neuroscience at CMU’s Dietrich College of Humanities and Social Sciences earned simultaneously with a medical residency in psychiatry, then professorships at the University of Pennsylvania and Princeton University and, eventually, to Google partner DeepMind in London.
“I took a sabbatical from Princeton in 2015. I wanted to find out what was going on at DeepMind and update my own understanding as to what was happening on the machine learning side,” he says. “I never left.”
At artificial intelligence company DeepMind, he leads the neuroscience team that is pioneering an interdisciplinary approach to the field by bringing together machine learning, neuroscience, engineering, mathematics, simulation and computer infrastructure. DeepMind programs have helped diagnose eye diseases more effectively, reduced the endless energy used to cool data centers and predicted protein shapes in ways that may change how drugs are created.
“CMU was the place to be to do neural network research. It was really the only place in the world where I could’ve done my graduate work, and it put me on a path to what I do now that I love so much.”
“I’m interested in the aspects of AI that are still out of reach for our engineering systems, which are clearly the things humans are good at — abstract reasoning, concept acquisition, understanding physics in an intuitive way,” Matthew says. “Sometimes it’s the simplest things for humans to do that are the hardest for AI to do.”
Along with his team of 50 staff members, he seeks to bridge the gap between neuroscience and AI and utilize each to inform better understanding of the other.
“I’m exploring the ways AI can help humans find better ways of cooperating and better ways of making decisions in their own interests,” Matthew says. “At the same time, we want the AI systems to avoid anything coercive or deceptive and enhance human welfare and human freedom at the same time. I’m really interested in how to get that right in a way that makes the world better for humans.”
As DeepMind and its world-altering ambitions enamored Matthew, CMU had a similar draw for him.
“CMU was the place to be to do neural network research,” Matthew says. “It was really the only place in the world where I could’ve done my graduate work, and it put me on a path to what I do now that I love so much.”
At CMU, Matthew learned from world-renowned professors at the Dietrich College — David Plaut and Marlene Behrmann.
“I simply wouldn’t be where I am today without CMU. The professors were really, truly available. I interacted with these internationally famous professors every day in meaningful ways. It was a hive of intellectual activity,” he says.
Matthew embraced what he calls the university’s “distinct interdisciplinary environment,” and loves that his career has come full circle with a similar atmosphere at DeepMind where from minute to minute he can be found shifting between neuroscience research, psychology explorations and AI.
“My ability to move that way throughout the day, I first learned at CMU,” Matthew remarks. “The academic lines or departmental lines didn’t matter at Carnegie Mellon. We were people who wanted to understand, and we followed that wherever it led.”
