Can a Computer Have a Mind ?

Can a Computer Have a Mind ? 

Lecture by Sir Roger Penrose.

                      Over the past few decades, electronic computer technology has made enormous strides. Moreover, there can be little doubt that in the decades to follow, there will be further great advances in speed, capacity and logical design.


                The computers of today may be made to seem as sluggish and primitive as the mechanical calculators of yesteryear now appear to us. There is something almost frightening about the pace of development. Already computers are able to perform numerous tasks that had previously been the exclusive province of human thinking, with a speed and accuracy which far outstrip anything that a human being can achieve. We have long been accustomed to machinery which easily out-performs us in physical ways. That causes us no distress. On the contrary, we are only too pleased to have devices which regularly propel us at great speeds across the ground- a good five times as fast as the swiftest human athlete or that can dig holes or demolish unwanted structures at rates which would put teams of dozens of men to shame. We are even more delighted to have machines that can enable us physically to do things we have never been able to do before: they can lift us into the sky and deposit us at the other side of an ocean in a matter of hours. These achievements do not worry our pride. But to be able to think- that has been a very human prerogative. It has, after all, been that ability to think which, when translated to physical terms, has enabled us to transcend our physical limitations and which has seemed to set us above our fellow creatures in achievement. If machines can one day excel us in that one important quality in which we have believed ourselves to be superior, shall we not then have surrendered that unique superiority to our creations ?


                 The question of whether a mechanical device could ever be said to think perhaps even to experience feelings, or to have a mind is not really a new one. 1 But it has been given a new impetus, even an urgency, by the advent of modern computer technology. The question touches upon deep issues of philosophy. What does it mean to think or to feel ? What is a mind? Do minds really exist ? Assuming that they do, to what extent are minds functionally dependent upon the physical structures with which they are associated ? Might minds be able to exist quite independently of such structures ? Or are they simply the functioning of (appropriate kinds of) physical structure ? In any case, is it necessary that the relevant structures be biological in nature (brains), or might minds equally well be associated with pieces of electronic equipment? Are minds subject to the laws of physics? What, indeed, are the laws of physics ? 

                 These are among the issues I shall be attempting to address in this article. To ask for definitive answers to such grandiose questions would, of course, be a tall order. Such answers I cannot provide: nor can anyone else, though some may try to impress us with their guesses. My own guesses will have important roles to play in what follows, but I shall try to be clear in distinguishing such speculation from hard scientific fact, and I shall try also to be clear about the reasons underlying my speculations. My main purpose here, however, is not so much to attempt to guess answers. It is rather to raise certain apparently new issues concerning the relation between the structure of physical law, the nature of mathematics and of conscious thinking, and to present a viewpoint that I have not seen expressed before. It is a viewpoint that I cannot adequately describe in a few words; and this is one reason for my desire to present things in a book of this length. But briefly, and perhaps a little misleadingly, I can at least state that my point of view entails that it is our present lack of understanding of the fundamental laws of physics that prevents us from coming to grips with the concept of 'mind' in physical or logical terms. By this I do not mean that the laws will never be that well known. 

                   On the contrary, part of the aim of this article is to attempt to stimulate future research in directions which seem to be promising in this respect, and to try to make certain fairly specific, and apparently new, suggestions about the place that 'mind' might actually occupy within a development of the physics that we know.

                   I should make clear that my point of view is an unconventional one among physicists and is consequently one which is unlikely to be adopted, at present, by computer scientists or physiologists. Most physicists would claim that the fundamental laws operative at the scale of a human brain are indeed all perfectly well known. It would, of course, not be disputed that there are still many gaps in our knowledge of physics generally. For example, we do not know the basic laws governing the mass-values of the subatomic particles of nature nor the strengths of their interactions. We do not know how to make quantum theory fully consistent with Einstein's special theory of relativity let alone how to construct the 'quantum gravity' theory that would make quantum theory consistent with his general theory of relativity. As a consequence of the latter, we do not understand the nature of space at the absurdly tiny scale of 1/100000000000000000000 of the dimension of the known fundamental particles, though at dimensions larger than that our knowledge is presumed adequate. We do not know whether the universe as a whole is finite or infinite in extent- either in space or in time though such uncertainties would appear to have no bearing whatever on physics at the human scale. We do not understand the physics that must operate at the cores of black holes nor at the big-bang origin of the universe itself. Yet all these issues seem as remote as one could imagine from the 'everyday' scale (or a little smaller) that is relevant to the workings of a human brain. And remote they certainly are! 

                      Nevertheless, I shall argue that there is another vast unknown in our physical understanding at just such a level as could indeed be relevant to the operation of human thought and consciousness in front of (or rather behind) our very noses! It is an unknown that is not even recognized by the majority of physicists, as I shall try to explain. I shall further argue that, quite remarkably, the black holes and big bang are considerations which actually do have a definite bearing on these issues!

                     In what follows I shall attempt to persuade the reader of the force of evidence underlying the viewpoint I am trying to put forward. But in order to understand this viewpoint we shall have a lot of work to do. We shall need to journey through much strange territory some of seemingly dubious relevance and through many disparate fields of endeavour. We shall need to examine the structure, foundations, and puzzles of quantum theory, the basic features of both special and general relativity, of black holes, the big bang, and of the second law of thermodynamics, of Maxwell's theory of electromagnetic phenomena, as well as of the basics of Newtonian mechanics.

