Marvin L. Minsky (1927–2016) a Founding Father of Artificial Intelligence

COMMENT OBITUARY Marvin L. Minsky (1927–2016) A founding father of artificial intelligence.

arvin Lee Minsky had no patience education that centred on teaching children for those who doubted that com- to program using the Logo language. puters could be intelligent at a In the mid 1970s, Minsky introduced Mhuman level or beyond. In the late 1950s, ‘frames’, a way of describing entities and building on the work of Alan Turing, along situations using a template-like representa- with computer scientists John McCarthy, tion. A frame describing a birthday party, Herbert Simon and Allen Newell, Minsky for example, would have a slot for the person COURTESY OF MINSKY FAMILY OF MINSKY COURTESY started the work that led everyone to think celebrated, that person’s age and a list of the of this group as the founders of the field of gifts presented, along with slots for time and artificial intelligence (AI). Were it not for place inherited from a ‘celebration’ frame. He their determined advocacy, AI might have also developed the idea of knowledge lines foundered. (K-lines) to address questions about how Minsky, who died on 24 January, was born information is represented, stored, retrieved in New York City in 1927. After serving in and used in the brain. He argued that K-lines the US Navy in the Second World War, he help us to solve problems by putting us back earned a degree in mathematics in 1950 into mental states that resemble those we were from Harvard University in Cambridge, in when we previously thought about similar Massachusetts, where he impressed the problems. mathematician Andrew Gleason by prov- In 1985, he brought these and many other ing fixed-point theorems in topology. Dur- ideas together in a book, The Society of Mind ing his doctorate on learning machines, at (Simon & Schuster). He wrote that intelli- Princeton University in New Jersey, he built His laboratory was an egalitarian utopia. gence emerges from the cooperative behav- one out of vacuum tubes and motors. He didn’t notice looks, gender, age or sta- iour of multiple agents, none of which is When Minsky finished his PhD, the tus. He cared only about ideas and ability. intelligent. Then, in 2006, Minsky published eminent mathematicians John von Minsky and his wife, Gloria, often welcomed The Emotion Machine (Simon & Schuster), a Neumann, Norbert Wiener and Claude students into their home, where several book about intelligence, creativity, emotion, Shannon all recommended him for appoint- pianos stood as a reminder that Minsky was consciousness and common sense. Multi- ment as a junior fellow at Harvard. During a musical prodigy, able to improvise fugues. plicity is a dominant theme. He noted, for this time, he became curious about how the Minsky’s attention span was short. When- example, that concepts such as intelligence brain works, but was frustrated by the limi- ever I explained an idea to him, he would leap are ‘suitcase words’, into which one can tations of conventional microscopy, which ahead of me, having worked the whole thing stuff multiple meanings. He wrote that our could not provide clear images of thick, out after a few sentences. Once, I suggested resourceful intelligence arises from multiple light-scattering neural tissue. This led to his that if we ever developed really intelligent ways of thinking on multiple levels, and from invention of the confocal scanning micro- machines, we should do a lot of simulation multiple ways of representing knowledge. scope, which uses lenses to focus light on before we let them loose in our world to be In recent years, Minsky found it ironic that successively small volumes. sure they weren’t dangerous. “And we’re the the doubters of the possibility of AI have been In the late 1950s, with McCarthy, he simulation?” he said, guessing my punchline. replaced by worriers about its consequences. founded a group that became the Artifi- “It isn’t going very well, is it?” He didn’t see a technical advance that would cial Intelligence Laboratory at the Massa- His laboratory built pioneering robots justify the change in attitude, attributing chusetts Institute of Technology (MIT) in as well as revolutionary programs. Minsky recent successes in AI to faster computers. He Cambridge. In 1961, Minsky published his invented a robot arm with 14 degrees of thought that not much real progress had been famous paper ‘Steps toward artificial intelli- freedom. He argued that space exploration made in the field for several decades, but he gence’ (M. Minsky Proc. IRE 49, 8–30; 1961), and nuclear-material processing would be had, nevertheless, no doubt that our species’ a call to arms for a generation of researchers. simpler with manipulators driven locally by greatest legacy will be the intelligent comput- Scientists flocked to Minsky’s laboratory to computers or remotely by human operators. ers that we create. take on the challenge of understanding intel- He foresaw that microsurgery would be done Minsky’s talks, papers and books are like ligence and endowing computers with it. by surgeons by using telepresence systems. diamond mines. The riches will take decades They benefited from Minsky’s wisdom and In the late 1960s, Minsky and MIT math- to cut and polish, inspiring researchers for enjoyed his insights, lightning-fast analyses ematician Seymour Papert worked on the decades to come. ■ and impish repartee. mathematics of perceptrons — simple neu- His students felt part of a scientific revolu- ral networks — showing what they could Patrick Henry Winston is professor of tion. They helped Minsky to develop high- and could not do, which raised the sophis- artificial intelligence at the Massachusetts level theories about how programs could tication of research on neurally inspired Institute of Technology, Cambridge, recognize structures made of toy blocks, mechanisms to a higher level. Minsky and Massachusetts, USA. He was a graduate answer questions about stories written for Papert collaborated into the 1970s and student of Marvin Minsky’s in the 1960s, and children, learn something definite from indi- early 1980s, developing theories of intel- thereafter an admiring friend and colleague. vidual examples and exhibit common sense. ligence and radical approaches to early e-mail: [email protected]