Socrates
Let’s get away from the disturbing problems of the present, in order to turn back to those much more basic and lasting ones which concern the nature of man. Mr. Marvin Minsky, you were the leading authority on Artificial Intelligence, glad to meet you in paradise!
Minsky
Oh, it’s you the philosopher? But let me tell you, we no longer need to philosophize about man in order to explore his true nature. For the first time in history, the exact sciences provide us with the means to use numbers and irrefutable proofs to make our statements irrefutable. What the natural sciences have achieved within the short lapse of merely three hundred years, namely making outward reality calculable so that we can master it, has now become possible with regard to man himself. Let me stress my point: for the first time in history.
Socrates
In other words, you don’t need philosophers anymore. We are mere dreamers, who created all sorts of fables and myths about man. But now you come with your measuring instruments, and in the end you will set up a handful of equations that will allow you to predict human thoughts and actions for the next fourteen days – just as you can now predict the weather for a fortnight and the orbit of Jupiter even for the coming millennia.
Minsky
Right, but I have no qualms in praising your past achievements. The conversations that someone like you used to have with clever young men in ancient Athens, oh, I quite loved to read that as a student. So much poetry and enthusiasm, really impressive! But let us be honest, your division of the state according to teaching, military and nutritional status had no practical consequences. Your most brilliant student Plato remained without political influence all his life; Syracuse’s infamous dictator did not even think of taking over his system but almost turned him into a slave for the rest of his life. You philosophers have provided the world with narratives, but your practical effect has been nil because it was not based on science. That is why the likes of you are now disappearing even from universities, while the natural sciences are booming.
Socrates
This makes you probably think that you will soon be able to describe human thinking and behavior so well that you can plan a community on the drawing board. Everything will then be as reliable and predictable as in a termite state? I am reminded of Watson and Skinner, the founders of Behaviorism, who proclaimed this sad utopia a century ago.
Minsky
That is your private quite malicious wording, because you want to suggest that freedom would be absent from such a state. But freedom is an illusion, as you should know. As long as man acts according to laws, there can be no freedom, but if instead he is not ruled by laws, then he is subject to blind and meaningless chance. Do you want to call that by the name of freedom? Your reference to a termite state is therefore nothing but misleading nonsense. Man has never been free, even if it is that what he imagines himself to be. Human scientists are merely demonstrating this lack of freedom by showing that it is quite obvious in artificial intelligence and in robots.
Socrates
I admit, you’ve done a lot of great things.
I Artificial man proves that we are on the right path to completely decipher his natural counterpart
Minsky
Indeed, that really deserves admiration. Our practical successes provide no less than conclusive proof that we are on the verge of completely decoding the human being. Our largest calculating machines carry out tasks within seconds that normal human beings have not been able to solve even in thousands of years. Our chess robots, equipped with artificial intelligence, now beat every human player, our robot doctors provide better diagnoses than professors with university degrees, our AI-equipped artificial lawyers have more knowledge and therefore more competence than any natural expert, our control systems in airplanes make pilots of flesh and blood superfluous – and reduce the frequency of accidents to that minimum that will always be caused by defective technology (as in humans as well, for instance, by heart failure). Cars controlled by Artificial Intelligence soon will allow fluent traffic, which in addition will be nearby accident-free.
Socrates
If I understand you correctly, your goal is to replace the fallible, comparatively dull human being so limited in his knowledge with his perfect counterpart, namely an artificial brain in an artificial robot – both considered far superior in every respect.
Minsky
That’s right, and we have already progressed so far along this path that most of the professions that still exist today will no longer be needed in a few years‘ time. Already now, pilots have become an expendable luxury, the drivers of trucks, busses and taxis will soon be dismissed as well. In diagnostics, doctors are already largely superfluous, only in therapy they still play a certain role, but robots will prove to be much better surgeons. Interpreters are hardly needed any more, as translation machines are now doing an excellent job. I confess that despite all the progress we can be very proud of, I am afraid that this might cause problems as so many people will lose their jobs.
