Marvin Minsky

Overview

Marvin Lee Minsky (1927–2016) was an American scientist who helped found artificial intelligence as a field. He worked on learning machines, knowledge representation and theories of how minds might be organised. He was born in New York City on 9 August 1927 and died in Boston on 24 January 2016, aged 88.^[1]

Education and an early neural network

After serving in the U.S. Navy during the Second World War, Minsky graduated from Harvard University in 1950. He received a PhD in mathematics from Princeton University in 1954, with a thesis on neural-analog reinforcement systems.^[2]

In 1951, during his first year at Princeton, he built what MIT describes as the first neural network simulator, known as SNARC. A neural network simulator is a machine or program that imitates networks of simple neurons so that their learning can be studied.^[1]

The Dartmouth proposal and MIT

Minsky was a Junior Fellow of the Harvard Society of Fellows from 1954 to 1957. While at Harvard he co-signed the August 1955 Dartmouth proposal on artificial intelligence with John McCarthy, Nathaniel Rochester and Claude Shannon, and he attended the full 1956 Dartmouth workshop.^[3]

He joined the MIT faculty in 1958. In 1959 he co-founded the Artificial Intelligence Laboratory with McCarthy. It later became MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL).^[1]

Perceptrons and the limits of simple networks

With Seymour Papert, Minsky wrote “Perceptrons”, published by MIT Press in 1969. The book is a mathematical analysis of the limits of single-layer perceptrons, a simple kind of learning network in which only one layer of connections is adjusted. Many readers took it as showing serious limits for this approach. It is often linked to the decline of neural-network research in the 1970s.^[2]

Knowledge representation and the mind

In his 1974 paper “A Framework for Representing Knowledge”, Minsky introduced frames, a way of storing knowledge as structured packages with named slots that can hold default values. Frames are one of the early ideas in knowledge representation, a branch of AI concerned with how a machine can store facts so it can use them. His books include The Society of Mind and The Emotion Machine, published in 2006.^[2]

Media Lab and other work

Minsky was a founding member of the MIT Media Lab, established in 1985, where he held the Toshiba Professor of Media Arts and Sciences post. He also invented the earliest confocal scanning microscope.^[2]

Honours

He received the ACM A. M. Turing Award in 1969, the Japan Prize in 1990, the Benjamin Franklin Medal and the BBVA Foundation Frontiers of Knowledge Award.^[1][2]

Legacy

Minsky is remembered as one of the people who gave AI its shape in its early decades. Perceptrons is still read as a key document in the history of neural networks, because its critique of single-layer models helped set the agenda for the research that followed.