English translation of the French transcript, made with AI and not yet proofread — click a paragraph to jump to it. Show the French original
Musical introduction: Kempff plays the Moonlight Sonata 0:00 Music
Hard sciences, human sciences, and artificial intelligence 1:42 Well, good evening, dear friends, good evening everyone. I'm very happy to be with you again, as usual, even if for some of you it might get a bit confusing, given… I'm told I don't stop producing content, well, listen, you shouldn't have pushed me, since you wanted me to make short formats. So we multiplied them a bit, and I want to thank Pierre-Alain, by the way, who takes a lot of his time, and he has other things to do, and he posts them online as we record them.
2:20 But this is our regular Wednesday evening workshop, once every two weeks, let me remind you, in case some of you are jumping on the train mid-journey, so there's a continuity to our series, so to speak, about the possibility of instilling into the so-called human sciences, the social sciences, the sciences of history, as we prefer to say, the possibility — we're not there yet — of establishing there the rigor that gives all the prestige, absolutely not undeserved, at least in that respect, to the so-called hard sciences.
3:00 That is to say, let's remember, this is my own framing here, the sciences of nature, plus mathematics, which are not sciences of nature. There you go. You quoted a little excerpt from Einstein the other day, who was surprised precisely that the sciences of nature became effective, that man was able, right, to predict, to appropriate, and given Descartes's dream, which we'll talk about again in a moment, to make man master and possessor of nature, starting from a discipline, right, that is based on notions which are neither natural, nor even sensible, nor accessible to the senses, like unity, number, the line, the point, the plane, to name only the simplest ones.
3:44 So this is the continuation of this series and it's also the continuation, as we promised last time, of a chapter of this series, on the relationship between hard sciences and human sciences, centered on artificial intelligence which, you don't have to be a wizard to realize, is situated precisely, almost always, at least in the discussion surrounding it and that it provokes, at the boundary of the two fields, right, of hard sciences and less hard sciences, right, to put it as kindly as possible.
4:26 So, we're going to resume our dialogue with Christophe, but first, I'd like a little animated sequence, after the introduction by a sonata, an excerpt from the finale of the famous "Moonlight Sonata" by Beethoven — it's Christophe's choice, and I'll ask him later why he chose this sonata — to continue our discussion on artificial intelligence. These are the last measures of the finale of opus 27 number 2, the so-called Moonlight Sonata; Beethoven never called it that — whenever there are extra-musical, literary notions like moonlight, it never comes from the composers, or so rarely it's worth mentioning; it comes most of the time from the publishers, who know that it sells to the public.
5:24 But before, beforehand, before our dialogue with Christophe takes place, which will be the essential moment of this session, I'd like to show a little filmed sequence, as a second part of our opening after the music, if you will, where we'll find again a film I had already mentioned — we showed the trailer for it a few workshops ago — it's a Space film, you'd say today a fiction or anticipation film, as it was called at the time, which is a reference among film buffs, so I'm not competent enough myself, but I love it a lot, I'm happy when film buffs tell me, well you're right, it's an excellent film — it's Forbidden Planet, right?
Forbidden Planet: the robot and the thinking machine 6:20 Forbidden Planet, it really is the literal translation for once, and in it there's a robot who is a bit of an ancestor of the robots you know much better from Star Wars and others, there you go. Except he's more archaic, they didn't have the tricks, right, of your era, and the special effects. But I'll spare you — he appears very often in the film, but there are two sequences I'm particularly fond of, and one I recommend, but watch it, Forbidden Planet, try to see it in full, because it's worth it.
7:02 But, including perhaps especially philosophically, there are two sequences I'm particularly fond of. It's right at the beginning, there's one of the crew members who comes from planet Earth, of course, NASA people basically, right. And among them there's the cook, who isn't the sharpest of the crew, as it happens, but who is at least the most likeable, well, the most likeable to me among the crew, because he's human, unlike those astronauts without a shred of feeling, who were recruited precisely because they had no feelings, I don't know.
7:35 With that composure — next to which even British composure seems far too emotional, that sort of thing. With those steel-blue eyes. And he's really a guy who's hit the bottle a bit. And when he saw that the robot on the planet where they land can reproduce anything it has analyzed — you just put into the little hole it has below what's supposed to be its head a sample of whatever you're holding, let's say, I don't know, you have chocolate, and poof, it makes you 20 chocolate bars in a few seconds.