Socrates
I agree that this could indeed turn out to be a tremendous problem. But it seems much more interesting to me to pursue a more fundamental question. Is your optimism at all justified that you will soon be able to completely decipher human beings and replace them with artificial superhumans?
I want to choose this topic because I don’t just think your thesis is exaggerated, as if you simply needed a little more time for further research. No, I think it is basically and demonstrably wrong. More than that, I consider it to be a great and dangerous illusion – despite all the undeniable practical successes that you already achieved and will undoubtedly still achieve in the future.
Minsky
Really? There speaks the philosopher opposing empirical knowledge with lofty speculation. But I will certainly not follow you on that path. The fact is that we are already very close to our goal of replacing natural man with his artificial substitute. We just don’t yet know which of two ways will allow us to reach it within the shortest possible time. Some want to make the elementary rules of logic the basis of artificial intelligence, i.e. to start from a few basic relationships and refine them more and more until we will not only imitate and perfect all natural operations of the human brain, but also carry them out much faster and more comprehensively (logic-based symbolic processing).
Others want to take a more pragmatic approach, combining all sorts of tried and tested problem-solving methods (deep learning). But in the end, our goal always remains the same. In the end, we will have created superhumans, who will not only perform all possible mental operations like any arbitrarily chosen person out of a world population of 10 billion, but who possesses much greater abilities, since his knowledge and his ability to react are capable of almost limitless extension.
Socrates
You want to say that you will be able to correctly predict what this artificial being thinks and how he acts?
Minsky
Of course. That is the final point and purpose of our research. Since it is us scientists who created this superhuman, we are of course able to predict his possible thought processes and actions. You will admit that this is a breakthrough of historic dimensions. So far science successfully managed to dominate inanimate nature, only man seemed to elude our efforts. Man remained unpredictable. The superhuman, whom we are equipping with artificial intelligence will change all this. He will be completely predictable, since he is ruled by the program we made for him. Given that his artificial brain largely surpasses the faculties of any natural human being, we will, of course, have no difficulty at all in decoding the latter even more easily. To be sure: Behaviorists like Watson and Skinner had already suggested the right direction. But now we are equipped with the required technical means to actually create fully computable and programmable human beings.
Socrates
You think that in the future you will succeed in predicting human thoughts and actions? I understand what you mean. If artificial superhumans are predictable, then this should apply even more to the two of us, who in comparison have only quite modest brain structures.
Minsky
Exactly! We must finally get rid of the absurd assumption that science may well calculate and dominate external nature, but remains helpless with regard to man as if the latter were not part of nature as well. That is nothing but a stupid, unscientific prejudice. After all, we humans belong to nature and are therefore at the mercy of all its laws. If we can predict and dominate the processes of outward nature, then it must be possible as well to predict and dominate the processes of our brain. Let me tell you: it can only be a matter of time before science deciphers my and your brain and knows what we will think tomorrow or in a week’s time.
Socrates
And I insist that any critical scientist must reject the thesis of the complete calculability of man and nature if he wants to remain a scientist, or in other words, that he can no longer be called a serious scientist if he accepts such a thesis.
II Robots cannot perform experiments
Minsky
I am aware that philosophers tend to surprise normal human beings. You will have to explain this statement to a layman like myself.
Socrates
It’s not as hard as you might think. It is well known that the empirical sciences conduct experiments in order to test their theoretical statements. I ask you: don’t they take for granted that experiments can be conducted at any time, in any place, and in any size and complexity? In other words, don’t scientists assume that they can arbitrarily bring about real processes at any time by their own volition?
Minsky
That’s right, that’s the way how experiments proceed.
Socrates
But if it is due to the will of human beings that an experiment should take place, then you will agree that its occurrence at that moment and in that particular place (e.g. in a laboratory) cannot be determined by the history of the universe or its overall state? We must even claim that, with regard to the latter, it is entirely coincidental, since it owes its occurrence solely to the will of some definite scientist?
