Part 7 of Turing: a natural philosopher (1997)
The training of thoughtDuring the year, besides training for marathon running to near-Olympic standard, Turing reflected on the 'indications' of mechanical intelligence, writing a report  for the National Physical Laboratory in 1948. Cambridge brought about contact with post-war biology; this and contact with 'cybernetic' thinkers probably reinforced his thesis that there was sufficient scope in the complexity of machines to account for apparently non-mechanical behaviour. But Turing's report quoted nothing from such sources; in fact its most conspicuous citation was from a book by, of all people, the religious novelist Dorothy Sayers, which encapsulated the naive notion of 'mechanical' behaviour. (It was a book he was reading in 1941.) And he argued less from biological theory than from his own life experience, in holding that modification of behaviour could be adapted from learning brain to learning machine.
If the untrained infant's mind is to become an intelligent one, it must acquire both discipline and initiative. So far we have been considering only discipline. To convert a brain or machine into a universal machine is the extremest form of discipline. Without something of this kind one cannot set up proper communication. But discipline is certainly not enough in itself to produce intelligence. That which is required in addition we call initiative. This statement will have to serve as a definition. Our task is to discover the nature of this residue and to try to copy it in machines.
The influence of the general behaviourist climate seemed to blend easily with his own public school background:
The training of the human child depends largely on a system of rewards and punishments, and this suggests that it ought to be possible to carry through the organising with only two interfering inputs, one for 'pleasure' or 'reward' and the other for 'pain' or 'punishment'... Pleasure interference has a tendency to fix the character, i.e. towards preventing it changing, whereas pain stimuli tend to disrupt the character, causing features which had become fixed to change... It is intended that pain stimuli occur when the machine's behaviour is wrong, pleasure stimuli when it is particularly right.
It is often supposed that computers began with heavy arithmetic, and that with this successfully achieved, computer scientists wandered to more ambitious fields. This may be so of others, but is quite untrue of Turing, who had always been concerned with modelling the human mind. (Besides, no computer in the modern sense performed a single addition until 1948.) That he now invaded the behavioural sciences is not in itself surprising; more surprising is that he so vehemently embraced the view that apparently non-mechanical steps of 'initiative' were only hidden mechanism, given his own experience of inspiration, and knowledge of the subtlety of computability. I find it surprising also that he used uncritically, almost gleefully, perversely, a primitive view of education. In his actual childhood experience he had ignored social training as much as possible.
Turing's ideas could not be tried out except on a very small scale as what he called 'paper machines' — the working through of programs by hand. But they anticipated the neural net or connectionist programme of artificial intelligence research, in which what Turing called 'unorganized machines' of sufficient complexity can be trained to perform tasks for which no explicit instructions have ever been written, and where indeed the evolving logical structure is unknown to the human trainer.
His 1948 report, unpublished until 1968, made no impression on the National Physical Laboratory, from which in any case he abruptly resigned. But the ideas resurface, expressed in more general terms, in the famous philosophical paper to which we now turn.
 A. M. Turing. 'Intelligent machinery', National Physical Laboratory report (1948). (See the Bibliography on this site.)
© 1997, Andrew Hodges.