8:12 Except that this time he gives it a little bottle of whisky, I don't know, and the robot, after saying, yes, there's fat, there's stuff, well anyway, and then 50 bottles of whisky come falling out — you won't see all that in detail. And then there's a very important sequence, all right, let's be serious now, but I was doing a sort of aperitif sequence, if you will, that's the word for it, but to get to some very important things. Here we are in the house of the earthling who lives on the planet and welcomes them, who is the sole survivor of an earlier NASA team that had already come — they're all dead — and precisely the purpose of the film, of the new team that arrives, is to see what happened, and the sole survivor is a scientist who lives there, alone on the planet with his daughter.
8:58 It's funny to see a pretty Hollywood star, a blonde, there, on this planet where there's absolutely nobody, except him and his daughter and a few animals. And there, he has capitalized, so to speak, on the manuscripts of a civilization that was on this planet long before our time and that allowed him to achieve things unthinkable on Earth, like this robot — we're going to talk about artificial intelligence through this robot — and also, very important for what follows, a machine that lets you, when you place your head on it, project in front of you, a bit like a hologram as we'd say today, whatever you're thinking of.
9:40 There's even a comic little moment because the head of the crew is asked to use the machine, and we see in the little jar, as in a crystal ball, we see the other man's daughter, of course — that's what was occupying his thoughts. So these aren't digressions. We're getting to the heart of the matter. And here we come to the sequence we're going to show you, the one where this earthling who welcomes them, who is their predecessor on the planet, explains to them a bit what I just told you, these civilizations he found and of which he realized a few small applications that are very far from the level of knowledge and technology they had, but which are already fabulous for us.
10:24 And in particular this robot — he presents it, shows its capabilities; it can do anything except destroy the person, its master, the one who made it. It's programmed in such a way that if it's given the order to destroy that person, it malfunctions. I won't say more, I'll comment afterward. But for now, let's play the sequence.
Asimov's laws, the piano-playing robot, and Kant's nightingale 17:24 I'll hand the floor to Bertrand, who sent me a very interesting article, I believe there's an excerpt from research, but he'll analyze it, I think, much better than me. I'll hand it over to him right away. Or if he'd rather we read Descartes's text first. Christophe, tell me… First of all, hello everyone.
17:58 In the excerpt, it echoes a lot Asimov's laws in the field of robotics. In fact, he was a science fiction author. And on robotics, there are also rules I can briefly recall. There are actually three rules. The first rule is that a robot may not harm a human being, nor, through inaction, allow a human being to come to harm.
18:30 The second rule is that a robot must obey the orders given by human beings, except where such orders would conflict with the first law. And the third law, so these are Asimov's three laws, is that a robot must protect its own existence as long as such protection does not conflict with the first and second laws. Now, on the first law, just to introduce it, it can be seen as incomplete, because a robot could harm a human being without knowing it.
19:06 An example, for instance, imagine a robot puts poison in a glass of milk, the first robot, then a second robot puts milk into that same glass, and the human sees it and dies or gets very sick. And perhaps the first robot was just programmed, it put poison in a glass of milk without any intention.
19:36 So, regarding Asimov's rules too, what's also quite interesting is to see that it's a science fiction author who also fed into science and robotics. So there are often back-and-forth exchanges between science fiction and science. Not just from a social point of view, because we often also see science fiction as imagining the evolution of societies — what's it going to become, are we going to end up in a Mad Max world, or I don't know — but it also gives rise to reflections on things, with the technological progress that we thought was truly in the realm of science fiction and that then comes into science.
20:25 And on the other hand, there are other things we perhaps dreamed of, and it's always interesting to look in the rearview mirror, to imagine how people saw the year 2000 back in the 1960s, in the sixties, and it didn't happen. So, regarding Asimov's laws, that also raises another concern: if we use machine learning methods, we can't be sure these rules are being verified, since we don't know the underlying workings.
20:59 So there are certain — we need to use other rules, more formal things to apply them. There you go. So, about the robot, there's been a lot. One last thing too, maybe important to note, is that you see, here, the robot doesn't have a very human shape. And there have been studies done, but I don't remember the name of the Japanese professor, that show we can really have, we can become aggressive.