Minsky
I guess that’s right.
Socrates
I am happy that we both are agreed on the matter, because from Laplace to Bertrand Russell, a completely different view has been taken for granted throughout the history of modern thought. French mathematician Laplace expressed this view in classical form to Napoleon: „An intellect, which, at a certain moment, would know all forces that set nature in motion, and all positions of all items of which nature is composed, if this intellect were also vast enough to submit these data to analysis, it would embrace in a single formula the movements of the greatest bodies of the universe and those of the tiniest atom; for such an intellect nothing would be uncertain and the future just like the past would be present before its eyes.“
You see that according to this view, there can be no events that cannot be derived from any preceding stage of the universe. A perfect intelligence should even be able to deduce the whole future from its current state. If you really think that further research on robots with artificial intelligence will soon allow us to predict human thoughts and actions, then you adhere to this thesis – consciously or not.
Minsky
Wait a minute, you have to be a little bit more specific.
III An irresolvable logical contradiction!
Socrates
Yes, of course, I am ready to describe your Brave New World somewhat more in detail. As of now we are already sitting in planes that are steered automatically and tomorrow we will be sitting in cars that reach their destination without human help, but at least we still have the choice of directing the car to self-determined destinations, for example to Bonn or Rome. But your artificial superhuman will no longer have this choice, because you predict his behavior just like a lunar eclipse. According to Laplace and your own conviction he has as little freedom as that vanishing moon and the whole of inanimate nature. You are convinced that the scientists of tomorrow will not merely know the laws governing planes and cars, but those laws as well that govern the brains of the people using them. In this way you will be able to predict the future behavior of both.
Yes, and here you suddenly fall into a logical trap – I mean into perfect self-contradiction.
Minsky
Oh, fine. Where do you think you find it?
Socrates
It is found within the scientist himself, who at the same time turns into the subject and the object of prediction. For he should be able to foresee what he himself will think in the future. With your chess players, with your robot doctors, with your lawyers equipped with artificial intelligence this has become true already now, because you have programmed their thoughts and behavior. So, you know exactly how they must react. But since you claim that these artificial beings embody the essence of man in a perfect way, your reasoning applies doubly to ourselves, those imperfect natural human beings. Which means, that the researcher claims that at some point he will muster the intelligence of Laplace to predict not only his own later behavior but even all the future results of his research.
Minsky
That is true, but it will of course take years, perhaps decades, before research actually gets to this stage. That’s why it makes little sense to talk about it now.
Socrates
No, on the contrary, we should assume that your goal has already been achieved, only then can we talk about its inevitable logical consequences – I mean its self-contradiction.
Minsky
I know what you’re going to say. If I could predict exactly what I will do tomorrow, I will certainly not do it anymore, because all the attraction of novelty is lost since I would, of course, also know my future feelings and sensations. Life would become completely superfluous, for I would be nothing more than a machine without any drive of its own.
But you are making a serious mistake. We can’t reject Laplace’s deterministic view just because we don’t like it.
Socrates
No, we must reject it because it is manifestly wrong. The freedom of man and nature is proven by scientific experiments. These conform to the laws of nature but cannot be derived from them as they owe their existence exclusively to human volition. Without the latter these events would not exist – not in this particular place at this particular time with these particular characteristics. A rocket flies to the moon according to laws that we have learned from nature, but the fact that it does so on Monday, June 23, does not obey any law of nature, but was decreed by the director of NASA or any other human agent.
Minsky
I admit that experiments presuppose an open reality where human will can intervene. But the evolution of nature takes place completely independently of our will. So, I see little use in choosing man-made experiments to explain reality.
Socrates
Objection! In fact, they are very useful for the purpose of demonstrating that we will never fully explain reality if we only consider laws and their deterministic effect.
