Minsky , Marvin Lee

(1927–) American computer scientist

The son of a surgeon, Minsky was born in New York and educated at Harvard and at Princeton, where he obtained his PhD in 1954. He taught at Harvard before moving to the Massachusetts Institute of Technology in 1957 as professor of mathematics, a post he occupied until 1962 when he became professor of electrical engineering. He also served as director of the Artificial Intelligence (AI) Laboratory (1964–73).

In the summer of 1956 Minsky attended a conference on AI at Dartmouth, New Hampshire. Here, it was generally agreed that powerful modern computers would soon be able to simulate all aspects of human learning and intelligence. Much of Minsky's later career has been spent testing this claim.

Under Minsky's direction a number of AI programs have been developed at MIT. One of the earliest, a program to solve problems in calculus, showed that most problems could be solved by a careful application of about 100 rules. The computer actually received a grade A in an MIT calculus exam. Other programs developed such topics as reasoning by analogy, handling information expressed in English, and how to catch a bouncing ball with a robotic arm.

But Minsky soon became aware that AI had a number of problems to overcome. For example, in one project a computer with a robotic arm and a TV camera was programmed to copy an assembly of bricks. Although it could quickly recognize the bricks and their relationships to each other, it found the stacking more difficult. It tried to stack the blocks from the top down, releasing brick after brick in midair. Computers simply do not have an innate knowledge of gravity or, he has pointed out, many other things we take for granted, such as that chairs painted a different color remain the same chair, or that boxes must be opened before things can be put inside.

He also noted problems with ‘perceptrons’, designed by Frank Rosenblatt in 1960 with the supposed ability to respond to and recognize certain patterns with the aid of an array of 400 photocells. In collaboration with Seymour Papert, Minsky published a critical account of this work in Perceptrons (1968) showing, in purely formal terms, that the powers of perceptrons were strictly limited. They could not, for example, be relied upon to tell when a figure was connected. A cat's tail protruding from a chair would prevent the perceptron from identifying either the cat or the chair.

In 1974 Minsky introduced the notion of ‘frames’. A frame is a package of knowledge stored in the mind, which allows us to understand many things about a certain topic. For example, the ‘dog frame’ includes what dogs look like, the sorts of things they do, and many other aspects of their nature and behavior. Because we possess numerous such frames we are able to communicate about the world without too much confusion, and to distinguish routinely between the ‘bark’ of a tree and the ‘bark’ of a dog. Only when a computer could be stocked with an enormous number of ‘frames’, some interlocking, others slotted hierarchically in other frames, could it begin to show signs of intelligence.

While speculating about developments in the 1990s Minsky has referred to ‘societies of the mind’. A computer capable of recognizing shadows would be unable to process perspective or parallax. Yet, the untutored human mind can normally handle all three. The aim should therefore be to write a program “that allows each expert system to exploit the body of knowledge that lies buried in the others.”