21:35 In fact, there have been robots, for example, that were attacked by humans, destroyed. Actually, because they resembled — they resembled humans a lot, but there was a small flaw that made an imperfection noticeable. And that small imperfection also caused anger on the part of the human being. So there's this aspect too: a robot that resembles a human being closely, when there's a noticeable imperfection or something like that, triggers a rejection from the human.
22:14 And on the first one, I think there might have been a small technical problem, but on the sonata, the question we can ask ourselves is: would we take as much pleasure listening to this sonata if it were performed by a robot, a robot pianist? Why not? And that maybe raises the same question: do we take the same pleasure listening to a mechanical piano, which might be very optimized, compared to a pianist who plays?
22:56 So, what can we draw from this? Actually, well, here I'm going to draw on some work — we had started doing work with my colleague, Judith Nicogossian, on communication systems. Who is an anthropologist, yes, an anthropologist. A specialist in the field of health, of new technologies. And who had recalled that, in the Palo Alto school, in fact, they had established — because she also works on the anthropology of communication — that there is digital communication.
23:34 For example, here, we could say it's the keys pressed by the pianist. And there's also analog, affective communication, which is more fuzzy, for example the face, etc. And both take part. There you go, in things. And it's also, I imagine, the imperfections, the changes in tempo, etc., of the pianist, that make us take pleasure in listening. And afterward, we could also ask other questions. Maybe, one day, we could discuss it.
24:04 Are we talking about music? We could also talk about poetry. Because you've seen that we can reproduce — the learning machines that are a bit more sophisticated than supervised machine learning can, for example, reproduce a master painting. But then the question is: does that remain art? Does that remain… if it reproduces a poem, does that remain poetry?
24:34 Or not. And the question isn't really, I think, obvious to settle. Well, the problems are laid out, and even clearly, thanks to these examples which speak for themselves, and which are going to lead us straight to the text we're going to work on, which is a text by Descartes. You'll tell me, but why bother taking a 17th-century text when we're talking about cutting-edge technology? But that's precisely what's damning for a large part — Christophe won't contradict me on this — a large part of the discourse surrounding artificial intelligence, which believes itself pertinent, raises issues it thinks are current and which are in fact almost as old as humanity itself, which is what we're going to see with Descartes's text.
25:21 By the way, before even getting to Descartes, I wouldn't want, going further back in time, to skip over Kant's head, but Kant already said, regarding your electronic pianist, your machine pianist — the example is famous in the Critique of Judgment, in his aesthetics, I alluded to it not so long ago — I'm walking in a park, I hear the song of a nightingale at dusk, I'm captivated by the beauty of this song, when the owner of the park who is hosting me tells me, "well, come back down to earth, actually it's a mechanism, an automaton that I hid in the tree," and there, he says, "suddenly my pleasure disappears, I'm very disappointed, I thought it was a natural nightingale." There, that already says a lot — what you just said, you see that I'm not solving any of these problems, but I'm showing you how long they've been raised.
Reading of the Discourse on Method, part five 26:13 And now I'll get to Descartes's text — if you'd be so kind as to display it, Pierre-Alain, I'd like to read it. Thank you. So, Descartes — I'll tell you right away what comes before the first words — it's the excerpt from the fifth part of the Discourse on Method; it's interesting to remember that this discourse was written, it appeared in 1637, so one year after the condemnation of Galileo, so Descartes is very cautious — I say this because he makes a small allusion to God at one point, he has no wish to alienate the church people who couldn't stand him, by the way, incidentally, he was watched very closely by the Jesuits and he knew it, and his motto, let's not forget, was "I advance masked" — I say this as a voice-over, but it's not irrelevant, especially with regard to what he's about to say.
27:05 Just before, he spoke about the body, about animals, etc., within an explanatory framework that we would since call mechanistic, right — it was all brand new since Galileo, with mechanism alone, that's classical mechanism, there's only extension and motion. There are two fundamental concepts, there's extension — space, as we'd say today — and motion, that's enough to explain everything regarding the order of bodies, of course, we're not talking about what's supposed to be thought, etc.
27:38 And so, once he's given a few examples from the animal world, he continues, and now I'm reading his text: "This will seem in no way strange to those who, knowing how many various automata or moving machines the industry of men can make, using very few parts in comparison with the great multitude of bones, muscles, nerves, arteries, veins, and all the other parts found in the body of every animal, will consider this body as a machine which, having been made by the hand of God, is incomparably better arranged and has in itself movements more admirable" — he's speaking of animals here — "than any of those which can be invented by men."
