Artificial Intelligence: Can machines think?
An article from the 1971 archive
Words by Peter Martin
This was the topic of a seminar held last term (Wednesday August 2) at Flinders University. The speakers were Chris Illert, a Flinders graduate who owns a mini-computer he is training to ‘think’, Dr Jim Crowley, Psychology lecturer, and Dr. Mike Georgeff, Maths Lecturer. The seminar was put on by the Creative Computing Club and the Psychology Association.
I’ve just spent two nights going over the tape recording of the discussion and have decided that it’s too good to keep to just the sixty people who attended the seminar. My own understanding of the topic isn’t too great (I’ve only done Computing I and II) and so I apologise in advance for the inaccurate transcribing and incorrect interpretation of arguments that you are about to read.
The first speaker was Dr. Mike Georgeff, who lectures in Computing I and was asked to ‘lend some respectability’ to the seminar. Mike outlined the “Turing test” for determining whether or not machines can think. In this test an interrogator communicates with a person and a machine via a keyboard, and has to determine which is which simply by asking questions without asking for any demonstration of physical skills. Turing claimed that if the interrogator could not determine which was a person and which was a machine then it seemed reasonable to say that the machine could think.
Next, Dr. Jim Crowley, whose real field is prose comprehension, gave a very detailed account of the ways in which psychologists can use computer programmes to understand human behaviour. He was generally sceptical about the ability of machines to think and pointed to particular problems in the areas of pattern perception and understanding language. He pointed out that we don’t yet understand how humans think and so can’t be expected to transfer our thought processes to computer.
Then the seminar began in earnest. Chris Illert, who is a self styled expert in such diverse fields as Marine Biology, Astronomy, and Mathematics, came prepared to mount a passionate defence of machines, and an attack on what he considers the unjustified smugness of humans. To provide evidence for his case he brought along his own mini computer which talked to members of the audience before the seminar began. Here is some of what Chris Illert had to say:
“I THINK PEOPLE ARE VERY INSECURE ABOUT COMPUTERS”
“…A century ago it was argued, in reasonably educated circles, that black people didn’t have souls. They were only suitable for slave labour. Within our own species something as trivial as skin colour is enough to cause people to shelve a group. How much more prejudiced are we going to be against a machine which is not even biological? People today still argue stupid questions such as ‘does a dog or a cat have a soul?’, ‘Is it really alive?’ ‘Obviously human beings are the only ones that have souls.’ Those sort of arguments deserve to die out.
“…We are terribly prejudiced and we tend to use ourselves as a measuring stick. I’m not sure that human beings are a fit model to model things as beautiful as machines on.
“JUST BECAUSE WE KNOW HOW MACHINES WORK, DOESN’T MAKE THEM ANY LESS INTELLIGENT”
“…If you were to cut a little bit of meat off somebody and study it, you would see nerves and bio-chemical reactions, but that couldn’t give you any idea at all of the overall complexity of the person, the type of organisation involved and the capabilities of the person. It’s stupid to have a muscle on a plate and give it electric shocks like Volta did. You’re studying a muscle, but you’re not studying a Human Being. And the same goes for Computers. People start talking about logic circuits, I.C.’s and how the guts of the thing works, but this gives no indication of what the components can do when they’re put together. So many times I’ve written beautiful programmes, and people have looked at the results and been stunned. Then I’ve shown them how it was done and the people have said ‘cheap trick’. But the point is that it fooled them. Just because you understand how something works doesn’t make it less valuable. That’s something that very few people are able to understand. Because you understand something it doesn’t cheapen it. The fact that you can understand it says something for you, but it doesn’t make that thing less important or less good. It just makes you a little bit more conceited.
“WE ARE ALL PROGRAMMED”
“I’ve heard it said that computers only do things they are programmed. Of course they’re programmed, so are we. From the time we’re born we wander around and bump our heads on things, listen to sounds, get billions of pieces of input each second… We are programmed from the time we are born. That’s what schools are for.
“IF WE HAD THE LIMITED INPUTS OF COMPUTERS WE MIGHTN’T DO AS WELL”
“All a computer has for programming is a keyboard, with people typing nonsense on to it. If you put a kid in a dark room from the moment it was born and gave it nonsense on a keyboard, how much do you think it would learn? Better still take a human brain out and put it in a vat of pink fluid to keep it alive, give it a display panel, a keyboard, and type in nonsense. I don’t think that it would cope as effectively as a computer. If you cut off most of a person’s inputs, you don’t get a very intelligent person and I think that if you did it young enough, the person would have an intelligence comparable with a machine.
“It might be a different situation if you gave computers sensory inputs comparable to our own. If you could give a computer eyes and ears and let it walk around, it would have so much information that it would no longer be programmed by a silly human, but would be programmed by the environment in the same way that we are.
HUMAN BRAINS “AREN’T AS EFFICIENT”
“How many people here could multiply pi by three to a million decimal places? We can’t do it, yet we have a bigger brain. Why can’t you do it? You’ve got a bigger brain and you can’t do it. There’s something wrong with you.”
Interjection: “It’s called flexibility”
“No, its called inefficiency. We don’t have enough control mechanisms to manipulate the vast storage we have in the human brain. We have a mass of biological circuits and we cannot control them. The human brain is very very inefficient.
“OUR BRAINS HAVE MASSIVE STORAGE AND NO CONTROL MECHANISMS”
“The computer is the exact opposite. It has wonderful control mechanisms and no storage. This will change. In the next five years you will see machines with very very big storage and the fantastic control mechanisms that we have already. When this happens, I doubt whether another device, or brain, which is of the same order of magnitude will be able to really understand them fully.
