COPElAND & PRINz | 329
a program that could be used to calculate an unambiguous move for any given chess position. He
had to act as human CPU and calculate the moves with pencil and paper—which you can guess
was a very laborious process and took quite some time. . . .
Computer chess before computers
The charming (if irascible) AI pioneer Donald Michie was just a schoolboy when he joined
the codebreakers at Bletchley Park in 1942. Fifty-five years later, after a lifetime of building
intelligent (and not-so-intelligent) machines, Michie was one of those present when DeepBlue
defeated Kasparov in New York in 1997. For the first time in history the human intellect seemed
on the run, with AI encroaching on what Goethe had famously called ‘a touchstone of human
intelligence’.^3 Michie, always an admirer of brainpower, had soon befriended Turing at Bletchley
Park. In 1998, relaxing at his home in Palm Desert, California, he told me (Jack Copeland) how
computer chess was born.^4
On Friday evenings, after a hard week battling against German codes, he and Turing used to
go for a drink together. Their usual pub was in the bustling little town of Wolverton, just a few
minutes away on the train from Bletchley Station. Finding seats in the poorly lit and smoke-
filled bar, they would begin a game of chess and turn to their favourite topic of conversation—
the mechanization of human thought processes. Michie explained:
What the codebreaker does is very much a set of intellectual operations and thought processes,
and so we were thoroughly familiar with the idea of automating thought processes—both of us
were up to our elbows in automation of one kind and another.
He was referring to Heath Robinson (see Chapter 14)—his own preoccupation at that time—
and Turing’s bombes (although Michie himself knew nothing about the bombes until long after
the end of the war). Given what the two were involved in at Bletchley Park, they found it entirely
natural to be spending their Friday evenings talking about automating the thought processes of
a human chess player—and from there Turing moved on to the idea of automating the whole
process of learning. Jack Good, another codebreaker and chess fanatic, would sometimes join
in the discussions, usually during Sunday morning walks with Turing and Michie. Good could
also remember an earlier conversation with Turing in 1941, before Michie arrived at Bletchley
Park, in which they ‘talked about the possibility of mechanizing chess’.^5
I asked Michie what he could recall of these historic discussions with Turing about what we
now call AI:
Michie : ‘There were three headings. One was: methods of mechanizing the game of chess
and games of similar structure. Another was: the possibility of machine algorithms and
systems which could learn from experience; and the third area was a little more general,
to do with the possibility of instructing machines with more general statements than
purely ground-level factual statements—which would involve, in some sense, the machine
understanding and drawing inferences from what it was told.’
Copeland : ‘Did you and Turing often play chess together at this time?’
Michie : ‘Being one of the few people in the Bletchley environment bad enough to give him a
reasonably even game, I became his regular sparring partner. Our discussions on machine
intelligence started from the moment that we began to play chess together.’
Copeland : ‘What specific proposals did Turing make concerning chess programming at this time?’