28:30 "And I had stopped here in particular to show that if there were such machines" — you see, he's anticipating today — "which had the organs and outward shape of a monkey or of some other unreasoning animal, for instinct does not have reason, we would have no means of recognizing that they were not of the same nature as these animals." There's nothing that would distinguish this machine from animals — all right, I said I wouldn't do voice-over. "Whereas if there were any which resembled our bodies and imitated our actions as closely as morally possible, we would still have two very certain means of recognizing that they were not for all that true men."
29:11 Two points, that's the word for it. "The first of which is that they could never use words or other signs by composing them, as we do, to declare our thoughts to others. For one can well conceive of a machine so made that it utters words, and even utters some in response to bodily actions which cause certain changes in its organs — as, for example, if it is touched in a certain spot, it asks what one wants to say to it, or if in another spot, it cries out that one is hurting it, and similar things — but not that it arranges them in different ways to reply to the sense of everything said in its presence, as even the dullest of men can do."
29:55 That's the first point. "And the second is that, although they might do several things as well as, or perhaps better than, any of us" — think of calculators — "they would infallibly fail in some others, by which one would discover that they did not act through understanding, but merely through the disposition of their organs. For whereas reason is a universal instrument which can serve in all kinds of situations, these organs need some particular disposition for each particular action."
30:28 "From which it follows that it is morally impossible for there to be enough different organs in a machine to make it act in all the occurrences of life in the way our reason makes us act. Now by these same two means one can also know the difference there is between men and beasts. For it is a very remarkable thing that there are no men so dull and stupid, not even madmen" — this refers here to the insane, crudely speaking — "that they are not capable of arranging together various words and composing from them a discourse by which they make their thoughts understood; and that, on the contrary, there is no other animal, however perfect and well-endowed it may be, that can do the like. This does not happen because they lack the organs, for we see that magpies and parrots can utter words just as we do, and yet cannot speak as we do, that is, showing that they think what they say — whereas men born deaf and dumb, who are deprived of the organs which serve others for speaking as much as or more than the beasts, are wont to invent for themselves some signs by which they make themselves understood to those"
31:38 "who, being ordinarily with them, have the leisure to learn their language. And this shows not only that beasts have less reason than men, but that they have none at all, for we see that very little is needed to know how to speak; and inasmuch as we notice inequality among animals of the same species, just as among men, and that some are easier to train than others, it is not credible that a monkey or a parrot which was among the most perfect of its species should not equal in this respect a child of the dullest sort, or at least a child with a disturbed brain, unless their soul were of an entirely different nature from ours. And one must not confuse words with natural movements, which testify to the passions and can be imitated by machines as well as by animals; nor think, as some ancients did, that beasts speak, though we do not understand their language, for if that were true, since they have several organs corresponding to ours, they could make themselves understood by us as well as by their fellow creatures. It is also a very remarkable thing that although there are several animals which show more skill than we do in some of their actions, one nevertheless sees that these same animals show none at all"
32:48 "at all in many others, so that what they do better than us does not prove that they have intelligence, for on that reckoning they would have more of it than any of us and would excel in everything else, but rather that they have none at all, and that it is nature which acts in them according to the disposition of their organs; thus we see that a clock, composed only of wheels and springs, can count the hours and measure time more accurately than we can with all our prudence." End of the excerpt. So, I'll start over from the beginning if you don't mind, Pierre-Alain, I'll go back to the beginning in a shorter way of course, but still with a few comments. First, from the second line, we already speak of automata — and yes, this was all brand new, these are the very first automata appearing, as it happens, at the moment the scientific revolution — that's what historians call it — had just occurred, in Italy first, right, with Galilean physics, the Galilean break as Bachelard's school calls it, for example, right — and which is the first irruption of applied hard science in the history
The first automata and the Galilean revolution 33:59 of the sciences, in order of appearance, since, well, 2000 years earlier mathematics had appeared, but here we're talking about applied sciences of mathematics — and I don't know if it matters for Galileo, who said that if there wasn't a mathematical law, a mathematical explanation, the experiments he could perform were not enough to convince him of the validity of the physical laws he was deriving — so for that I refer you to the admirable Galilean studies of Alexandre Koyré — I'll apologize for last time, I've been citing for 20 years Canguilhem's History of the Concept of Reflex, I don't know why last time I attributed it to Koyré — here it really is Koyré, the Galilean Studies, I refer you to that — it's not our topic directly, but anyway everything Descartes says here presupposes this scientific revolution that had just occurred, this mechanical physics which retains nothing, which no longer looks in bodies for a soul, a hidden will, mysterious occult forces as they used to say, but only matter and motion, or extension and motion, extension being the category
