The value of Artificial Stupidity
Rupert Wegerif puts computers in their place.
In the fevered imaginations of product marketers, computers and intelligence seem indissolubly linked.
For example, toys for toddlers with computer chips in them all seem to have names such as `IQ builder', `L'il Genius' or `Brain Booster'. On examination these toys do little more than make a certain noise, or flash a light, when the child presses a particular button. There seems little that is intelligent about such simple causal connections. On the whole the more intelligent an animal or a person is, the less predictable are the outputs that will result from repeating the same inputs. But there is a sense in which simple interactive toys are illustrative of computing in general. Causal rules linking inputs to outputs in a regular way are essentially all that computers do. And, while I would say that this makes them pretty dumb, there are those who argue that computers can be made to be `intelligent'.
The British mathematician, World War II code breaker and father of computing, Alan Turing, was one of these optimists and he gave his name to the test of artificial intelligence (AI) that is still applied today. Put crudely, the 'Turing test' claims that artificial intelligence is achieved when a person having a conversation with a computer cannot tell that the latter is really a machine and not a human being. While, by all accounts, Turing was a genius, the Turing test itself may not have been one of his better ideas. In 1966 Joseph Weizenbaum wrote a little interactive programme called ELIZA as a kind of subversive joke to demonstrate that the 'Turing test' was more likely to be met by Artificial Stupidity (AS) than by anything that could reasonably be called intelligence. ELIZA imitates the way in which psycho-therapists will take the words of their clients, twist them a little, and then feed them back in order to get clients to think more deeply about their own problems. So, for example, if the client makes a statement ELIZA will take the key word or phrase, repeat it and say 'that sounds interesting, tell me more about it'. If the client asks ELIZA a question the response may be 'Why do you ask me that?' or 'We are talking about you, not me'. ELIZA was created using crude pattern-matching without any pretence to intelligence. To respond to the presence of the word 'mother' with the letter string 'tell me more about your mother' as ELIZA does, is simply the equivalent to playing a 'moo' sound when the picture of a cow is pressed on a children's toy. Yet the programme worked. Intelligent, educated people who were well aware that ELIZA was just a bit of software nonetheless spent long periods working with it on their personal problems. A commercial version was quickly produced and various descendants of this original are now widely employed for therapeutic purposes all over the world. Proponents of artificial intelligence still try to simulate intelligence and some of them argue that their artificially- intelligent software is, or will be, the best way to teach thinking. The main idea behind this claim is that it is easier for an artificial intelligence, a set of programmed rules, to articulate and communicate the process of effective thinking to students than it is for human teachers. Weizenbaum's ELIZA cut through the nonsense and mystique that surrounds ideas such as 'AI' and the 'Turing test.' Computers are just dumb machines all the way through and that is that. AI does not exist. However ELIZA also suggests a constructive way forward in the use of computers to support the teaching of thinking. ELIZA prompted, reflected, probed and challenged everything that was given to it, forcing the exercise of intelligence back onto its human interlocutors. It helped people to think about their personal problems precisely because it had no competing intelligence of its own. The useful message of this experiment is that, rather than trying to design computers that can think for us, it might be better to design them to more effectively support us in thinking for ourselves. Rupert Wegerif