Hugo Steinhaus on Generative AI

"It's unknown whether one should rejoice or worry that the near future will provide us with a companion in the form of a box that would give banal answers to our banal questions."

I’m still reading the diaries and memoirs of the mathematician and inventor Hugo Steinhaus, the main subject in my last post (other than cake). Steinhaus is witty and insightful, so when I saw a reference to him publishing an article titled “Na Marginesie Cybernetyki” (“On The Margins of Cybernetics”) I looked it up. I read it via machine translation, which is an irony that would have amused him. Feel free to scroll down past my summary and just read the translation, if you’d like.

It’s a great essay—it holds up astonishingly well for something written in 1963, often demonstrating clearer thinking than some prominent scientists and commentators today. When Gary Marcus, for example, wrote in 2018 that deep learning was overhyped and probably approaching a wall, he was wrong, but at least not directly contradicted by evidence—GPT-2 was released in 2019. The fact that he’s still writing that GPT-2 will never be invented shows that the fallacies Steinhaus discussed are alive and well.

Steinhaus argues that new technology constitutes a proof that some supposedly magical process is actually mechanical. Before the airplane, people could get away with thinking that birds flew “because they’re birds.” Then the Wrights built a machine and told the world that it worked using the same principles that let a bird fly. So birds fly because they’re flying machines. Before the gramophone, music was mystical, etherial. Now, you could listen to a gramophone make noise and have the same experience as if it were a human singing. Gramophone operations can be described using simple equations—ergo, so can the human’s song.

That’s the correct way to respond to a new invention, Steinhaus says. There are two popular wrong ways—you can try to argue that a gramophone playing “isn’t actually music,” whatever that means. Or you can decide the gramophone must have a soul—in fact, since it can play any song, a superhuman ability, it must be something like a god.

So now, turning to computers… Steinhaus says it’s obvious that, while it might be very difficult, there’s no fundamental reason we won’t be able to invent a device that doesn’t just record music, but can read sheet music and lyrics and transform them into a song. Such a device, he says, will also be able to carry on a conversation, although it’ll be limited due to being trained on human writing with all of its fallacious “received wisdom.” If you ask one whether it can think, it will respond “Unfortunately, no, because I don’t have a soul.”

On the Margins of Cybernetics

Hugo Steinhaus (published in Znak Monthly, 1963)

Three hundred years have passed since the death of Blaise Pascal, but our minds return to this independent spirit, too early for his own epoch to understand him in full. He died at forty. He invented a calculating machine before he turned twenty. In principle, this arithmometer did not differ from those that have served as calculators in shops, banks, and offices since the beginning of our century - it too could quickly add multi-digit numbers. But the prototype built by the author of the Pensées did not come into use; in his time, gear trains were unknown, which today are an essential feature of such machines - in the 17th century there was no geometry or kinematics of gear wheels yet, and they did not know how to cut such gears in hard metal; instead, they used discs equipped with pins around the circumference - at high rotation speeds the pins broke, and this technological deficiency decided the failure of the invention. But our age knew how to deal with such problems of precision mechanics, and calculating machines, in various versions and under different names, won their rightful place in the utilitarian world; there was nothing mysterious about them, so they aroused no greater admiration than locomotives and steam mills. Only the Second World War posed a task that the most efficient arithmometer could not handle: How should artillery shoot down aircraft carrying death and destruction at speeds of several hundred kilometers per hour? For this it was necessary to calculate from the aircraft's position and speed such a direction of fire that the winged enemy would fly into a fountain of shells; an automaton was needed that would in a fraction of a second perceive the aircraft's position and perform calculations, while directing all barrels so that the aircraft, despite its speed, would not escape unharmed... Here the technology of the "age of steam and electricity" was no longer sufficient - there are no gears that would function at such speed, and it is even harder to conceive an apparatus that would withstand the enormous starting acceleration. There was only one solution: electron tubes, commonly known as components of radio apparatus - they can change their electrical conductivity hundreds of thousands of times per second, and this advantage was exploited in modern calculating machines; hence the name electronic machines.

