Julian Baggini on ... Artificial intelligence: Sometimes the future seems so old-fashioned. The idea of people in silver clothes driving hovercars and eating trays of sludge is now more dated than Rayon flares, prawn vol-au-vents and Babycham.

Sometimes the future seems so old-fashioned. The idea of people in silver clothes driving hovercars and eating trays of sludge is now more dated than Rayon flares, prawn vol-au-vents and Babycham.

The quest for artificial intelligence (AI) is beginning to look just as tired. Last week, a computer called Elbot won the eighteenth Loebner Prize for artificial intelligence. The competition is based on the famous Turing test, named after the brilliant mathematician who set it. He argued that if a machine could converse in a way that was indistinguishable from a human, we would have to say that it was thinking. Elbot failed the test, but got closer to passing it than any of the competition.

To say it got close already shows you just how far the goalposts have had to be moved. Communication is via tapping on a computer keyboard, which is hardly natural human behaviour. To make things even more confusing, it seems Elbot only won because it did a good impression of a human pretending to be a machine. For instance, it replied to the question "How's it going?" with: "I feel terrible today. This morning I poured milk over my breakfast instead of oil, and it rusted before I could eat it."

But the deeper problem is that the Turing test was devised in 1950. Since then we should have learned that we need to go beyond binary distinctions such as thinking/not thinking or conscious/not conscious. Intelligence and consciousness turn out to be more fragmented and messy than we once thought.

Take, for example, the phenomenon of "blindsight". People with damage to the visual cortex of their brain often lose part or all of the field of vision, claiming they cannot see anything in the affected areas. However, if you ask them to guess what is in front of them, they are often able to correctly say what is there. The visual information seems to be getting to part of their brain - just not the part that should actually process what they see.

Or take the example of when you're driving and go on "auto-pilot". You might not be actively aware of what you're doing - but if you do need to stop suddenly or swerve, you almost always do so. How could you do that if you were not in some sense aware of what was going on around you?

Both the blindsight and auto-pilot examples suggest that consciousness is not a simple on/off mechanism. At any one time, at some level, we are aware of all sorts of things, but only a small subset of all this mental activity is at the front of our minds and saved in memory.

There are numerous other examples of how our minds work in bitty ways. People who suffer from prosopagnosia, for example, can identify all the features of a face, but cannot recognise them, even when they are of friends and family. They are not stupid, they just have a cognitive blind spot.

For all these reasons and many more, the question "can a machine think?" is looking increasingly simplistic. The answer always depends. Set the bar low, and you can claim that computers are already intelligent. The American scientist Marvin Minsky, for example, claims that even a thermostat has three beliefs: it's too hot, it's too cold, it's just right. The apparent absurdity of the claim reflects the truth that intelligence and consciousness come in a variety of forms and degrees.

Set the bar high, however, and you can always point to something a computer can't do but a human can, and so deny the machine has a "mind". But since there are lots of things that most conscious humans can do but some can't - like recognise faces or teapots - this seems to be an unfair stipulation.

The Turing test is simple and appealing, but more and more people are coming to realise that it's time we stopped using it as the Holy Grail for AI research. Igor Aleksander, emeritus professor of neural systems engineering at Imperial College London, insists we have to stop thinking that machine consciousness, should it ever arise, would be anything like human consciousness.

More importantly, we need to live with the fact that human consciousness is not uniform or simple anyway. A person is not a neat, unified centre of consciousness. The mind does not steer the self like a rider on a horse. Rather, the self is like a pack of unruly dogs, kept together only by a fragile harness, under no firm control at all.