home | non-tech | cs | ee | about
Abstract New

Walter Pitts - A tribute to the unknown genius behind cybernetics

It was very many years ago, in grad school at Chicago, that I first heard the name of Walter Pitts. The professor was expounding on the perceptron when he mentioned the enigmatic autodidact, Walter Pitts, who worked with Warren McCulloch to come up with the first mathematical model of the neural network. I remember a student piping up - "A real-life Good Will Hunting". Fascinated, I dug up more details on his life, only to find that his was a life far more extraordinary. A life that culminated in a tragedy.

I recently found my notes from more than 15 years ago on Pitts - collected from various papers, professors, and the then world wide web. I thought committing them to the memory of cyberspace would be useful to some. So here goes.

Not much is known about Pitt's early life, save for the fact that he was born in Detroit. He was an autodidact who taught himself logic and could read Latin, German, and Sanskrit. According to some sources, at the age of twelve, he mastered the Principia Mathematica and sent letters to Bertrand Russell on what he considered serious issues with the first volume of the work.

Lettvin, Carnap and Rashevsky

At the age of 15, he ran away to Chicago to listen to Russell's lectures at the University of Chicago. Pitts then joined the physicist Rashevsky's group as an unregistered junior. He was homeless and made his living by ghosting papers and working other menial jobs.

The Russian born Rashevsky was himself a child prodigy, having completed his doctorate before he turned twenty. After a brief stint at the Russian University in Prague, Rashevsky immigrated to the United States in 1924 to work at Westinghouse Electric. In 1934, after being awarded a Rockefeller Fellowship, he joined the University of Chicago as an assistant professor. Rashevsky was a pioneer of mathematical biophysics and laid the foundation for Pitts' later work in computational neuroscience.

Around this time, anecdotal evidence suggests that Bertrand Russell corresponded with him and encouraged him to study logic. Rudolf Carnap was one of the eminent philosophers teaching logic at the University of Chicago. Pitts studied Carnap's work and walked into his office with notes on his seminal work - "The Logical Syntax of Language". He outlined what he felt were inconsistencies in Carnap's work and suggested solutions. Carnap, greatly impressed by this meeting, tried to find Pitts so he could talk to him again. Sources vary, but generally agree on the fact that Carnap, after searching for weeks, found Pitts homeless and working menial jobs at the university. Carnap got him a job. Around this time, the then homeless Pitts, met Jerome Lettvin, who would become one of his life long friends.

Warren McCulloch

It was in 1942 that Pitts first met Warren McCulloch. McCulloch had degrees in philosophy and psychology and received an MD in 1927. His interests ranged from poetry to engineering. In 1942, after having worked at Yale for almost 8 years, he moved to the University of Illinois at Chicago.

McCulloch and his exceptional wife Rook McCulloch took the then homeless Pitts along with Lettvin into their home. McCulloch became Pitts mentor and together they published their seminal paper - "A logical calculus of the ideas immanent in nervous activity" where they presented a simple mathematical model of a neuron. This paper laid the foundation for the field of artificial neural networks. If anyone had an insight into Pitts psyche, it would have been McCulloch.

Norbert Wiener

Lettvin introduced Pitts to Norbert Weiner at MIT. Born to a Harvard professor, Norbert Wiener was a celebrated child prodigy of his time. He had entered Tufts College as a eleven year old graduated with a BA in mathematics at the age of fourteen. He was awarded a Phd by Harvard when he was a mere eighteen years old. He travelled the world participating in many international conferences and even spent an year teaching in China. He joined MIT in 1919.

For a decade Wiener worked with Pitts. During this time, Pitts worked for a brief while at the Kellex Corporation - a company formed by MV Kellog company in 1942 to establish a team to work operations on the Manhattan project. Not much is known about Pitts' work here. He later returned to MIT and continued working with Weiner.

Pitts continued to collaborate with McCulloch and wrote an influential paper on the perception of auditory and visual forms. In 1952, McCulloch and Lettvin moved to MIT to join Wiener and Pitts.

Later years and decline

In 1952, Wiener broke all relationships with McCulloch and people associated with him. It is not clear if this had anything to do with Pitts emotional decline. According to some, in Lettvin, he had found friendship, in McCulloch a mentor, but in Wiener he found a father figure. Completely broken, Pitts went into an emotional downward spiral. He burnt some of his work, including a paper on three dimensional neural networks. He tried hard not to be found by any of his friends and colleagues, completely dissappearing from the scene towards the end of 1950s.

After having faded into obscurity, Pitts died in 1969, alone in a boarding house in Cambridge.

That was the story of Walter Pitts, the genius who came up with the first mathematical model of the neuron. Walter Pitts, the autodidact without degrees, who influenced the giants of his time.

No comments:

Post a Comment

© 2014 - 2015 abstract new. All rights reserved.