Biography · Marvin Minsky

He set out to disassemble the mind into countless little machines — and ended by freezing, with one book, another road that led to the mind.

A Neural Network That Could Not Think

Princeton, 1951. A twenty-four-year-old graduate student and his classmate Dean Edmonds packed three thousand vacuum tubes, the surplus autopilots of forty B-24 bombers, and great coils of cable into a single room. The machine they built was called SNARC — Stochastic Neural Analog Reinforcement Calculator — the first hardware simulation of a learning neural network in history. It could simulate a rat in a maze: when the rat went the right way, the weights of the correct synapses were strengthened; when wrong, weakened.

The graduate student was Marvin Minsky. SNARC became his doctoral dissertation. It was not a thinking machine, but it was the first time learning had been pulled out of the mysterious territory of biology and into the workshop of the engineer. The young Minsky believed that the mind was no miracle, but a process that could be reproduced.

Minsky was born in New York in 1927, son of an ophthalmic surgeon. As a Harvard undergraduate he majored in mathematics while wandering between physiology lab and concert hall — a lifelong pianist, a lifelong devotee of the brain's neural circuitry. This double training, engineering and biology, ran through six decades of his research life.

The irony is that eighteen years later this same Minsky would, with his own hand, press the embers of the neural network movement into ash.

The Summer at Dartmouth

In the summer of 1956, in a small building at Dartmouth College in New Hampshire, John McCarthy, Claude Shannon, Nathaniel Rochester, and Minsky convened ten scholars for two months. The meeting would later be called the birthplace of AI.

Minsky was twenty-nine. He brought no thunderclap of a paper, but in that room he was the man who most firmly believed the mind could be built. His optimism was nearly religious — he was certain that within a few generations machines would attain human-level intelligence. That optimism carried him through the next sixty years, and through every spring and winter the field would suffer.

After the meeting, Minsky and McCarthy went together to MIT. In 1959 they founded the MIT Artificial Intelligence Project, which would become the MIT AI Lab — perhaps the single most important monastery in AI's history.

Perceptrons: One Book, Twenty Years of Frost

In 1957 Frank Rosenblatt built his Perceptron at Cornell — an electronic machine that could learn, hailed by The New York Times as "the embryo of an electronic computer that … will be able to walk, talk, see, write, reproduce itself." Rosenblatt promoted his work loudly; Minsky watched coldly. The two had been classmates at the Bronx High School of Science.

In 1969 Minsky and Seymour Papert published Perceptrons: An Introduction to Computational Geometry. The book was a mathematical monograph of forbidding rigour. It proved that a single-layer perceptron could not even solve XOR — a function that simply distinguishes whether two inputs differ. The two authors hinted darkly that multi-layer networks would do no better.

The influence of the book outran its intent. Funding for neural network research dried up almost overnight. The connectionist line in the United States and Britain went nearly silent. The field was frozen for almost twenty years. Only in 1986, when David Rumelhart, Geoffrey Hinton, and Ronald Williams published the backpropagation paper in Nature, did the road thaw — by which time Rosenblatt had been dead, in a boating accident, for fifteen years.

Minsky later defended himself: he had never denied the potential of multi-layer networks, only that no one then knew how to train them. History has not been kind to the defence. In the memoirs of many connectionists, Perceptrons is a book that "killed a generation of research." This is Minsky's original sin and the most famous methodological dispute in the history of AI.

Frames: A Proposal for Knowledge Representation

The other Minsky was one of the deepest thinkers in the symbolic camp. In 1974 he circulated the MIT memorandum "A Framework for Representing Knowledge," proposing frame theory. Humans, he argued, do not understand the world by isolated propositions but by structured templates — the "birthday party" frame, with its slots for cake, presents, candles, the celebrant, the song, and their default values. When we walk into a strange room, it is these prior frames that allow us to make instant sense of what we see.

Frame theory shaped the early designs of knowledge representation, semantic networks, and object-oriented programming. Roger Schank's script theory was a near cousin. Today's ontology engineering and knowledge graphs trace their ancestry back to that memorandum.

The Society of Mind

In 1985 Minsky published The Society of Mind. Its central claim was radical: the mind has no centre, no commander-in-chief, no single "I." The mind is a society of vast numbers of simple, clumsy agents — each of which can do only some tiny thing, like "detect an edge," "grip a cup," "avoid collision." Intelligence emerges from the cooperation, competition, and switching among these agents.