35:10 of everything material at that time in physical thought. Now, immediately the hypothesis arises — and not only the hypothesis, they had already begun manufacturing the first automata of the time, which would soon give rise, barely a century later in the 18th century, to a mechanical chess player — not yet electronic but still mechanical, right — so, "to those who, knowing how many various automata or moving machines" — you see, still motion, mechanical in Greek — "the industry of men can make, using very few parts," well, "in comparison with the great multitude of bones, muscles, nerves," well, in other words, imagine, imagine a machine that had as many parts as any of the animals, right, that surrounds us — it would be even more incomparably efficient, and we'd have even more trouble distinguishing it from the living animals around us. That's what he's saying, right. And so I skip to the paragraph that comes after, you see, and "I had" — you see, at the end of the line — "and I had stopped here in particular to show that if there were such machines"
36:20 "which had the organs and outward shape of a monkey" — they hadn't yet seen Planet of the Apes — "or of some other unreasoning animal, we would have no means of recognizing that they were not of the same nature as these animals" — in other words, we couldn't tell them apart from animals if these machines were perfect, if their mimesis, as we say, their imitation of animals, was very well done, we couldn't distinguish them from animals — "whereas if there were any" — let's apply this now to human beings — "which resembled our bodies and imitated our actions as closely as morally possible" — and here's where the symmetry breaks down — "we would still have two very certain means of recognizing that they were not for all that true men." Now, before recalling these two means, I mean, I can already see those who will say, yes, Descartes doesn't want to give up on anthropocentrism, he absolutely wants to preserve some specificity — well, listen, I recommend you reread Rimbaud's A Season in Hell, that brings us much closer to the modern world already.
Something and someone: against naturalism 37:30 And Rimbaud, in A Season in Hell, says at one point that sentence — it's so short, he has the genius of saying the maximum of things in the minimum of words — Rimbaud: there's no one here and there's someone. So when we're incapable, aren't we, of distinguishing something from someone — you're free, if I come into your office and I come to shake your hand, ask about your children, you're free to tell me that I'm a something. Myself, I mean, that's not where I put my self-respect, but still, I mean, we have to own it if there's nothing that distinguishes man — because it's very fashionable, that's what I'm in the process of saying, there's a prejudice that's as old as the world, moreover, which is naturalist, it's probably the oldest of all thoughts even though it believes itself the most up-to-date, especially when it gives itself pseudo-scientific guarantees — there is this thought which consists in denying the difference between something and someone. But personally, I'm willing — I don't decide, how could I decide — but you have to own it. The Nazis owned it, didn't they, for them people, their prisoners, were somethings.
38:40 If you weren't first of all of the right race, you weren't even a someone yet, you were a something, and you know that the notion of man was vigorously attacked by Nazi thought. The notion of man is that of universal man, with his universality, that is to say applying just as much to pygmies, to Japanese, as to blond dolichocephalics — was fiercely attacked by the Nazis, in the continuity of an old European thought that appears as early as the counter-revolution after the French Revolution, and which was first formulated by Burke, and would be taken up again by all our French counter-revolutionaries in the style of de Maistre and so on, which consists in saying: I have met in my life Irishmen, Iroquois, Frenchmen, Bretons, but I have never met man in general. You see, it's an old file. Why did I make this parenthesis, why did I dwell on this comment — because the word universal is not going to be long in appearing on Descartes's pen, and Descartes is ahead of these thoughts that remained, which I've just denounced, isn't he?
First criterion: composing signs to express one's thought 39:47 Well, which I've just recalled are very ancient — they may be right, we have to own the political consequences, we really have to own them, we mustn't play on every board. I say that because now I've seen people who claim to be on the left, on the left, and who value Popper — fine by me — and neopositivism in general. So I go back to Descartes's sentence: 'We would still have two very certain means to recognize that they would not for that be true men: the first, that they could never use speech.' [laughter] I can already hear some contemporaries: 'but he doesn't know about dolphins and bees' — those are people looking behind them, pointing at the straggler who is far behind them, when the person speaking is far ahead of them. They would need to turn their heads like Janus. Descartes is well beyond their thinking. 'The first, that they could never use speech, nor other signs by composing them' — he speaks of signs, not signals, like dolphins or bees, 'as we do to declare our thoughts to others; for one can well conceive that a machine could be so made' — well made, implied — 'that it utters words.' You see, he seems to contradict himself — no, a machine can utter words, and even utter some of them.