As often happens, a decisive leap in military technology brought consequences in other fields that seemingly were distant from the original task. American mathematician Norbert Wiener became interested in another aspect of shooting, namely the necessity of gathering information, acting on the basis of this information, determining the error of this action, and eliminating this error in a new attempt at action. This “trial and error” method has probably been known to artillerists since the invention of gunpowder. But Wiener was the first to realize its universality. A cyclist uses it, a car driver uses it when he constantly compensates with slight steering wheel movements for the vehicle's deviations from a direction parallel to the highway, and a helmsman acts no differently - precisely the ship's rudder gave Wiener the best example of the general law he called "feedback" - the Polish equivalent of this term, “sprzężenie zwrotne,” sounds rather colorless. On large ships, steering is automated; when the rudder is set so that the ship sails straight, e.g., northward, and an accidental cause throws it off course, the gyrostat axis maintaining a constant northern direction will deviate from the ship's axis - this will cause steam to flow to the appropriate side of the cylinder, so that the piston will turn the rudder and cause the ship to rotate toward the correct course - when it has already restored the northern direction to the ship, the valve will close the steam supply. This clever device is not, however, the exclusiveprivilege of contemporary automated ocean giants: for a solitary sailor on an ordinary boat, the oar serves as a rudder, and a lighthouse on shore orients him as to direction - when he sees that the boat has gone off course, he corrects the direction with the oar; here the cycle consists of several links: the lighthouse casts its image on the rower's retina, the optic nerve transmits the image to the brain and arouses an impulse there, which by another route runs to the muscles holding the oar, and having reached there evokes contraction of those muscles, which causes a change in the oar's angle relative to the boat. This feedback, though used for many thousands of years, was not - until the last years of the Second War - the subject of conscious scientific analysis; it is incomparably more refined and universal than the modern gyroscopic rudder.

Now Wiener was the first to realize the fundamental similarity of these two seemingly different mechanisms. He organized a team of scholars who decided to systematically cross the boundaries of their specialties. In his book “Cybernetics, or Control and Communication in the Animal and the Machine,” he reports on the work of this team from the outbreak of the Second War until the date of publication of this work - he cites in it primarily the American physiologist Arthur Rosenblueth as his wartime collaborator, later replaced by other physiologists, doctors, mathematicians, and engineers. Most interesting is Chapter IV, devoted to feedback and oscillation - there for the first time he compares a patient in a neurological clinic with a sick automatic-type rudder. For the first time he states that every malf unction of the rudder has its clinical counterpart in the form of one of many diseases of the neuro-motor apparatus well known to doctors. For the first time it was indicated what defects of regulators and control automata correspond to such diseases as - for example - tabes dorsalis or Parkinson's tremor. The similarity becomes visible when we try to describe mathematically models of such common activities as reaching for matches or grasping a door handle. Wiener and his companions created precisely such models - today their technical terms “input,” “connecting channel,” “output,” “noise,” “amount of information,” “feedback” have passed from specialists' vocabulary to the general press, enriching it without ballast. But it's not these names that matter here, but an amazing occurrence; medical history had to record a strange tale of strange doctors: the patient was a rudder, the symptoms were determined by a mathematician, and neurologists knew the disease picture before engineers invented automatic rudders, i.e., patients! In the history of philosophy one can find equally momentous events - today it's already hard to evaluate them with the proper measure, but in their time they were milestones of wisdom, though later the direction of the road was to change many times. Descartes' statement that animals are only automata required of the thinker who uttered it no small independence of mind; this ultra- materialistic thesis asked for application to humans, so closely related to higher mammals by complete similarity of structure of all organs (not excluding the neuromotor apparatus), that a mathematician could give a dog and its master as an example of a pair of topologically equivalent creatures. But from Descartes to Pavlov - three centuries passed...

