Peter Cave

On 7 June 1954, Alan Turing, aged 41, ate an apple injected with cyanide. Thus died a man who laid the foundations for computer science; who, through his secret cryptological work, significantly contributed to the Allied second world war victory; and who founded the artificial intelligence programme, a source for much of today’s philosophy of psychology. Nobody can be sure of the exact reason for the suicide, but his homosexuality was key. Such sexual activity was, for some, almost de rigueur where Turing studied – King’s College, Cambridge, whose Keynes had earlier promoted the ethics of higher sodomy. In the wider world, though, homosexuals received public vilification, predictions of eternal damnation, and even earthly imprisonment; and in 1952, a Turing sexual affair was exposed to the police. Turing was dismissed from his cryptological post and – to avoid imprisonment – he submitted to chemical treatment to submerge his sexuality.

Turing read mathematics at Cambridge, finding King’s life congenial, with its progressive intellectualism and libertarianism. He turned specifically to mathematical logic and computability. After his wartime success in breaking the Enigma code, he moved to Manchester, pioneering further computer development.

What has captivated many philosophers is Turing’s belief that, one day, we will have engineered machines that think. What counts as engineering here? The answer "when by mechanical means" is no help: sexual copulation is often mechanical, yet churns out people.

Investigating ideas of effective procedures, Turing conceived of "Turing machines" – hypothetical abstract devices, with infinite storage, but finite rules or algorithms which determine, given a prior state and input, subsequent states and output. Manifestations of Turing machines, therefore, would behave predictably. Although (physical) computers these days look pretty different from machinery with cogs and pistons, they follow determinate procedures. Computers, like humans, lack infinite memory storage, of course; but they operate according to finite algorithms and can mimic.

If computers can be constructed to think as intelligently as humans, then it is possible that our thinking operates mechanically and that we are part of the physical, causal world. It remains a practical matter about what stuff can realise the algorithms. Silicone chips can do so; biological brains too – but maybe they do much more in human thinking? After all, machine intelligence seems contradictory, for we contrast mechanical with intelligent behaviour. To deal with this, Turing devised a discriminatory test – The Turing Test.

An interrogator, receiving only printed answers, interrogates two hidden individuals, one intelligent human, one machine. If he cannot tell which is the machine, then the machine is intelligent. Of course, the interrogator needs intelligence himself. Asking simple arithmetical questions alone, for example, would lead to no difference being identified between persons and pocket calculators with suitable interfaces. Further, the machines need to be intelligently programmed to avoid tell-tale signs of their non-human nature: so, they can, for example, delay answers to complex questions, print out hesitant "er"s and even make mistakes. Turing anticipates that, once we grow accustomed to intelligent machines, their programmes could give quirky outputs, leading us to note, "My machine said such a funny thing yesterday."

Even if machines pass Turing’s test (some have, but were their interrogators sufficiently discerning?) they might be merely simulating or modelling. After all, computers can model weather changes, but there’s no danger of computers raining. In contrast, simulating steps of reasoning to right answers, with right answers given – as also, sometimes, acting sex scenes in films – amounts to the real thing. So, are machines that pass Turing’s test akin to weather modelling or to realistic sex acting?

The Turing Test hides seeming irrelevancies, such as whether the "thinker" is metallic or flesh; but it hides much more. Our thinking engages a fleshy life – riding bicycles, coaxing lovers who have headaches, using chairs as coat-hangers. Simply knowing that something or other, about bicycles, coat-hangers and lovers, captures neither our thoughts’ richness concerning these items, nor our knowing how to use them, nor how to judge what is relevant to say about them. Computation, however, is symbol manipulation – symbols that we interpret as meaningful. Can we even grasp how sheer computational complexity could give rise to interpreting symbols – to understanding their meaning?

Turing’s ideas continue to stimulate research into mentality as nothing but functional relationships between inputs and outputs, with the mind akin to software, not hardware. This gives rise to controversial and fertile thought experiments – from replications of oneself to tales of Chinese rooms and of brains consisting of the Chinese nation. Turing, with his yen for the provocative, radical and comic, would surely have approved.

Suggested reading

The Philosophy of Artificial Intelligence, ed M Boden (Oxford University Press)
Alan Turing: The Enigma, A Hodges (Vintage)
Codebreakers, Eds F H Hinsley and A Stripp (Oxford University Press)

Peter Cave is an associate lecturer at the Open University