War Games and Mind Games
As described on an earlier Scrapbook page, Alan Turing was based at Bletchley Park, Buckinghamshire, while acting as the leading cryptanalyst of German ciphers during the Second World War.
It was during this period that he formulated the ideas that emerged after the war as Intelligent Machinery, and would now be called Artificial Intelligence. He must have been influenced by the astonishing power of mechanical methods at Bletchley Park.
The codebreaking work at Bletchley Park was highly secret and discussed only with those directly involved; but Turing used game-playing, particularly chess-playing, as a close analogy.
Chess and Go
Alan Turing talked at Bletchley Park with his younger colleague Jack Good about what we would now call chess-playing programs. They got the idea of searching decision trees for the best move.
Turing knew the game of Go as early as 1936: in fact he explained it to Christopher Morcom's mother shortly before he left for America. At Bletchley Park he showed the game to others including Jack Good. According to the British Go Association's page it was Good who twenty years later popularised the game in Britain.
|In the later years of the war, he talked with a young student, Donald Michie, about the prospect of machines 'learning.' In other conversations he talked about 'building a brain.' It seems that it was during the war that he came to the view that anything done by the brain must be a computable operation, and so could be simulated by a Turing machine. |
I. J. (Jack) Good, who also worked for GCHQ after the war, became a prominent mathematical statistician. |
Donald Michie (1923-2007) later became a British pioneer in Artificial Intelligence.
Turing's post-war ideas reflected the discussions he had enjoyed during the wartime period. After 1945 he often used chess-playing as an example of what a computer could do, and in his 1946 report on the possibilities of a computer, made his first reference to machine 'intelligence' in connection with chess-playing. In 1948 he met Donald Michie again and competed with him in writing a simple chess-playing algorithm.
This short history of computer chess begins with Turing, and there a longer history with detail about Turing's chess programming on chessbase.com.
Turing included an example of his chess program in action, when he wrote a discussion of computer chess in the 1953 book Faster than Thought. It is reproduced on this page from Der Spiegel OnLine, in conjunction with a feature article on Turing's work at Bletchley Park and the beginning of AI and ChessBase.
Turing and neural networks
Turing gave some important comments on the prospects for machine intelligence in a talk in 1947 but his most constructive proposals were put in an unpublished report for the National Physical Laboratory that he wrote in 1947-8.
One of his ideas was to define logical systems modelled on interconnected neurons. He considered starting with an initially unorganised system, then training it or letting it evolve into some form of organisation. He imagined simulating this process on a computer.
At the time he wrote this, no computer had actually been built. Rather surprisingly, he does not seem to have followed up this idea even when he was able to use the Manchester computer.
| Craig Webster and William Fleming in New Zealand have studied Turing's structures and programmed them on a modern computer. |
Below is the outline of a net, based on Turing's ideas, which has evolved so as to perform the addition with carry of two binary digits, or equivalently, to compute both the XOR and the AND of two bits.
|0 0||0 0|
|0 1||0 1|
|1 0||0 1|
|1 1||1 0|
This is described in more detail on Craig Webster's page, which is part of a
short report on Turing's ideas.
A parallel project by Christof Teuscher at the Swiss Federal Insitute of Technology, Lausanne, has now come to publication in a book Turing's Connectionism, supported by a website.
Theory and Practice of Intelligent Machinery
The general idea that machines should be able to learn, modifying their behaviour with experience, was a vital part of Turing's thought on machine intelligence. It features strongly in Turing's most famous paper Computing Machinery and Intelligence, published in 1950. This is the subject of another Scrapbook page.
But to try any of this out in practice, Alan Turing had to build a practical version of his Universal Machine — and for this it was necessary to have electronic speeds. The last two years of the Second World War gave him just the experience with electronic technology that he needed.
Continue to the next Scrapbook Page to see the
practical side of 'building a brain.'