It was only a hundred years ago that it was understood that the chemistry of organic matter transformations is subject to the same laws as the chemistry of dead substances, and that the principle of conservation of mass and energy also applies to living organisms! It was not easy to overcome such psychological resistance of the era when it was established that the movements of people and animals are governed by the same dynamics that were perfectly known in the 18th century and soon after tested on machines practically. But the most interesting episode occurred much later, in 1903, when Wilbur Wright traced the first circle in the sky with a motor airplane, and to the question of how his machine stays in the air though it is heavier than air, he replied: “Like a bird” - just like a bird. How should we understand this? Because a bird is a flying machine, says Wright, I built our machine imitating a bird... In this sentence of the first aviator lies faith in Descartes - before 1903, most people, even educated ones, believed that a bird flies because it is a bird, and it is created for flying, and no apparatus can match it because it is not a bird...

Wright bet on a different card: a bird flies because it is a machine for flying - therefore one should build an apparatus equipped with wings and give it a gasoline motor instead of muscles... Wiener's path was parallel to Wright’s, but the direction was opposite: An automatic ship’s rudder sometimes exhibits so-called “hunting” - it can be described mathematically; a human being after brain damage must have a steering apparatus, and it is subject to defects similar to the defects of an automatic rudder - so says Wiener; Wright, on the other hand, told himself that since there exists a living flying machine, one can build an artificial one that will imitate the living one in this function. Wiener knew the theory of the rudder and its defects; he also knew that the human neuromotor apparatus is subject to diseases - therefore he could describe the symptoms of Parkinson's disease without looking into a neurophysiology textbook; his discovery did not require inventions, unlike the discovery that a bird is a machine.

This kind of discovery is rarely recorded in the history of human thought - one can include Newton's discovery of gravitation among them. It did not make a great impression on people; even the educated, even later ones, such as Goethe, did not realize what hour had struck on the clock - perhaps because the models of the theory were planets and the moon, distant and mysterious objects. When a few years ago artificial earth satellites were successf ully placed in orbits, the majority became concerned about what would happen when the artificial ones fall to earth, and a minority was amazed that they don't fall. When someone noticed that the real moon also doesn't fall to earth, they were told: it doesn't fall because it's the moon - this almost literally repeated the thesis of aviation skeptics “a bird doesn't fall because it's a bird,” and only a few understood that if the moon doesn't fall, then another object set in orbit around the earth also won't fall.

Even fewer were those who noticed that they were witnesses to the first experiment in history perfectly confirming Newton's theory on objects made by human hands and visible to the naked eye - one had to wait for this almost three hundred years. Again, technological difficulties were the cause of the delay; in Newton's time there were no such explosive materials that could give a projectile at least thirty times greater speed than that of a cannonball of that time. Even today there is still a lack of laboratory-scale experiment; one cannot show students in school the law of gravitation on models as one shows, for example, pendulum motion. Recently I found in a newspaper information about observations that MARINER II transmitted to Earth's inhabitants; the information showed the impossibility of organic life on Venus - it was printed in small type. The rocket ran for over a hundred days and covered about 80 million kilometers to confirm the old aphorism that there are more things between heaven and earth than can be dreamed of in school philosophy - indeed, it never could have occurred to me that this confirmation would be printed in small type where the thickest capital letters give news that even after being provided with a negation sign would not become true.

Philosophers have long asked whether man is a machine. Marionettes give a partial answer, which evoke applause by their resemblance of posture and movement to the human prototype - the secret of the wooden Punch performing somersaults on stage is his obedience to the same laws of mechanics to which the living clown in the circus is subject; but let's not think that it was always easy to believe in this truth, sounding so natural today...

It wasn't long ago that the Paris Academy of Sciences was puzzled by the paradox of a cat falling on four paws when all four threads holding it suspended on its back are cut simultaneously - rational mechanics, as its great creators, members of this same Academy, called it, has long been able to explain this really interesting experiment, but the very fact of the controversy testifies that there were many supporters of a different explanation of the feline salto mortale - anyone can guess what this explanation must have been: a cat always falls on four paws because it is a cat; this irrational mechanics pleased them more than the rational one.