Please look here: experiments are underivable from past stages of outward reality, but this applies to the evolution of nature as well. Our 83 primordial elements cannot be derived from the knowledge we gained of the primordial plasma at the initial state of the universe; they must only conform to those initial conditions. In their turn, molecules cannot be derived from primordial elements, they are merely subject to the laws that applied up to that point. And further: from the molecules of inorganic nature we cannot derive the later developed organic ones, although these remain subject to the laws of physics and chemistry. Finally, from biology we cannot derive the mental sphere, although the latter too cannot disregard the laws of biological life.
Let me put the matter in a metaphorical garb: a cosmic experimenter could have thought up all this, just as earthly scientists think up experiments. Both are limited in the creation of new realities by the laws already existing, and yet they create something new at a certain time in a certain place.
This is why we must unequivocally reject the idea of Laplace and his many successors up to Bertrand Russell. We are the inhabitants of an open world where new realities constantly emerge, realities which cannot be deduced from existing laws even though they cannot transgress them. Every experimental scientist creates reality by the sheer force of his will.
IV Your artificial superhuman cannot act, he can only re-act
Minsky
Interesting, surprising and for me at first sight even worth considering, although you are contradicting a three hundred year old tradition, which, as you certainly know, has meanwhile been relativized by quantum physics. The latter accepts chance and the underivable new. But let me mention this fact in passing. To me it seems more important that you come back to our topic. What do experiments and the new in nature have to do with artificial intelligence and robots?
Socrates
That’s what you ask? But the connection is obvious! Didn’t you previously insist that man is part of nature? Of course, I fully agree with you. But this means that what applies to the latter must also apply to man. If new reality can be produced at any time within existing reality by way of experiment or any other kind of human action, then there must be such islands of newness in human thought as well. In other words, we have to assume that thoughts constantly pop up in our brains (and lead to actions) that we cannot derive from the past of our thinking or from its overall state. For this reason, we will never be able to derive future thinking from its present shape.
Minsky
All right, but supposing you were right with this statement, I still don’t see what significance this statement has for artificial humans. To put it bluntly, I don’t see how you can criticize our great project, the project of finally creating the perfect human being, the artificial superhuman, who has all the advantages of the natural man but without suffering from his shortcomings. What is wrong with this superb project?
Socrates
I’m not questioning this, your goal. I’m just saying you’ll never achieve it because you can’t. The artificial robot equipped with digital intelligence differs from its natural counterpart in that it can do no more than re-act to reality.
Certainly, in many areas it does infinitely better than its natural counterpart. That’s a lot, but that’s all. The chess robot reacts according to a programmed algorithm to every move by its opponent, the diagnostic robot evaluates the radiologist’s photos according to a built-in program. The flight pilot reacts with the greatest reliability to the totality of situations with which he was previously fed. But no artificial human being creates new realities – and this for a simple, logically compelling reason: programmers are incapable of deducing the new from the old, i.e. the unknown from the familiar. But man does exactly that: he creates the new. Just think of how, over the past two hundred years, he has succeeded in radically reshaping both his natural and social environments.
Minsky
No, I definitely contest your thesis that artificial intelligence can’t create things new. We have law and we have chance. Laws dictate the rules, but a random generator can very well imitate chance and be built into our artificial humans. The robot then not only reacts to external stimuli and conditions, but also acts like a natural human being by taking new, unpredictable actions. I maintain that we can imitate and probably even surpass human beings in this respect as well.
V Chance coexistent with laws
Socrates
Fine, this brings us to a key concept of our joint investigation: chance.
Minsky
Right, and this concept helps us to get rid of freedom, which objectively doesn’t exist as it represents no more than a somewhat gratifying illusion. Remember what I said before: as long as man acts according to laws, he cannot be free, but if he does not act according to laws, then he is ruled by chance – so, he can’t be free either. We, the scientists studying man, are now demonstrating by means of artificial intelligence and robots that we do not need freedom in order to explain human behavior. Using the random generator, we create chance artificially – that’s all we need.
Socrates
I know that’s what you think, but in reality you are wrong again. You are quite incapable of explaining let alone imitate human behavior.