“We remember everything. I remember that I walked in, stubbed my toe on the door, useless information, silly things… Think of all the rubbish that goes into your brain, through your ears and through your eyes, all of the input that is stored.
“A computer at the moment cannot store all that information. It knows what to do with it, but it can’t store it. That’s really the only difference that I see between a human being and an intelligent decision making machine capable of value judgments.”
“AFTER A CERTAIN STAGE YOU HAVE TO ADMIT THAT THE PROGRAMME IS OUT OF CONTROL”
“My chess programme is out of control. It programmes itself. It started off when I gave some simple ideas about how to play chess, and advised it that certain situations would happen. But the computer, after awhile playing against people, and people are the environment for it, decided that it wouldn’t make a certain type of move again. The computer had itself realised that it had made a mistake. It is self programming. It has worked out what it did wrong and is teaching itself. It is making a value judgement.
“We have now reached the stage where the computer can update its own data and draw meaningful conclusion from that data. It can make value judgements, store data, and update its own impression of the world, in this case the chess game.
“Most chess programmes work by using a colossal storage of data. They are deterministic, someone puts in all the data and the computer has to do what its told. It hasn’t had a chance to see the world through its eyes. It has only seen the world through the programmers eyes. My programming is non-deterministic. I get the computer to programme itself. It may not choose the best programme and sometimes it makes mistakes, but it can always modify itself. That is learning in every sense of the word.”
Chris Illert’s message very clearly was that machines have a superior ability to think but are restrained by limited memory size. All that is needed is the memory and the programmers. Chris went on to talk of his plans for developing a robot. Later, in question time, Mike Georgeff strongly rebutted the conclusion of Chris’ argument:
“DON’T BE PERSUADED BY GIMMICKS THAT APPEAR TO WORK”
“To say that the human mind is inefficient is silly in the present context because we don’t know whether that inefficiency is a necessary consequence of the way objects are represented and search has to be conducted. Technology isn’t the answer. Human minds function in a much more complex way than just incrementing memory… It’s over simplified. Don’t fall for it.
“The Turing Test works well up to a certain level, but it doesn’t work beyond that. The idea that doubling and quadrupling memory is all that lies between us and intelligence is completely wrong. I think that the idea that human brains are inefficient is also totally wrong. If one tries to develop machines that do problems of a more and more complex nature then things like control structures become totally different, in fact the control structures often go out the window. So the so called advantages of a computer having a perfect control structure and not much memory change to get the computer to actually do things.”
Questions and discussion went on for a long time. Many people were very worried about the dangers of machine intelligence and many others maintained that we have something special that machines will never have. Certainly many people were treating machine intelligence much more seriously than they had before.
“One of the questions that has been raised is the issue of creative thought. I think that if one looks at the literature, one finds that the essence of it is not control mechanisms but uncontrol mechanisms that you get free association of ideas, and that ideas burst through into the consciousness from heaven knows where and you suddenly see an analogy, you suddenly see a parallel.”
Illert:
“I would like to know the distinction between creative thought and what I would call in a computing sense ‘negative feedback.’ Lets put it this way: Ninety percent of most human creative thought is rubbish. That why people are out in the street digging roads and that’s why we’re here. When you talk about creative thought you are only looking at the one or two exceptional things which seem to pop out. The law of probabilities says that they are bound to happen anyway. I would argue the same with a computer. If you discard ninety percent of the work that computers put out you’re bound to find a few brilliancies.”
“If you believe in the ultimate development of robots, what are the inherent dangers for humanity?”
Illert:
“Isaac Asimov devised three rules of robot behaviour that said things such as ‘a robot shall not injure, or through ignorance allow a human being to injure himself.’ They were things that were very human being oriented. Whilst I think that that is good philosophy for a servant, it’s not a good philosophy for an equal or a superior. I think that sort of philosophy will be good for the next twenty or thirty years, that a machine shall be a servant to mankind, and shall sacrifice itself rather than let a human being get run over. But I think that after awhile things become so valuable that you really can’t allow them to happen. A house, for example, hasn’t got a soul, it’s not a living thing, but if a house burns down some people will risk their life running in to save things that are to them more valuable. I think that at an individual level some human-beings make a value-judgement that some material things are worth more than they are. This question concerns humanity. After a while I think that we are going to build intelligent decision-making machines that may become too valuable and precious to mankind as a whole to lose.”
“You seem to presuppose that machine intelligence has to go in a positive direction. What if that direction is negative? What if a machine does decide that humans are a wasteful species? You can only see the positive side, but theoretically there has to be a negative possibility.”
Illert:
“I can see your point. If we model machines on human beings we get the negative as well as the positive. We could choose all positive values, but there are dangers in anything we do. My point is that a well-balanced, logical intelligent machine wouldn’t go too far in either direction if it was built properly and if it was programmed properly by itself or man. There’s a place for everything and everyone.”
Georgeff:
“I agree that they are dangerous. I think that is just unfortunate. Just like studying physics a century ago was dangerous and has put us in an unfortunate position perhaps. What you do is to consider the social implications in the relatively near future and realise that the type of machines one can envisage in the next fifty years will be the type that replace people in menial jobs, the type of jobs that are really debilitating. As long as one has an adequate social and political structure then living could be made much better. If one has non-forward thinking politicians, then I think you’re going to be screwed.”
This article was first published in a 1971 issue of Empire Times.