DONALD MICHIE ON MACHINE INTELLIGENCE, BIOLOGY, AND MORE
edited by Ashwin Srinivasan.
321 pp. Oxford University Press. 25 pounds.
978 0 19957 304 2
From the early 1960's until his death in 2007, Donald Michie was one of the leading British researchers in the area of artificial intelligence (AI). In 1963, he founded the AI research group at the University of Edinburgh; he worked there until 1985, then left to found the Turing Institute in Glasgow. He did groundbreaking research in automated planning, problem solving, and robotics. Perhaps his best-known accomplishments were the application (1968) of automated reinforcement learning techniques to the robotic problem of balancing an inverted pole on a movable cart; and the FREDDY robots (1973) which combined perception, tactile sensing, manipulation, and planning to do automated assembly of complex objects.
Earlier, in the 1950's he did seminal research in biology, together with his second wife Anne McLaren. He studied the genetics of mice, and wrote an influential paper arguing, contrary to the common wisdom of the time, that cross-bred mice made better subjects than inbred mice. He and McLaren also studied fertilization and transplantation; their research laid the basis for important techniques in "in vitro" fertilization. Before that, during the Second World War, he was part of the secret team at Bletchley Park, under the leadership of Alan Turing, that use the Colossus computers to break the German Enigma code.
This book is a fitting memorial tribute to this remarkable scientist. Mostly, it is a collection of his own writings, which reflect his wide range of interests: retrospective essays on his work at Bletchley; essays evaluating the progress and the future promise of AI and arguing for better funding for AI in the UK (though none of his technical AI papers); excerpts from his biological research papers; some of his essays and columns on science and society; and finally a long critique of the British involvement in the Iraq war, written in 2003. Michie was politically engaged all his life; he joined the Communist Party during the cold war and regularly attended political congresses and demonstrations. He was also the science correspondent for the Daily Worker for many years.
In addition to Michie's writings, there are transcripts of two interviews, and, in the foreword, four memorial essays: The most interesting of these is by the late, great statistician I.J. Good, who recalls his days working with Michie at Bletchley Park.
Although Michie, in one of his essays, argues that research papers should be less technical and more entertaining, his own writing tends to be dry, and many of the essays deal with controversies long since settled. But the book is certainly a valuable resource for the histories of Colossus, of mid-century genetics research, of AI research, and of science and public policy in postwar Britain.
Ernest Davis is Professor of Computer Science at New York University. He is the author of "Representations of Commonsense Knowledge" (1990).