Minsky
Why not? Are you going to deny that there are random generators?
Socrates
No, I’m not denying that at all, I just maintain that they don’t produce chance. First of all, the kind of chance they produce is radically limited from the outset. It would be genuine chance if, for instance, a billion dollars were suddenly to appear on my account or if a strong gust of wind carried me to the hammock in my garden. But, of course, that’s not what a random generator achieves. It can only produce sequences of numbers.
Minsky
But it certainly suffices that these are completely irregular.
Socrates
To be sure, this would be enough for the purpose of demonstration, but they cannot be random if scientists want to produce them systematically.
Minsky
That’s a claim I am hearing for the first time.
Socrates
Let me ask you. Can we use a rule to create something bereft of rules?
Minsky
Of course not.
Socrates
But that’s what happens when we program a generator to produce chance – all planning is inevitably based on rules.
Minsky
All right, I can see your point. No algorithm is capable of producing a sequence that does not obey any algorithm. That’s elementary logic. But you missed an important issue. We may, for example, plan the randomizing device in such a way that it looks through the window where there’s normal pedestrian traffic. Let a sensor measure the size of each passing pedestrian and multiply it by the number of seconds that elapsed since the last pedestrian appeared. Then we get real chance. Because nobody can imagine a law that determines the frequency, the different sizes and the time distance between pedestrians.
Or take another case. We know the half-life after which the amount of radiating radium will be reduced by fifty percent, but we do not know the point in time when a single alpha particle will be sent out. This moment is ruled by chance. Which means that we could likewise feed our random generator with these irregular occurrences in order to get a sequence of numbers that cannot be calculated by any algorithm.
Socrates
Quite true, but do you know what that means? You are asserting – and quite rightly so – that man is incapable of producing chance, he can only produce events obeying some rule. If he wants to produce or to imitate chance, he has to take it ready-made from nature.
VI Chance remains a pure, unintelligible secret. This applies to chance in non-human nature as well as in man.
Minsky
And what is it you’re trying to say that is so terrific? Isn’t this but a bunch of philosophical subtleties that won’t get us anywhere?
Socrates
On the contrary, this is a basic insight. It says that chance is a mystery to us. We understand only rules, we recognize only laws, but everything irregular, everything lawless remains terra incognita for human understanding.
Minsky
Right, and that is precisely why it is of no concern to science, and why science speaks of „blind and meaningless“ chance removing it completely from its mental horizon.
Socrates
But how can we call something blind and meaningless, if we fail to know what it is? And how do you want to build something into artificial intelligence and robots that is beyond our understanding? And last not least, how do you want to achieve the self-imposed goal of predicting human behavior when chance eludes predictability?
Minsky
I see your point. That’s not possible, of course.
Socrates
So, we are back to our starting point. Artificial human beings, i.e. AI-equipped robots, can never be congruent with natural ones, because, unlike the latter, they always follows definite rules or algorithms. As mentioned before, we cannot program them with algorithms that do not obey any algorithm. Therefore, they will always remain second hand imitations, which may greatly exceed the performance of natural humans in terms of special performance, but do not possess that characteristic endowment of true human beings and of nature as well, which consists in getting beyond rules (or laws). For both are dominated by chance and by laws to the same extent – and chance is by no means blind and meaningless. We simply don’t know what it is.
Minsky
This reasoning is hopelessly abstract and philosophical. An example, please!
Socrates
I already used an example. The laws of nature known to us determine the path a rocket will take from Earth to Mars. But the exact time of its launch depends on the will of a certain Mr. Meyer, on financial means available at the moment, the state of the art and last but not least even on weather conditions at the launch pad – and of course on a thousand other conditions, all of which cannot be calculated in advance. Chance and law are therefore equally involved in this specific event as in any other. Laws are only valid under certain conditions.
Minsky
But that does not exclude the possibility that science may in due time explain all reality, including the human being acting within it, according to laws!