The controversy of today’s era is the calculating machine, or speaking more ominously, the electronic digital computer. It is not a playing cabinet or a ticket vending machine; modern computers have higher ambitions: they play chess, compose dance music, and determine the authenticity of signatures - admittedly they do this worse than living specialists, but it's not easy to find a human who would single-handedly replace them in all three functions, and - as a side occupation - practice solving a hundred equations with a hundred unknowns in a few minutes. Belittlers of machines claim that whatever an electronic computer does, it will always be executing a program written by human hand and inserted into the appropriate drawer of the machine by a human, whose blind tool the machine is. But the machine's defender will point out to them that the metaphor “blind” is inappropriate in their mouths, because after all the eye is also a tool, and yet it is not blind; here the evolution of knowledge was similar to the discovery of cybernetics, because the invention of the glass lens and discovery of its ray-focusing function preceded anatomists’ recognition of the role of the eye’s lens. Then the anti-machinist will charge the computer with lack of consciousness: it's hard to imagine a machine that is aware of its actions, and even harder one that has the privilege of free decision. Instead of defending machines, let me be allowed to recall what Pascal writes about such games as heads-or-tails. According to him, every player of some intelligence will beat a naive opponent in this game if he correctly assesses his naivety - he will know that a simpleton, having lost on heads, will bet on tails; a more refined partner will manage with an averagely clever one, who can only cope with a completely dim-witted one...

To benefit from Pascal's theory, one must assess correctly the partner's degree of intelligence based on the first few throws - and perhaps also from the partner's appearance. I once saw in a bar in New Mexico how my friend played with a chance opponent and beat him very effectively applying Pascal's advice. But what will happen if the partner is a computer? D. W. Hagelbarger gave an answer to this question through appropriate modification of the machine, which he ordered to guess whether the living partner would place the coin heads up or tails. After a large number of trials it turned out that the machine wins on average 55 games out of 100. Its secret is this: it records in memory the entire history of the game from the beginning, because its caretaker informs it each time about the result. At first this information gives no visible benefits, but after several dozen records the machine bases its decision on experience: it searches in memory for game sequences composed of several (e.g., three) consecutive decisions by the player and as many machine responses, such that the entire cycle (six moves) does not differ from the last cycle (six moves) - if it finds such cycles, it concludes that the living player will currently act as he acted in the majority of those selected cases, and announces this guess as its response. Here Pascalian finesse won't help the living player: even if he tries to avoid his previous principles and change tactics, sooner or later he will fall into routine and lose in the long run - such a player will fare worst who thinks he is dealing with an automaton cyclically repeating signals once recorded on tape. It's the opposite: the electronic computer considers the living partner an automaton and is not so wrong - the machine's positive balance forces the human to confess that rather the computer saw through him, and he, the living player, underestimated the apparatus's intelligence. Thus the electronic gambler becomes richer and wiser, and this cannot be said of the living one...

Two great wars embedded in our memory the name “Torres y Quevedo.” This Spanish scholar from before the First War did not have electronic technology at his disposal, but he constructed an electrical player mating with king and rook against a solitary king, led by a living partner who positioned all three pieces arbitrarily at the beginning; the electrical player always won in the shortest number of moves determined by theory long known to living chess players. Ulam and Stein decided to investigate whether N. Wiener was not mistaken in his prophecy that electronic computers would equal skilled chess players of the “homo sapiens” brand. In a report published in 1957 they explain why computers still play very poorly. If a machine is to play tolerably, it must - like a human - look into the future at a distance of at least three of its own moves and as many opponent responses; this corresponds - according to rough estimation - to 64 million chains connecting the initial state with what will be on the chessboard after those six moves; even the most efficient machines would have to ponder several hours over each move, so that the game would drag on for whole weeks...

