The Making of a Scientist
Salk professor Terry J. Sejnowski has been elected to the National Academy of Engineering, making him one of only ten living individuals who are members of all three branches of the National Academies. For him, it is all about asking the right kinds of questions.
"So, Terry. What would it look like if there were a really big black hole in the middle of our galaxy?"
This first big question was posed by theoretical physicist John Wheeler, best known for coining the terms worm hole, quantum foam and black hole. Propelled by his interest in astrophysics, Sejnowski had joined lab Wheeler's at Princeton University as a physics graduate student just weeks earlier. At the time, black holes were nothing more than a mathematical curiosity based on the theory of general relativity, which predicted their existence. Undeterred, Sejnowski went off and found the answer by simulating the orbits of stars as they fell into a big black hole.
(All his predictions were later confirmed. He only missed the so-called accretion discs, Saturn-like rings formed by the stellar debris.)
"Nobody had ever asked that question before, and it was a transformative experience for me," he remembers. "It taught me that you need to ask a good question if you are going into a new area where there aren't lot of facts on the ground yet. When I became interested in the brain, in a sense it was like a black hole since back then not much was known about how things like memory work."
During his first hot and muggy summer in Princeton, as Sejnowski was cramming for his physics qualifying exams, he often went to the library to cool off body and mind. His recreational reading included picking up books on the brain, and soon the question that would change the course of his career arose: "Are the signals in the brain regular, like signals in a computer, or random?"
Sejnowski quickly found the answer. (They are random.) But the uncharted depths of the human brain had gripped his imagination. "I thought to myself, 'My god, here we are trying to understand the mysteries of the universe, when the mysteries of the brain are just as exciting,'" he says and adds with a laugh that "the advantage of biology is that you can actually do experiments, but you can't do experiments on the universe."
After completing his master's degree with Wheeler and his doctoral degree with John Hopfield at Princeton, he quickly immersed himself in neurobiology, learned how to record from neurons and before long was invited to join the Harvard lab of Steve Kuffler, who is often referred to as the "father of modern neurobiology." Remembers Sejnowski, "It was like jumping from a swimming pool into the ocean, but it was a fantastic experience."
As a young assistant professor at Johns Hopkins University, he continued to work on synapses, the specialized connections between neurons, but he also picked up some of the theoretical work on computational models of neuronal networks that he had done during his thesis.
Together with his lifelong friend Geoffrey Hinton, he developed the first neural network capable of learning to solve difficult computational problems. By overcoming a logjam that had hindered meaningful progress on neural networks for decades, the so-called Boltzmann machine helped spark the neural networks revolution in computing in the 1980s. "It was a beautiful model and actually one of my most influential projects from the perspective of the impact it had," he says.
Throughout his quest to understand behavior through the lens of neuronal circuits, Sejnowski has combined experimental and computational tools. "When I started, there were only a few people who were thinking about the brain in a theoretical way, and there was no real connection with biology," he says. "But I realized that to make progress you really have to connect it to the biology, and that's why I took the neurobiology summer course in Woods Hole after I got my Ph.D. From there it snowballed."
Today Sejnowski is regarded as one of the world's foremost theoretical brain scientists and celebrated as the founding father of the field of computational neurobiology.
His advice to the next generation of scientists?
"Find a really good question, and follow it through as far as you can push it. If it is a really good question, you are bound to come up with something truly interesting."