Father of artificial intelligence in Britain.
Donald Michie was a man of many parts, irrepressible energy and great personal charisma. His first significant scientific contribution came in the field of genetics. With his second wife Anne McLaren, tragically killed in the same car accident in which he died, he published research in the 1950s that helped lay the groundwork for modern reproductive technology, and later earned him a Pioneer Award, with McLaren, from the International Embryo Transfer Society. But even while developing those ideas, he became inspired by a very different scientific passion: intelligent machines.
He had first encountered this revolutionary idea in the company of Alan Turing at Britain's code-breaking centre, Bletchley Park, during the Second World War. The two did not cooperate closely, for whereas Turing was busy breaking the Enigma codes, Michie focused on using the Colossus computer to crack messages generated by the Germans' 'Tunny' teleprinter machines. They became friends nonetheless, and regularly played chess. They often discussed the possibility of intelligent machines that could play in their stead — and do mathematics, use language, interpret photographic input, learn, and even (Turing suggested) wander around the countryside unaided.
This shared vision of artificial intelligence (AI) became Michie's guiding principle, and by 1948 he was writing a paper-and-pencil chess program in his spare time. As he said years later, recalling how Turing's speculations had gripped him, “I resolved to make artificial intelligence my life as soon as it became feasible”. He himself helped to do that, not only intellectually, but also commercially, writing the first marketable 'expert-system shell' for logic programming in the 1960s. In Britain, he founded AI virtually single-handed.
His first task was to overcome what he later called “the national computer-blindness”. Even Britain's science minister in the early 1960s knew nothing of the wartime code-busting efforts, and thought that 'computing' meant desk calculators.
During his relentless lobbying, Michie persuaded the Royal Society to provide “a few hundred pounds” to enable him, with Bernard Meltzer, to set up a small AI research group at the University of Edinburgh in 1963. Its existence was made official as the Experimental Programming Unit in 1965, and in 1967 it became the Department of Machine Intelligence and Perception. Right from the start, the department received frequent visits from AI pioneers in the United States based at the Massachusetts Institute of Technology, Carnegie Mellon University and Stanford University. Its successor is still a leading centre in the field.
Those early years were not free of problems. The initial offer of a deconsecrated church as lab space was withdrawn when Michie's intention to build an intelligent robot became known. But he remained indefatigable, and in the ensuing years fought lustily for the infant discipline, confronting academia, research councils and industry. He also did much to raise the profile of AI with the general public, holding popular lectures and penning one-off articles for the press, as well as a regular column for the trade magazine Computer Weekly.
Sometimes his optimism and enthusiasm, for instance in describing the achievements and potential of the Edinburgh robot, went too far, and prompted a backlash from sceptics and rivals. One result was the UK Science Research Council's notorious Lighthill report of 1973, which in effect pronounced research in AI to be a waste of time and money. Its criticisms were so fierce as to cause a scandal. It was the only 'internal' research-council report to be published, together with extensive rebuttals — including nine pages from Michie.
The Lighthill report stalled AI research in Britain for a decade. Morale and funding reached a low ebb, with several prominent researchers fleeing to the United States. Michie was also sidelined at Edinburgh: a new Department of Artificial Intelligence was formed and inherited most of its predecessor's resources, whereas Michie was put in charge of an independent Machine Intelligence Research Unit, and forbidden to work on robotics.
Research in AI was officially rehabilitated ten years later, when in 1982 Japan launched its 'Fifth Generation project', a huge financial and industrial commitment to base its future economy on AI, predicting world supremacy as a result. The response, in Britain as in the United States, was an injection of government funds into both military and civilian AI research. Michie's urbane comment was that the Lighthill incident had been a “mishap of scientific politics” due to all-too-human frailties — specifically, “nothing but ignorance at the top”.
Michie's contributions to the theory of AI began in the 1960s, when a colleague had insisted that learning machines were impossible, and had challenged Michie to prove him wrong. Edinburgh still lacked a digital computer, and so Michie built “a contraption of matchboxes and glass beads” to master the art of noughts and crosses (tic-tac-toe). He called it the Matchbox Educable Noughts And Crosses Engine, or MENACE. He followed that in 1968 with the first reinforcement-learning program, the 'pole balancer', involving a pole balancing on a cart. This was no simulation, but controlled a real pole balanced on a real cart. In the 1970s, he capped this with research into chess endgames.
Michie's pioneering Graph Traverser of the 1960s provided ideas that are now standard in heuristic search algorithms, and live on in widely used AI planning techniques. His last important contribution was his StatLog project of the early 1990s, a highly insightful comparison of various models for learning algorithms, from statistical approaches through symbolic tree-building to dynamical systems.
From his long career, Donald Michie has left a generous academic legacy: the establishment of AI in Britain. That was a product of his mercurial intelligence, but those who knew him personally will also remember his wide learning, his wit and his charm.