|
|
Dave Walsh assesses the progress of reasearch into 'artificial intelligence' (AI), chess computers and conversation machines and finds that reality falls far short of public expectation. According to the Bible, 'God created man in his own image', but mythology, science and literature have demonstrated mankind's striving to create other beings in its own image. In Greek legend, Prometheus created an army from clay with fire stolen from Zeus; Rabbi Löw created a golem (clay automaton) in the Jewish ghetto of Prague; and Mary Shelley's Frankenstein, the 'Modern Prometheus', assembled his creature from divers human remains.
Today, artificial intelligences (AIs) both humanoid and non-humanoid in form are common characters in film, television and innumerable science-fiction novels; and often stereotyped as godlike entities, totalitarian tyrants or, less frequently, as helpful, witty friends of man see 'Artificial Pals' panel. These familiar depictions have fixed certain ideas about 'artificial intelligence' in popular culture: for example the common technique of accompanying the machine's polite yet patronising monotonic voice with images of a particular part of it, or its console or even, as in the case of Red Dwarf's 'Holly', a human face. Arguably, two of the best-known fictional accounts of AI do not involve conventional ideas about computers. In the case of the movie Blade Runner based on the Philip K Dick novel, Do Androids Dream of Electric Sheep Tyrell creates a race of artificial humans (androids) called 'replicants', who are genetically engineered to have a short lifespan as slave labour. They are made by man in his own image, so to speak, and, mirroring the Bible, turn their back on their creator. With some irnoy, Tyrell's motto is: "More human than human." The other account is Idoru, the novel by William Gibson, whose eponymous character is an 'intelligent software agent' who exists only in the electronic realm of the 'net matrix' and is holographically represented in the 'real' world as a beautiful Japanese girl. Gibson's other hero a Chinese-Irish rock star announces his plan to marry the Idoru, a proposal which tips his entourage and fans into a state of chaos.
Curiously, last year, a Japanese record company created a computer generated female pop star called 'Kyoko Date' So how do these mass-culture interpretations match the reality of artificial intelligence research? What does the term really mean? According to the Concise Oxford Dictionary of Current English, 'artificial intelligence' is "the theory and development of computer systems able to perform tasks normally requiring human intelligence such as decision making and speech recognition." The precise definition is being disputed while others ask: "How can we hope to create artificial intelligence when we don't fully understand the structure of our own brains and the nature of our own intelligence?" As Professor Marvin Minsky of Massachusett's Institute of Technology (MIT) puts it and not, one suspects, without a sense of irony "artificial intelligence is the science of making machines do things that would require intelligence if done by men". In 1950, British mathematician Alan Turing began exploring the question of whether a machine could think. He deduced that it could and that a more meaningful question was: "How could we tell?" Turing concluded that "if the responses from the computer were indistinguishable from that of a human, the computer could be said to be thinking." Turing devised a test, in which the interrogator's task is to decide whether or not his correspondent is human by gauging the level of 'realism' in their half of the conversation. If the interrogator can reach no conclusion, then the computer can still be said to be 'intelligent'.
Much criticism has been directed towards this test, as it takes no real account of conversational skills, typing ability and cultural background so that there is the possibility that a human subject could be mistaken for a computer. In addition, the test requires the computer's responses to be compared with those of a 'control group' and the conversations are a kind of 'blind date' (as the 'opponent' cannot be seen). This has lead some to question why we should bother testing a machine for intelligence when it can only communicate through a terminal. Which brings us back, again, to the definition of 'intelligence'. Many people involved in researching the subject see its goal as the creation of an exact replica of living intelligence in a non-organic entity. So far, this goal has not been attained; many believe that it never will. As Jorn Barger so eloquently puts it: "It's been more than 40 years since the term 'Artificial Intelligence' was coined in 1956. During that time, AI has pretty successfully maintained an aura of arcane impenetrability, funnelling off billions of dollars in research grants... while delivering almost nothing that can really be called 'intelligent'!"