How does it happen that on 64 squares man still reigns unchallenged over computers? After all, the imperfection of human memory is - unfortunately - well known to us! How then does man find in the course of one or two minutes a move that radically destroys a position worked out by the opponent over many hours? He manages only because he knows how to think in broad strokes: when he sees, for example, that his king finds itself in a difficult position on the right wing, he will not think about a possible move of his pawn on the far left; he also won't count on such an opponent response whose pointlessness he sees directly - this already reduces the field of choice so radically that the necessity of reflection rarely occurs. In this art of distinguishing essential things from inessential ones, and in the art of painting the situation with a few broad strokes chosen from thousands of details, which the machine analyzes consecutively all without exception, the chess player maintains his superiority over the computer. This is a strange human advantage, of which until recently no one was aware, so that it still lacks a name: walking on uneven terrain, estimating by eye the width of a ditch, guessing from a few strokes whom a caricature represents, understanding the content of a letter read hastily - these are examples of specific human efficiency that machines would env y, when they someday learn to env y. Following this line St. Ulam proposed “synergy,” which consists of division of labor: man divides the field of vision into several sectors and classifies them according to importance; he rejects those he easily recognizes as meaningless in the chess struggle or other problem requiring his decision, and from the remaining part of the field he selects several possible decisions and orders the machine to study their consequences - in this way he can choose the best one and after a few moves without the machine's cooperation again review the situation. In bridge, the machine should be limited to matters of probability, and man should be left to choose the card after hearing the electronic prompter's report on card distribution. A universally known example of synergy is the cooperation of a car driver with the motor, which handles the energy aspect, leaving the driver to choose direction. When we compare Torres y Quevedo's apparatus with a computer playing chess with a full set of pieces, the question arises why the old-fashioned electrical apparatus manages splendidly, while a computer a thousand times faster is so helpless? Simply because chess players long ago solved the problem of the game of king and rook against solitary king and provided rules determining in each situation of such a trio the best move for White - since Torres's apparatus plays White, it was enough to build these rules into the apparatus; there is nothing similar in full chess, because it's not even known which color has theoretical certainty of not losing (although it's known that there is such a color). War differs from chess in that here the chessboard is visible only through thousands of telegraphic, telephone, radio-telephone, and other communications, and the so-called theater ofwar encompasses entire continents and oceans; therefore some experts on contemporary war problems want to raise the significance of “common sense” of commanders at the highest levels; one expert, Sir Solly Zuckerman, fears that overly refined automatization of all decisions might - given the enormous number of parameters whose precise knowledge is impossible - lead to fatal errors if everything depended on an inexorably logical machine and some surprise occurred whose importance the machine would not notice, but which any reasonable and experienced commander could easily assess if he didn't have to trust machines more than peasant reasoning. In short, the English expert would be a supporter of synergy if he knew such a theory exists, and if it were mature for practice - one can doubt both assumptions. But how can one rely on peasant reasoning when the main food of such reasoning is experience, which no one has had the opportunity to acquire - no one has seen atomic missiles and rockets in action encompassing populated areas and reacting as living targets. Today's war, in which one side believed in the infallibility of machines and the opposite in the advice of Mr. S. Zuckerman, would be like a chess game in which on one side of the board sat a computer and on the other an inexperienced chess player - as we know from Ulam and Stein's studies, the chances are then equal...

The NATO expert also regards modern game theory with distrust, which he considers a chapter from statistics and probability calculus, in which he is wrong. Given that we (fortunately) know very little about future war, we must - if we already accept the possibility of such war as a subject of reasoning - learn from experiences during the struggle; we cannot entrust such registration of events and constant “feedback” to anyone except the machine of the future - the owner of this machine will find himself in a similar situation to the constructor of the machine winning 55% to 45% in heads-or-tails.