Socrates
You are wrong. It is precisely this goal that science will never reach – not because of inadequate human knowledge, but in principle. We already stated that experiments as such constitute unpredictable interventions of human volition. We are thus forced to conclude that chance does not just exist alongside natural laws, but must be seen as a second dimension of reality, and a necessary one at that. For science presupposes experiments, that is, arbitrary interventions in nature. For this reason, nature cannot follow a deterministic course for in this case it would not permit such intervention. We cannot conceive natural laws without freedom (chance) and freedom (chance) without natural laws.
Minsky
Quod erat demonstrandum! Bravo! But now I would like to see that extraordinary experiment that will prove your thesis. Because otherwise all this will remain just theory – not to use the word speculation. And what’s more, I don’t like the fact that now you speak of freedom and chance as if they were interchangeable concepts.
Socrates
Sorry, I have to disappoint you as to such an experiment. Experiments can only confirm rules or prove that we were wrong when we tried to confirm them. But there is no conceivable experiment with which we could prove the existence of irregularity with regard to any specific event. Let’s turn back to the sequence of passing pedestrians in front of our window. There is no experiment to prove that we could not in the distant future arrive at an infinitely complex algorithm that faithfully describes such a sequence. Nor can there be any experiment that conclusively demonstrates that the decay of radium does not obey a rule still hidden from us today.
On the other hand, there is no conceivable experiment either with which we could prove the opposite, namely that all reality is deterministically dominated by laws, as Bertrand Russell still believed. It was precisely because such an experiment is impossible, that Karl Popper spoke of the principle of causation as a metaphysical assumption.
Minsky
Well, in this case you are only confirming what I stated before, namely that your thesis is based on mere speculation, because it cannot be corroborated by experimental science.
Socrates
No, you’re wrong this time again! For at this stage it is the philosopher who comes into his own, or any scientist who asks about the preconditions of human thinking. Earlier, we have seen that scientists get entangled in irresolvable contradictions when they negate freedom. Even if we speak of chance instead of freedom, we cannot conceive reality without it. But we will never know what freedom or chance „really“ or „objectively“ are, because then we would have transformed them into something regular that can be grasped by means of analysis. But we cannot reduce what is new and as yet unknown to something familiar.
Minsky
So, we must finally accept the verdict that freedom is nothing but chance, and therefore blind and meaningless?
Socrates
No, not at all, it’s just that in many cases we can’t give freedom a meaning that we understand – and in this case it becomes mere chance to human understanding. From the perspective of a distant observer of cosmic events, the arbitrary explosion of an atomic bomb is mere chance, as would be the bursting of a supernova. But from the perspective of scientists conducting such an experiment, it constitutes a meaningful event.
The evolution of the cosmos from primordial plasma to the human spirit appears to today’s science as being the result of an endless sequence of blind and meaningless chance. That is why the whole of nature including all living beings is often apostrophized as totally meaningless. But against this point of view the zoologist Rupert Riedl rightly asserts: „Would it not be utterly presumptuous if the tick wanted to imagine the blood vessels of a mammal, a dog the international drug scene or if we imagined the laws beyond the cosmos?
When speaking about chance we only describe the limits of our understanding, because we can attribute meaning only to purposeful actions. But human meaning manifests itself in ever new forms as „Creative Reason“ is constantly creating new manifestations. Ultimately that is what history consists of: newly evolving meaning.
Obviously, meaning is not the same today as it was in the times of hunter-gatherers, nomads or early peasants. We cannot imitate these differences by equipping robots with a random generator – in this case they will either act according to the rules of yesterday or they simply act in a completely meaningless way. Artificial superhumans are condemned to be mere caricatures of natural ones.
*In this essay, I resume some of the conclusions arrived at in my book „Creative Reason – A Synthetic Philosophy of Freedom in Nature and Man (Homage to William James“. What is completely missing in this essay is, due to lack of space, the historical perspective, which occupies a large part of my book.