IF EVERYTHING had gone as planned, Walter Pitts would surely have become one of the most famous names of 20th-century science. He had already done seminal work in computing and neuroscience, and his peers were on the edge of their seats awaiting the next instalment. But in 1959 his world view was shattered by an experiment on a frog. He burned all his work, quit science and died alone 10 years later, largely forgotten.
If Pitts’s exit from science was unlikely, his entry was even more so. He was born in 1923 into a rough neighbourhood in Detroit, where his father was a boiler-maker used to solving problems with his fists. He wanted his son to quit school and get a job. Instead, the young Pitts hid in the local library to escape bullies and read books on Greek, Latin and Sanskrit. He was especially drawn to mathematical logic, the antithesis of his chaotic life.
When he was 12, he spent three days reading Bertrand Russell Principia Mathematica and then wrote to Russell pointing out what he thought was wrong with it. At 15, he ran away from home and gatecrashed lectures at the University of Chicago, where he stuck around doing menial jobs like a real-life Good Will Hunting . He never spoke to his parents again.
Advertisement
Pitts flourished in Chicago, impressing everyone he met with his intellect. He soon repeated the stunt he had pulled on Russell, this time with Rudolph Carnap, a philosopher who had made major contributions to mathematical logic. Carnap’s Aufbau was something of a manifesto for the group of mathematicians, scientists and philosophers known as the Vienna Circle. Either oblivious or indifferent to Carnap’s reputation, Pitts marched into his University of Chicago office with an annotated copy of the book pointing out its flaws.
Russell and Carnap helped Pitts to get a research position at the university and Warren McCulloch, a charismatic and up-and-coming scientist interested in the brain, took him into his home. They began a collaboration that would take science down a whole new path. Pitts was 17.
McCulloch wanted to know how the brain works – in particular, how the collective activity of individual neurons produces thoughts. Pitts threw logic at the problem, converting the messy behaviour of brain cells firing signals at one another into the discrete actions of a logic circuit: cell on, cell off. It was the first time anyone had seriously considered the brain as a computer. “For the first time in the history of science we know how we know,” McCulloch said at the time.
The brain-as-computer model was enormously influential in both neuroscience and computing. McCulloch and Pitts’s major work, ““, published in 1943 when Pitts was 20, inspired a generation of cyberneticians and early computer scientists. Their model of neuronal activity also led more or less directly to the invention of neural networks, underpinning today’s machine learning revolution. “The 1943 paper was epoch-making,” says at the University of Missouri-St Louis, who has researched McCulloch and Pitts as part of his work on computational theories of mind.
The pair would go on to create the first mechanistic theory of the mind, the first computational approach to neuroscience, the logical architecture of modern computers, and the pillars of artificial intelligence. The father of modern computers, John von Neumann, based his design for the EDVAC – the first computer that could store programs – on Pitts and McCulloch’s logical model of the brain.
Pitts was on the way to becoming a household name. In 1954, Fortune magazine placed him alongside James Watson and information theory founder Claude Shannon in an article on the world’s 20 most talented young scientists. But he was seldom at ease. Colleagues noticed his shyness and discomfort in social settings.
“For the first time in the history of science, we know how we know”
He decided to write his PhD dissertation on probabilistic three-dimensional neural networks, an ambitious extension of his earlier model of the brain and one that would demand ridiculous mathematical acrobatics. His peers were floored by the proposal, but Pitts was confident he could pull it off.
His interests were also broad. He dazzled contemporaries with his genius in everything from physics and chemistry to history and botany. But his curiosity would lead him to a discovery he could not handle.
Working again with McCulloch, Pitts started to study the visual system in animals, leading to a 1959 paper called “What the frog’s eyes tell the frog’s brain”. What they found overturned everything Pitts had thought to be the case up to that point.
It turns out that a frog’s eyes process visual information before passing it on to the brain, filtering things like contrast and movement. Not only was the brain not the sole processing unit, but the processing done in the eye was analogue – messy and approximate. Even if logic still played a role in the brain, it wasn’t nearly as important or central as Pitts had thought.
It was the straw that broke him. He had been drinking heavily for some time to cope with a series of social upsets and chronic depression. With the undermining of his intellectual work, he unravelled. He burned his magnum opus and all the notes that went with it. Years of much-awaited work were gone.
For the next 10 years, Pitts did little but drink, publishing only one more paper – this time on a frog’s sense of smell. In May 1969, aged 46, he died alone in a boarding house in Cambridge, Massachusetts, of alcohol-related causes.
Pitts’s breakdown was a major loss to science. His published ideas about neural networks paved the way for a revolution in artificial intelligence decades later. Though we can never know the contents of his burned manuscript, Piccinini is convinced the whole field would have progressed much more quickly had Pitts completed his work.
So why have we forgotten him? In a sad twist, it may have been his association with McCulloch. Pitts’s collaborator was known for being sloppy and an alcoholic himself. “The 1943 paper is in many ways a mess,” says Piccinini. “It is very difficult to follow both conceptually and mathematically, and full of bold but unsubstantiated claims.”
Worse, it seems other researchers deliberately stayed away. John McCarthy, who coined the term “artificial intelligence”, told Piccinini that when he and fellow AI founder Marvin Minsky got started, they chose to do their own thing rather than follow McCulloch because they didn’t want to be subsumed into his orbit.
Without McCulloch, Pitts would not have done the work he did. But with him, his work may have become tainted – at least until later generations began to rediscover it.
This article appeared in print under the headline “AI’s lost genius”