The end of the 19th century repeated “Ignoramus et ignorabimus” - we do not know and we shall not know - these words of the famous physiologist Emil du Bois-Reymond were the title of a list of unsolvable problems. A century earlier, Immanuel Kant called them antinomies of pure reason. Did the world have a beginning, or was it always? Will it last forever? Is it limited or boundless? Can living substance arise from dead, or only from living? The appearance of electronic computers reminded us of these questions of the Berlin physiologist, who placed the answers beyond the boundaries of science, as inaccessible to human reason. Against the background of these classical dilemmas, the question of whether an electronic computer is a living being will seem to everyone comical in its absurdity. It is worth posing, however, because it forces reflection on what is the characteristic feature of living organisms. Ever new discoveries of mechanisms in the organic world and in man himself awaken in us age- old dreams of the homunculus; if man is a machine, then why shouldn't a machine be man?

The name “electronic brain” displaced from the dictionary “robot” from 30 years ago - both are expressions of the same aspiration, and Adam molded from clay is its oldest reminiscence. This metamorphosis is unidirectional: to make the living from the dead - the opposite transformation is a curse and punishment. Epochs past are also familiar with the coexistence of free will with absolute determinism - Islam is an example of this.

The aesthetic criterion fails, because some people (but almost exclusively men) are inclined to see specific beauty in machines - Huysmans, whose famous “À rebours” closes a certain era, was the first to notice this techno-romanticism of the generation that followed, and the genius creator of expression prophesied the defeat of the generation that would sell itself to the machine. And perhaps body too? Who wins in the war of machines against man? How many people died, how many machines in recent wars? This statistic is known, though machines don't boast.

Naturalists are inclined to recognize as a criterion of life the metabolism of matter between a living being and its environment, as well as the continuation of living individuals through reproduction. I would add death here as a feature of life: everything that lives will die. Spontaneous movement is not a necessary attribute of life, but a su fficient one. There are machines that can pretend to life by virtue of movement. The name “automobile” literally means something that moves itself; the translation “samochód” [self-walker] is consistent with this name - it is very significant that the first witnesses of automobiles saw in them precisely what is most essential: previous experience had taught them that a wagon requires a draft animal, which walks by itself because it lives; amazement at the sight of a wagon that also walks by itself found expression in the name. Anyone who knows cars in the West, especially in the United States, must agree with the assessment of local sociologists that no rational reasoning explains such widespread distribution of these machines as we see there. There are practically no roads for pedestrians, and streets are so crowded with cars that serpentine overpasses must be built for them. Admittedly, the price of a car, especially a used one, is not high, but the costs of maintenance, repair, insurance, and especially parking are burdensome. Driving a car over long distances is exhausting, and the trouble with it in a foreign city is considerable. This irrational relationship to the machine can only be explained by seeing in it something more than a soulless lump of matter. I know only one example of machine dominance over man even more complete than cars: the rotary press - here, however, the dominance is indirect, through the product, which also took on the character of the press; this example requires a separate study that cannot fit in the margin.

Spontaneous movement requires free will. Does a machine have it? Let us ask whether man has it. I was once a witness to a post-hypnotic experiment, that is, one in which the subject is to perform after awakening certain orders imposed on him during hypnosis. The experiment was arranged so that all possibilities of deception were excluded. After awakening, the subject behaved completely normally. Those present knew, however, when the time would come to return home and that then the subject would demand to be lent a pen lying on the table - the host deliberately made difficulties, but to all his arguments the subject argued and insisted on its will, which was alien - was it not then similar to a stone kicked by a tourist's foot from a mountain slope? And perhaps such a stone has consciousness of free choice and if it could speak in human language would say: “I'm going downhill because I agreed with a fellow stone!”

