LECTURE

ROBOTS:THEY CAN'T REALLY BE INTELLIGENT, CAN THEY?

A Joint Lecture with the British Association for the Advancement of Science by Professor Kevin Warwick on 12 May 1998

Professor Warwick of the Department of Cybernetics at the University of Reading is a leading international expert on the robots of the present and future. His recent book “March of the Machines” generated a lively debate.
Why are humans in control of so many things on Earth? They are not more powerful or faster than other creatures, but they are more intelligent. From the
evolutionary point of view this is a short-period occurrence; although it is unlikely that another animal will surpass us, what about the machines we make?
Let us assume that a machine could be more intelligent than a human; then machines would be in control. How would they treat humans? - as we treat animals now, on farms, in zoos, hunted, culled, slaughtered if they become diseased or domesticated as pets or slaves. So if machines can be more intelligent we have a problem: is it possible?
What is intelligence? Slugs and snails have 9 - 200 neuron brain cells; bees and wasps, 10,000; dogs and cats, 10 million; humans, 100 billion. A typical Personal Computer now has around the same brain power as a bee; in four years it may match a cat; in nine years a human; in 20 - 30 years 10,000 times that of a human. Animals take centuries to evolve; machines take minutes using computer-aided design and manufacture.
Robots can now clean lavatories, play soccer, walk on two legs upstairs or pushing a truck, drive a car, play from sheet music with individual differences in performance on different occasions. Are they intelligent? Do you consider a child who plays music, intelligent?
Current robots are programmed by humans but humans are programmed by their genes, for example, to breath and develop. Both humans and machines (if allowed) can learn and adapt. At Reading they study how machines learn using a number of robots which can communicate. Recently, one of them by itself programmed another in New York by way of the Internet. These have 40 - 50 ‘brain cells’,equivalent to a slug. With 500 cells they could ‘see’.
Deep Blue, the IBM chess computer, beat Kasparov, the World Champion, in two games out of three. Chess involves strategy; so does warfare and financial dealing. These are the fields in which computers are being used to learn how to improve their performance. They may now do only one thing well - play chess or music or clean lavatories - but they can be made bigger and able to ‘walk and chew gum at the same time’! They do not have to think as humans think, just to 'think' faster or better to beat us. The military already find that robotic planes without pilots can win over planes with pilots.
As long as a human can switch off a machine the human is in charge, but we are approaching the stage where the machine will be too intelligent to let us switch it off! (Would you let a cow switch you off?)
We are quickly getting into a very dangerous situation; we have less than ten years before it becomes a desperate problem.In the discussion it was asked: are we stupid enough to make intelligent machines? Probably, because short-term advantage in finance or war drives the development without consideration of the long-term consequences. Political international agreement is necessary to prevent it - is that likely?
Don Lovell (from tape)