Artificial Intelligence (Part II)

Task 1. Read the text and check the difference between the information in the I and II Parts of the text.

1. There is interest in the results of AI from at least four directions. In particular there is the study of robotics which is concerned, to a large extend, with the practical requirements of industry for mechanical devices which can perform «intelligent» tasks – tasks of a versatility and complication which have previously demanded human intervention or control – and to perform them with a speed and reliability beyond any human capabilities, or under adverse conditions where human life could be at risk. Also of interest commercially, as well as generally, is the development of expert systems, according to which the essential knowledge of an entire profession – medical, legal, etc. – is intended to be coded into a computer package! Is it possible that the experience and expertise of human members of these professions might actually be supplanted by such packages? The question of whether the computers can simulate genuine intelligence clearly has considerable social implications. Another area in which AI could have direct relevance is psychology. It is hoped that by trying to imitate the behaviour of a human brain (or that of some other animal) by means of an electronic device – or by failing to do so – one may learn something of importance concerning the brain's workings. Finally, there is the optimistic hope that for similar reasons AI might have something to say about deep questions of philosophy.

2. How far has AI been able to progress to date? It would be hard to try to summarize. There are many active groups in different parts of the world. Nevertheless, it would be fair to say that, although many clever things have indeed been done, the simulation of anything that could pass for genuine intelligence is yet a long way off. But to give an optimistic example it should be mentioned some of the impressive achievements, and then some remarkable recent progress with chess computers.

3 One of the first AI devices was W. Grey Walter's «tortoise», made in the early 1950s, which moved around the floor under its own power until its batteries ran low. A somewhat separate line of development is exemplified by the computer program, designed in 1972, which can converse sensibly about a collection of blocks, of various shapes and colours which, in simulation, it places on top of one another in different orders and arrangements. Despite such early successes, it is remarkable how difficult it proved to design the control of even a simple, jointed «robot arm», which has to manoeuvre around, avoiding crowded obstacles, where the required motions may seem «obvious» to a human looking on. The Interpretation of visual scenes generally is also an area where the procedures of Al have so far got nowhere close to what the human brain (and, indeed, most other animals) seems «effortlessly» to be able to achieve.

4. Perhaps more close to the human intelligence is the computer program developed in the mid-1960s, which simulates a psychotherapist – so successfully that some patients find that they prefer the computer to a human therapist, and are prepared to reveal more of themselves to it than they are to its human counterparts. Though this may give an impression that the computer has some understanding, in fact it has none, and is merely following some fairly simple mechanical rules.

5. Chess-playing computers probably provide the best examples of machines exhibiting what might be thought of as «intelligent behavior». In fact, some machines have now reached an extremely respectable level of performance in relation to human players approaching that of «International Master». (These computers' ratings would be a little below 2300, where, for comparison, Kasparov, the world champion, has a rating greater than 2700). In particular, a computer program by Dan and Kathe Spracklen has achieved a rating of 2110 and has now been awarded the «Master» title. Even more impressive is «Deep Thought» which has a rating of about 2500 and recently achieved the remarkable feat of sharing first prize (with Grandmaster Tony Miles) in a chess tournament (in California, 1988) actually defeating a Grandmaster (Bent Larsen) for the first time! Chess computers now also excel at solving chess problems, and can easily outstrip humans at this endeavour.

6. Chess-playing machines rely a lot on «book knowledge» in addition to accurate calculational power. It is worth remarking that chess-playing machines far better on the whole, relative to a comparable human player, when it is required that the moves are made very quickly; the human players perform relatively better in relation to the machines when a good measure of time is allowed for each move. The computer's decisions are made on the basis of precise and rapid extended computations, whereas the human player takes advantage of «judgments», that rely upon comparatively slow conscious assessments. These human judgments serve to cut down drastically the number of serious possibilities that need be considered at each stage of calculation, and much greater depth can be achieved in the analysis, when the time is available, than in the machine's simply calculating and directly eliminating possibilities, without using such judgments.

Task 2. Define the number of blocks where you can find information about:

a) the program simulating a psychotherapist;

b) the results of active groups working on AI;

c) the differences in chess-playing strategy between chess-playing computers and humans;

d) the main directions that are of interest as to the results of AI development;

e) the impressive success with chess-playing machines;

f) the attempts to construct the first robot;

g) the score of the last chess match “Computer-Kasparov”.