After all, Aristotle explained the falling of bodies by the predilection of all matter for the vertical downward direction. For materialists there is no sharp boundary dividing the world into organic and inorganic - it's only a matter of complexity, because organic cells surpass in structural richness everything we consider dead matter. The liver consists of millions of cells, each of which is such a complex assembly of various elements that an electronic computer is something very primitive compared to this machinery - the physiological tasks that fall to the liver are diverse and difficult; wanting to solve them, nature did not try the power of the magical “because it is a liver,” but built an enormous complex of miniature automated factories, which is proof of the universality and unity of nature's laws. Electronic machines are incomparably poorer technically than these factories; let us not wonder then that they play chess poorly, but let us note that nature too does not have brilliant shortcuts when it must solve difficult problems. Believers in Scripture resolve the controversy between living and dead matter by giving a living and immortal Being the power to create from chaos first cosmic order and then organic life, whose pinnacle is man; if so, it would not be strange if this man in turn could construct objects endowed with abilities of perception, coordinated movements, feeding, reproduction, decision, play, and struggle for existence. The recently deceased J. von Neumann, one of the leading mathematicians of the past two decades, posed the problem of whether an electronic machine can produce another better and more efficient than itself - he achieved certain results testifying to such a possibility. Von Neumann's name is linked with game theory, and games of chance with Pascal's creativity: he found the proper theory of gambling in the form of a separate doctrine, which is probability calculus. But von Neumann, and before him several other scholars, among whom the French mathematician and politician Emil Borel was the earliest, were interested in a different aspect of games than Pascal, namely the principle of optimal behavior. It would seem that one cannot determine such behavior if one does not know the principles by which the opponent is guided. But he too does not know what principles we hold to. Mathematics knows how to master such situations; game theory, unknown before World War I, became after the second an established branch of science. It is known how to escape the vicious circle through optimal strategy, but in real situations finding it requires long calculations. Therefore electronic machines assist game theorists - games are not only chess and checkers, but also disposing of freight cars on a national railway network scale, or determining drug strength from several dozen trials can be reduced to this theory.

The problem of what a machine can and cannot do appeared a hundred years ago in the dream of automata speaking with human voice - the good-natured barrel organ born from the lyre but deprived of poetic aureole, how far it was from the baritone heroes and soprano heroines whose names appear on marble tablets of theaters - those who placed them there did not know that it would be better to engrave with a needle on wax the voice of the famous Malibran than with a stylus on stone her name. Of all the machines that the second half of the 19th century gave us, the gramophone is the simplest: a rotating disc (first wax, then ebonite) and a stationary needle, whose tip rests on the groove, and the other end is embedded in a metal tube. Despite this simplicity, Edison's invention could be so perfectly realized already in prototype that when it was demonstrated for the first time in the Paris Academy of Sciences, many academicians looked around for who among those present was a ventriloquist - doctors appealed to their authority in matters of throat, larynx, vocal cords, etc., and ruled that this natural phonetic apparatus has so many wonderful properties that no machine, let alone some disc with a nail, would ever speak with human voice given to us by the Creator.¹ Again the force of faith in the hereditary privilege of Adam's descendants was revealed: man speaks because he is man, and no machine can speak with human voice. But the machine spoke with human voice, and its mimicry was perfect - let the novel “L'Ève future” testify to the impression, whose hero constructs an artificial woman with Edison's help. Since the gramophone faithfully renders pitch and its timbre, individual features of pronunciation and voice, as well as the acoustic image of all emotions we hear in word and song, doesn't the artificial woman surpass the living one in that her answers are exactly those of which the adorator who built her dreams? But everything that evokes joy and despair, hope and doubt, lies in the disc in the form of a groove carved by the needle during disc recording. This trace can be represented by the relation y = f(x), where x denotes time, and y the depth of the trace... thus everything that gives the audience the charming voice of a tenor singing the famous aria from Tosca can be expressed through a continuous function of one variable, and the gramophone is experimental proof of this theorem. It fulfills a role reserved for human beings. Hence the reverse conclusion: man is a speaking and singing machine, and his organism contains - among many other apparatus - also a vocal apparatus. Today everyone knows this, but before Edison few would have accepted such brutal formulation of facts. Is it not worthy of philosophers' attention, this persistent defense of every human monopoly and inability to draw a general conclusion from defeats inflicted on us by ever new machines? To this question we will hear the answer that it was man who built them. This was indeed so, but there is no fundamental reason that would guarantee us eternal patent rights to all possible inventions about which we have not yet dreamed...