                    Questions of philosophy and psychology will have their clear role to play when it comes to attempting to understand the nature and function of consciousness. We shall, of course, have to have some glimpse of the actual neuro physiology of the brain, in addition to suggested computer models. We shall need some idea of the status of artificial intelligence. We shall need to know what a Turing machine is, and to understand the meaning of computability, of Godel's theorem, and of complexity theory. We shall need also to delve into the foundations of mathematics, and even to question the very nature of physical reality.

                    If, at the end of it all, the reader remains unpersuaded by the less conventional of the arguments that I am trying to express, it is at least my hope that she or he will come away with something of genuine value from this tortuous but, I hope, fascinating journey. 

THE "Mr. TURING" TEST 

                       Let us imagine that a new model of computer has come on the market, possibly with a size of memory store and number of logical units in excess of those in a human brain. Suppose also that the machines have been carefully programmed and fed with great quantities of data of an appropriate kind. The manufacturers are claiming that the devices actually think. Perhaps they are also claiming them to be genuinely intelligent. Or they may go further and make the suggestion that the devices actually feel- pain, happiness, compassion, pride, etc. and that they are aware of, and actually understand what they are doing. Indeed, the claim seems to be being made that they are conscious.

                     How are we to tell whether or not the manufacturers' claims are to be believed? Ordinarily, when we purchase a piece of machinery, we judge its worth solely according to the service it provides us. If it satisfactorily performs the tasks we set it, then we are well pleased. If not, then we take it back for repairs or for a replacement. To test the manufacturers' claim that such a device actually has the asserted human attributes we would, according to this criterion, simply ask that it behaves as a human being would in these respects. Provided that it does this satisfactorily, we should have no cause to complain to the manufacturers and no need to return the computer for repairs or replacement.

                   This provides us with a very operational view concerning these matters. The operationalist would say that the computer thinks provided that it acts in distinguishably from the way that a person acts when thinking. For the moment, let us adopt this operational viewpoint. Of course this does not mean that we are asking that the computer move about in the way that a person might while thinking.

                  Still less would we expect it to look like a human being or feel like one to the touch: those would be attributes irrelevant to the computer's purpose.

                  However, this does mean that we are asking it to produce human-like answers to any question that we may care to put to it, and that we are claiming to be satisfied that it indeed thinks (or feels, understands, etc. ) provided that it answers our questions in a way indistinguishable from a human being.

                 This viewpoint was argued for very forcefully in a famous article by Alan Turing, entitled "Computing Machinery and Intelligence', which appeared in 1950 in the philosophical journal Mind (Turing 1950). (We shall be hearing more about Turing later.) In this article the idea now referred to as the Turing test was first described. This was intended to be a test of whether a machine can reasonably be said to think. Let us suppose that a computer (like the one our manufacturers are hawking in the description above) is indeed being claimed to think. According to the Turing test, the computer, together with some human volunteer, are both to be hidden from the view of some (perceptive) interrogator.

               The interrogator has to try to decide which of the two is the computer and which is the human being merely by putting probing questions to each of them.

               These questions, but more importantly the answers that he receives, are all transmitted in an impersonal fashion, say typed on a keyboard and displayed on a screen. The interrogator is allowed no information about either party other than that obtained merely from this question-and-answer session. The human subject answers the questions truthfully and tries to persuade him that he is indeed the human being and that the other subject is the computer; but the computer is programmed to 'lie' so as to try to convince the interrogator that it, instead, is the human being. If in the course of a series of such tests the interrogator is unable to identify the real human subject in any consistent way, then the computer (or the computer's program, or programmer, or designer, etc. ) is deemed to have passed the test.

            Now, it might be argued that this test is actually quite unfair on the computer. For if the roles were reversed so that the human subject instead were being asked to pretend to be a computer and the computer instead to answer truthfully, then it would be only too easy for the interrogator to find out which is which. All she would need to do would be to ask the subject to perform some very complicated arithmetical calculation. A good computer should be able to answer accurately at once, but a human would be easily stumped. (One might have to be a little careful about this, however).

              I hope it is clear to the reader that in my opinion there is a great deal more to the understanding of mental qualities than can be directly obtained from AI (Artificial Intelligence). Nevertheless, I do believe that AI presents a serious case which must be respected and reckoned with. In saying this I do not mean to imply that very much, if anything, has yet been achieved in the simulation of actual intelligence. But one has to bear in mind that the subject is very young. Computers will get faster, have larger rapid-access stores, more logical units, and will have large numbers of operations performed in parallel. There will be improvements in logical design and in programming technique.

             These machines, the vehicles of the AI philosophy, will be vastly improved in their technical capabilities. Moreover, the philosophy itself is not an intrinsically absurd one. Perhaps human intelligence can indeed be very accurately simulated by electronic computers essentially the computers of today, based on principles that are already understood, but with the much greater capacity, speed, etc. , that they are bound to have in the years to come. Perhaps, even, these devices will actually be intelligent; perhaps they will think, feel, and have minds. Or perhaps they will not, and some new principle is needed, which is at present thoroughly lacking. That is what is at issue, and it is a question that cannot be dismissed lightly. I shall try to present evidence, as best I see it.