Part 3

Read the text and answer the questions below.

The long road to artificial intelligence

A

In 1950, the computing pioneer Alan Turing published a paper on the potential mind entitled “Computing Machinery and Intelligence”. It was the first concept of Artificial Intelligence (AI). Turing predicted that by the year 2000 the majority of people would be able to speak of machines thinking without expecting to be contradicted, envisaging a world very different from the one he lived in.

B

Despite Turing’s prediction, human-level artificial intelligence had not been achieved by the year 2000. Nevertheless, one significant milestone in artificial intelligence had recently been attained. In 1997 Deep Blue, a computer developed by IBM, defeated the champion Garry Kasparov. Kasparov had beaten previous computer chess programs, which he had deemed predictable and mechanical, but he already said he was “seeing intelligence” on the other side of the board when he played against Deep Blue.

C

It is instructive to stand back and ponder this moment in the history of AI. The field had accomplished something that half a century beforehand might have been considered an astonishing achievement. Humanity had been outstripped by a machine. Of course, a car can move faster than the fastest human sprinter, and a crane can carry more than a champion weight-lifter. But intellectual prowess is what sets human beings apart from the rest of the animals, and chess is a supremely intellectual pursuit.

D

Yet we consider we are no nearer to human-level AI in Turing’s time. How could that be? The problem with Deep Blue was that it was a specialist. All it could do was play chess. Contrast this with a typical human adult. Take the office worker who had just walked past the window of the café where I am sitting with my laptop. Her day has no doubt been a busy patchwork of activities – making a packed lunch, reviewing the children’s homework, driving work, composing an email, using the photocopier, and so on. Each of these activities required the exercise of multiple concentration skills. Consider the task of making a packed lunch. This involved retrieving utensils and ingredients from various places, opening packets, chopping, cutting, spreading, and so on.

E

In short, a human being is a generalist. A human chess champion can do a whole lot more than just play chess. Moreover a human being is adaptive. Fixing photocopiers is not an innate capability. It is learned. Had the office worker been born in a different country or a different culture, she would have acquired a different set of skills. And if the job machine to lose her present job, she can be trained for another one. The achievements of AI research in a variety of specialist domains (chess being just one among many excellent stories) contrast starkly with the field’s aim: a machine with general purpose, adaptive intelligence. So our basic question is, how could we produce artificial general intelligence?

F

An essential feature of biological intelligence is embodiment. Unlike Deep Blue, a human being is an animal with a body, and its brain is part of that body. The brain of an animal has evolved to prevent any harm from coming to that body and to perpetuate the ability of the body that has evolved, enabling it to survive and reproduce, so that its actions can be made to react to the state of the environment, the better to achieve its goal. The brain shapes the animal’s actions according to what it perceives. Human intelligence, for all its glories, an extension of animal intelligence, and the human capacities for language, reason, and creativity all rest on a evolutionary foundation.

G

So while the endeavour to create artificial general intelligence might succeed without that it is essential to biological life, such as metabolism and reproduction, perhaps embodiment is a methodological necessity. Perhaps the need to interact with the environment is not just a contingent feature of many intelligent systems, but intelligence is at root interactive. The only way to make a subject of the actions of an artifact is to arrange for it to behave in an environment like our own. And one way to achieve human-level AI, according to this way of thinking, is through robotics. Our basic question can then be reformulated: How can we produce a robot with general intelligence?

H

One possibility is that general intelligence is simply the sum of many specialist communicator skills, and up to now, AI hasn’t replicated enough of them. So when robots have been given certain critical sets of skills, general intelligence will emerge somehow. But it is very difficult to make the proposition that happens by itself. The products of specialist intelligence have the advantage of general intelligence. But nobody would be fooled for very long. The multi-specialist robot is going to get stuck as soon as it has to face a problem that is outside its area of expertise, an inevitable occurrence in our ever-changing world.

I

Perhaps the capacity to learn is enough to plug the gaps here. In an unfamiliar situation, a new skill can be learned. Indeed, learning as its various forms is the broad deep into intelligence. But that learning is time-consuming and risky. The biological basis of general intelligence is the ability to adapt its set of behaviour. If intelligence in a robot is one challenge, and to do so without recourse to trial and error or to training by a third party. Only where that can be achieved will artificial intelligence truly match human intelligence.

Questions 27–30

Choose the correct letter, A, B, C or D. Write the correct letter in boxes 27–30 on your answer sheet.

27 When playing chess against the computer Deep Blue, Kasparov felt ☐ had been predictably outwitted, as usual ☐ was able to see into its own thought processes ☐ used powers that were potentially dangerous to humans ☐ was fundamentally different from others he had encountered

28 What is the writer’s main idea in the third paragraph? ☐ In the last 50 years machines have advanced more quickly than expected. ☐ There are certain features of chess that machines cannot replicate. ☐ Skill at chess is more significant than other machine achievements. ☐ Machines will always be more successful at some tasks than at others.

29 The writer describes the case of an office worker to illustrate the point that ☐ everyday human actions involve a complex combination of skills. ☐ some types of mechanical activities require little conscious thought. ☐ some people are able to carry out multiple tasks at the same time. ☐ certain human tasks are more intellectually demanding than others.

30 In the fifth paragraph, the writer claims that ☐ ideas of what success means vary across cultures. ☐ machines are most effective at game-like activities. ☐ machines are better at specialisation than at adaptation. ☐ it may be impossible to isolate different skills involved in an activity.

Questions 31–36

Complete the summary below. Choose ONE WORD ONLY from the passage for each answer. Write your answers in boxes 31–36 on your answer sheet.

Embodiment: an Essential Feature

Unlike machines, animals (including humans) have a body, including a brain which has evolved to provide 31 ________ for that body and allow its genes to be passed on. An animal’s senses provide it with information about its 32 ________ and then through 33 ________ it can react physically to what it senses. The interaction of the body and brain therefore allows the creation of a link between 34 ________ and action. The 35 ________ which make us human and which distinguish us from animals have their 36 ________ in that link between the brain and the body.