The gramophone is the acoustic equivalent of photographic film, and like it is only a recording and reproducing apparatus - it cannot be classified among universal machines, such as contemporary computers. Today's problems demand an apparatus that would read notes and lyrics converting them into song, as a singer looking at a score does... there are no fundamental obstacles on this path, though technical difficulties might be considerable - it's unknown whether one should rejoice or worry that the near future will provide us with a companion in the form of a box that would give banal answers to our banal questions... materials for such conversations can be found in G. Flaubert's “Dictionary of Human Stupidity” - to the question whether a machine can think, we will receive the polite answer: “unfortunately not - we have no soul” (nb. the term “electronic brain” was invented by a journalistic brain as unrequited flattery for the machine).

The concept of feedback appears in calculating machines in the form of control devices; one could also imagine its use in gramophony. One can suppose that the principle of synergy will soon find application in computers. But not only these, here parenthetically formulated tendencies characterize our era.

The popularization of problems facing our generation is itself a problem. The number of scholars belonging to it exceeds their total in all previous generations - a popularization will struggle even to reach every scholar. Few humanists realize the essence of mathematical problems (recently I read a serious article in which one of the lightning calculators multiplying multi-digit numbers in a few seconds was called a “mathematical genius;” the author evidently did not know the work of French psychologists from the end of the 19th century, who noticed that among such calculators there is a large fraction of mentally deficient persons - perhaps even fewer are mathematicians familiar with the results of today's genetics).

We are like Croesus, who could not count his riches. We live in an age of explosion, not only atomic. European monarchs of the first quarter of the 19th century sent their couriers by relay horses - the Egyptian pharaohs also knew this method a few thousand years earlier; from the moment when man first mounted a horse until the moment of putting Stevenson's locomotive on rails, the speed of communication did not change, but that moment doubled it. Doubling the speed of the first trains required only 50 years - then came 8 such doublings of transportation speed, and the transition from jet aircraft at the speed of sound to rockets carrying cosmonauts around the earth required five doublings and was accomplished in six years - this means that currently man doubles his transportation speed every year! Population growth of the globe also takes on the character of explosion - currently the period of its doubling is about 50 years; if one accepted this period as constant, it would mean that before Christ there were no people on earth at all - this reasoning ad absurdum shows that the period of population doubling is decreasing rapidly. Economic, and therefore political, consequences cannot be grasped by the observer's eye: they will appear before us suddenly, as an accomplished fact. This near future will appear as a result of rapid technological development, which will be based on new energy sources. From 1643, when Pascal devised his machine, to arithmometers about 250 years passed, and from them to computers only 50; currently the doubling of digital machine speed also has an explosive character, so that what I write will be obsolete when it appears in print.

Thus it is difficult to predict the role of digital machines in the future. When cinematography appeared at the beginning of our century, viewers delighted in the sight of a train entering a station or a horse taking an obstacle - everyone thought that cinematography would show interesting events filmed in motion, but no one predicted that it would become a producer of artificial events, and thus theater in universal and commercial format. Today's computers calculate pensions and workers' wages or look for contradictions in tax testimonies - one can hope that the fact of machines' existence will evoke other questions, unknown today, but incomparably more important...

Will they solve all problems? I think not.

Will they solve every problem? I think so.

How will the game between man and machine end? I don't know, but I know it's a game with infinitely high stakes...

1. Via an Edison collection on rutgers.edu, one example of the response was a college professor in New Haven, who wrote to Edison urging him to repudiate the hoax before it damaged his reputation: “The idea of a talking machine is ridiculous, but…some persons are so ignorant of the first principles of science, that they will be apt to believe it true, unless you deny it.”