Marvin L. Minsky

Marvin Minsky (New York, United States; 1927 – Boston, United States; 2016) graduated in mathematics from Harvard University (1946) and went on to complete his PhD at the University of Princeton (1954). In 1959, he took up a professorship at the Massachusetts Institute of Technology (MIT), where he was co-founder, with John McCarthy, of the Artificial Intelligence Laboratory. Appointed Professor of Electrical Engineering and Computer Science at MIT in 1974, he combined this post with that of Donner Professor of Science until 1989.

From 1990 until his death in 2016 he has held the Toshiba Professorship of Media Arts and Sciences, devoted to the most advanced IT research. His many honors include the Turing Award (1969) and membership of the IEEE Intelligent Systems Hall of Fame (2011), and he was member of prestigious scientific societies including the American Academy of Arts and Sciences or the American Association for Artificial Intelligence.

Minsky also influenced future generations through his PhD mentoring work, with many students going on to occupy top research positions in the AI and computer science worlds.

Marvin Minsky, “father” of artificial intelligence, author of essential contributions in mathematics, cognitive science, robotics and philosophy, and recipient of the BBVA Foundation Frontiers of Knowledge Award in Information and Communication Technologies, remained convinced that we will one day build machines at least as intelligent as human beings. He did not believe it will happen soon, having avowedly seen “nothing that surprising” in recent years. But if past experience is anything to judge by, the wait will have been worth it. For over the last few decades, the quest to create non-human intelligence has inspired ideas and applications that have transformed our society.

Minsky was Professor of Electrical Engineering and Computer Science at Massachusetts Institute of Technology (MIT), where he was also Toshiba Professor of Media Arts and Sciences. Since his time as a student at Harvard University, where he graduated with a degree in mathematics in 1950, he was fascinated by the human mind and the emergence of its cognitive functions. It was while pacing the library floor leafing through Mathematical Biophysics by Nicolas Rashevsky that he knew his future must lie in the construction of learning machines. Rashevsky’s book mathematically described the workings of biological systems — the beating of the heart, the firing of electrical signals among neurons, etc. — and informed the mechanistic view of the human mind that defines Minsky’s work.

His PhD thesis, which he defended at Princeton University in 1954, explored how to construct neural networks capable of learning. This line of work bore no relation to the computers of the time; giant calculating machines performing highly specific tasks for a few major institutions. Minsky described them thus in a 1982 essay: “…when computers first appeared, their designers intended them for nothing except to do huge, mindless computations. That’s why the things were called ‘computers’. But even then, a few pioneers were envisioning what we now call artificial intelligence or AI. They saw that computers might possibly go beyond arithmetic, and maybe imitate the processes that go on inside human brains.”

Artificial intelligence officially came into being as a discipline at a computer science conference in Dartmouth College (New Hampshire, United States) back in 1956. Accompanying Minsky on the founders’ platform were John McCarthy — inventor of the term — Allen Newell and Herbert Simon. These were times of optimism. Advances in programming were beginning to yield results, in the form of software for integral calculus, algebra, etc. And Minsky was quoted as saying in the 1960s that “within a generation, the problem of creating artificial intelligence will substantially be solved.” Even then, however, he knew that the strategy being pursued to achieve AI had its Achilles’ heel: the machines could solve advanced math problems, “but we didn’t have a system [capable of understanding] that you can pull with a string, but you can’t push, or that normally people don’t like to get wet so they take umbrellas when it rains, but not when they go swimming…,” he explained in an interview on Web of Stories.

What is missing is common sense, and Minsky led the quest to endow computers with this elusive quality. How to teach a computer tasks that come easily to the youngest child? “We rarely recognize how wonderful it is that a person can traverse an entire lifetime without making a really serious mistake, like putting a fork in one’s eye or using a window instead of a door,” wrote Minsky in one of his best-known works, The Society of Mind. In 1959, Minsky joined the faculty at MIT, where he and McCarthy co-founded the Artificial Intelligence Laboratory. His curriculum was by then already rich in practical achievements: the first neural network learning machine, SNARC, a device made out of vacuum tubes (1951); and the confocal scanning microscope (1955), still widely used in biology for its ability to reconstruct 3D images.

And the list would continue with new contributions like the first head-mounted graphical display (1963); a robotic arm (1967), now on show at Boston’s Museum of Science; or a small computer-controlled robot, the LOGO “turtle” (1972), which draws as it moves. And of course Minsky did not confine his creative impulse to the lab: a fan of science fiction, he was an advisor to Stanley Kubrick on 2001: A Space Odyssey (1968) and gave Michael Crichton the underlying idea for Jurassic Park (1990).

However, the success of the Artificial Intelligence Laboratory could not disguise the fact that the discipline was struggling to meet its formidable goals. In the 1980s and 1990s, it became increasingly evident that it is easier for a machine to solve complex operations than to develop common sense, and that it would take time to emulate the human brain. Minsky, meantime, continued to do his bit on the conceptual plane. In 1974, he published a theory postulating that knowledge can be represented by previously acquired stereotypes or frames. Frames function as organized deposits of earlier knowledge and experience that enable us to process information. And the language of frames is now a well-established AI tool.

Next came The Society of Mind (1987): intelligence, Minsky argued, arises from the interaction of myriad non-intelligent parts. And emotions — The Emotion Machine (2006) — are simply the product of a different level of mental processing, another way of solving problems. What this implies is that a human-like intelligence may even arise from the natural flow of progress: as we come to understand more about the workings of the brain we will build smarter machines that, in turn, will teach us more about the brain, and so forth. The loop will continue until we are faced with the dilemma of whether or not to create machines more intelligent than us. “We are fortunate to be able to leave that decision to future generations (…). But right now one thing is sure: there’s something wrong with any claim to know, today, of any basic differences between the minds of men and those of possible machines.”

Until this future arrives, the present belongs to the technology inherited from AI. Drones and driverless cars, medical diagnostic systems, applications in robotics… And ruling them all, computers. An explosion of digital diversity has transformed these single-purpose machine minds, exclusive to large corporations, into the first universal and ubiquitous device, man’s best friend in every area of life. Will there be some future HAL that traces its genealogy to our PCs, tablets and smartphones?

Scientist Marvin Minsky in conversation with filmmaker Kike Maillo