The book was written in an odd form: one idea per page, half aphorism, half philosophy textbook for children. It offered no rigorous algorithm, but a unified picture of mind. Researchers in multi-agent systems and behaviour-based robotics drew freely from it.

In 2006 he extended the model in The Emotion Machine, folding emotion into the agent picture: emotion is not the opposite of reason but a mechanism by which the brain switches modes of thought.

Kubrick's Consultant, and a Final Shadow

Minsky's reach went beyond academia. In 1968, when Stanley Kubrick was making 2001: A Space Odyssey, he hired Minsky as the film's AI consultant. The film's machine intelligence, HAL 9000, drew on many of Minsky's ideas about machine cognition — and on his unease about its possible failures.

The students he trained were a constellation: Patrick Winston, who would later run the MIT AI Lab; Terry Winograd, who wrote SHRDLU; Roger Schank, who created script theory; Carl Hewitt, who invented the Actor model; Richard Stallman, who launched the Free Software movement. After he and McCarthy, the MIT AI Lab raised half a generation of AI history within its walls.

In 1969, at forty-two, Minsky received the Turing Award.

But honour does not erase shadow. In 2019, while The New York Times and others investigated the Jeffrey Epstein case, it emerged that Minsky had visited Epstein's private island multiple times and was named in a victim's testimony. Minsky had died in 2016 and could not respond. His family denied the allegations; the MIT Media Lab was caught up in a related scandal of donations, and its director Joi Ito resigned. The late years of an AI elder thus settled into the historical record in a way that is uncomfortable to acknowledge.

Selected Works

Year	Work	Significance
1951	SNARC (with Edmonds)	First hardware simulation of neural network learning
1961	"Steps Toward Artificial Intelligence," Proc. IRE	Early road map of AI; shaped a generation of researchers
1969	Perceptrons (with Papert)	Showed the limits of single-layer perceptrons; froze connectionism for nearly two decades
1974	"A Framework for Representing Knowledge" (MIT AI Memo 306)	Proposed frame theory; founded knowledge representation
1985	The Society of Mind	Multi-agent theory of mind; influenced multi-agent research
2006	The Emotion Machine	Folded emotion into the agent framework of cognition

Historian's Note

Historian's Note

Minsky was both founding father of AI and the magistrate of connectionism's winter. He lit the first spark of neural networks with SNARC, then with Perceptrons he pressed it into ash; he stitched together the most refined map of symbolic AI through frames and agents, and never lived to see that road wholly bypassed by deep learning. He sought all his life to understand the mind, and so dismantled the mind into countless little machines; and yet, with one book, he blocked another road to the mind for twenty years. Methodological disputes are common; the depth and cost of Perceptrons lay an unshakable responsibility on him beyond his accomplishments. As for the Epstein shadow at the end of his life, that is a separate moral ledger — genius does not exempt anyone from human filth. There is no commander-in-chief in the society of mind; there is no perfect man in the history of AI.

Eyewitness Accounts

Call for contributions

If you knew Marvin Minsky personally or have firsthand sources or recollections, please contribute on GitHub.

References

1. Minsky, M. (1954). Theory of Neural-Analog Reinforcement Systems and its Application to the Brain-Model Problem. PhD dissertation, Princeton University.
2. Minsky, M. (1961). "Steps Toward Artificial Intelligence." Proceedings of the IRE, 49(1), 8–30.
3. Minsky, M., & Papert, S. (1969). Perceptrons: An Introduction to Computational Geometry. MIT Press.
4. Minsky, M. (1974). "A Framework for Representing Knowledge." MIT AI Laboratory Memo 306.
5. Minsky, M. (1985). The Society of Mind. New York: Simon & Schuster.
6. Minsky, M. (2006). The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and the Future of the Human Mind. New York: Simon & Schuster.
7. McCorduck, P. (2004). Machines Who Think (2nd ed.). Natick, MA: A K Peters.
8. Olazaran, M. (1996). "A Sociological Study of the Official History of the Perceptrons Controversy." Social Studies of Science, 26(3), 611–659.
9. Markoff, J. (2016). "Marvin Minsky, Pioneer in Artificial Intelligence, Dies at 88." The New York Times, January 25, 2016.
10. Aldhous, P., et al. (2019–2020). Reporting on Jeffrey Epstein and the MIT Media Lab. The New York Times / The New Yorker.