40:57 'Regarding bodily actions that will cause some changes in its organs' — look how nuanced he is, he's no longer content to say it can't utter a word — not just the animal, but even a machine can utter words, Descartes is saying. We're at the very heart of our problem of artificial intelligence. 'As if, being touched in some spot, it were to ask what one wants of it; if in another, it were to cry out that one is hurting it.' So Descartes, back in the 1630s, anticipates the fact that we can very well build machines that cry out when we're supposed to be hurting them, or pressing on a sensitive spot, etc.
41:33 'But not that it should arrange its words variously so as to respond to the sense of everything said in its presence' — we get to the universal there — 'as the dullest of men can do.' And the second — here we jump right in — 'is that although it might do several things as well, or perhaps better, than any of us' — they give the example of the clock at the end, but I'm thinking of powerful computers with staggering computing power, aren't I, doing billions of operations per second, I think — Christophe, don't add too many zeros — you can add even more, I thought I was making a hyperbole but we're always still short of the true, of the real — 'and the second is that although it might do several things as well, or perhaps better, than any of us' — mustn't take Descartes for a fool — 'it could not do them all' — that's what he's going to say — it can only do that. The machine that plays chess and beats the world chess champion, it can only do that; it's not also going to play Beethoven's sonata, and above all it's not going to compose one. I'm very pleased that Christophe quoted his anthropologist colleague who says there are also analog components, you were saying, that get communicated
Rubato and the diversity of performers 42:44 between the human pianist, this time, and the audience — and I had the word on the tip of my tongue when you quoted your colleague — it's a term musicians, musicologists, of course, but even many music lovers know: what's called rubato in music generally and in classical music in particular. Rubato, 'to rob,' steal, in English, isn't it — remember in the Ring, Das Rheingold, one must steal the gold from the Nibelung, from the proletariat — rubato means what is stolen. What's stolen? The tempo. It means that a tempo that's too regular, too metric — because you can tell a machine, right, go play this Beethoven sonata at 125 eighth-notes a minute, that's doable, that's not a problem, that's not a problem, anything quantitative the machine can do perfectly — but not that the machine should choose the spots where you need to slow down a little. There are no two pianists who have the same performance, precisely, including among the greatest — and I take this opportunity to say, Christophe, you served us the interpretation
43:55 of Wilhelm Kempff — I appreciate that very much, because those who know this repertoire know that Kempff is one of the finest achievements of the Beethovenian repertoire, and not only that, he's also an immense Schubertian — I recommend his complete Schubert sonatas — and I was saying earlier, off record, before we started our broadcast, that Kempff had training as an organist, which is quite rare among pianists, which gives his playing this horizontal clarity, because with the organ, as with the harpsichord, you can play loud or soft and it doesn't change the volume, so you're forced into a perfect clarity in horizontality — that transparency which suits so well works as complex as Opus 27 No. 2, which we started with despite the bugs. So, there is — but our luxury, we who consume this music, or even who sometimes try to perform it, much worse than these great performers — we have the luxury of being able to hesitate between the interpretation of a Kempff, of an Yves Nat — some say he's the greatest Beethoven interpreter, he's French
45:05 appearances notwithstanding, it's not only Germans who play Beethoven — of a Schnabel, who was the first to record the complete Beethoven sonatas back then on 78 rpm records — I'm not saying one is better than the others, there I'm in the Himalayas, all these are peaks beyond comparison, and our luxury precisely is to hesitate according to our own sensibility — one will find this one a bit more introverted, and that one perhaps better in rhythmic and dynamic aspects, and that one perhaps more lyrical — but each time it's fabulous. How could you want a machine to have this diversity of affect and sensibility? So I'm very pleased, Christophe, that you cited the analog parameters, so to speak, that your colleague was alluding to. So I go back to Descartes quickly — I seem to be talking fast but there's a lot of material, and I'd like to hand the floor over to you all shortly, so I don't want to eat up your time. 'And the second is that' — yes — 'for whereas reason is a universal instrument' — and there the word is out — 'which can serve in all kinds of situations' — it's not that it's vague, it's that it applies what he's just said
Reason as universal instrument and the abyss of desire 46:15 there we really are at the heart of anthropology, because this doesn't apply only to knowledge, this faculty of the universal. The fact that a polar bear can't live long in the Amazon — it would be, I'd say, spotted like the white wolf, isn't it, its camouflage in the ice is what it owes its survival to — and a badger, in terrain that's all granite, is done for, because it's equipped to dig holes and burrows. So the animal is entirely determined by the particularity of its here, of its now, of what's rightly called in zoology its biotope — in the life sciences generally there are biotopes, temples too — whereas man lives in the rainforest, builds stations at the poles and even stations in space, in environments that weren't at all made for him. In other words, a fundamental difference: every animal adapts to the here and now that constitutes its immediate environment, whereas man adapts the environment to his will, whether it's a space capsule
47:25 a boat — already, you know, our ancestors who launched frail skiffs, which must originally have been tree trunks, to cross a river — already, navigating on water, we're not made to float on water, the human body, to begin with, you see. So this universal faculty is crucial, and I said that not only for what I just said, the difference between the universal and the particular at the cognitive level, but also — I'm going to surprise you — at the trade-union level, or, since I'm becoming topical, you know, in other words at the point where we start finding that the wage-earner demands too much. Because after all, a Neanderthal had solved the problem of food — he's one of the greatest hunters ever seen, the Neanderthal. If human action were guided by instincts, that is by survival, that is by biological necessity, we would long ago have hit a ceiling at Neanderthal, and maybe we'd even find Neanderthal quite sophisticated, since he made spears after all, compared to the animal world.
48:36 But if technical progress — I say this without any value judgment, you see, I'm being honest — whether one deplores it, finds it a catastrophe measured against wars, or finds it a path of liberation, a necessary but not sufficient path — technical progress, if it — and by contrast it is undeniable, it can be measured, that is, the time it takes me to make this object, isn't it, with a machine now it takes a few seconds, it used to take months of exhausting labor before — if this progress has no limit, if it never stops developing, it's because human desire is just as universal as human knowledge, as Descartes has just described it. That is to say, it's an abyss, human desire, which no limited and particular satisfaction, even that of a billionaire — look at Citizen Kane, how well the film illustrates this, isn't it — no particular satisfaction can come to fill. We're saying serious things here today, I say it for those who would want to suppress the human race — the Nazis wanted to suppress the Judeo-Bolsheviks
49:47 the Jews, the communists, the homosexuals, certain mentally ill, the people with Down syndrome, whatever you like, isn't it — but not the human race. Now there are those for whom the human race really is the one species too many, isn't it, on the planet. So there are those who want to suppress the human race, who say man is at most one species among others. That's what's at stake in what we're discussing here — it's important to reaffirm that there is a human race, which doesn't mean — and not transhuman either, that's what it doesn't mean — that we're going to treat animals like a machine or like inert matter. There are very beautiful passages where Descartes — of whom it's said that he speaks of animal-machines with appalling cynicism — not at all, Descartes denies that the animal has reason, but not at all that it has feeling, I'll refer you to those texts another time. So — 'whence it follows that it is morally impossible that there be enough diverse' — well, I won't spell it out, his reasoning is simple: if there were, in a machine of any kind, in a robot, even with the degree of miniaturization there is today, isn't it — if there were enough parts to speak the way Descartes
Speech, deaf-mutes, and Helen Keller 50:58 as there are occasions for us, isn't there, to make our reason function on the encounters of daily life — it would need, isn't it — it's so implausible that it's not even worth thinking about. 'Now by these same two means one can also know the difference there is between men and beasts, for it is a very remarkable thing' — it's not that there's no man so dull and so stupid, not excepting even madmen, who is not capable of arranging various words together and composing a discourse by which he makes his thoughts known' — you see, this is more precise than earlier where he said they emit words — no, but by which they make their thoughts known — 'and that on the contrary there is no other animal, however perfect and however happily born it may be, which does the like' — which doesn't happen — which they lack, for want of organs. And he knows very well that parrots — and not only parrots, there are birds that speak much better than parrots in terms of vocal clarity. I remember, in Geneva, I had gone into a watchmaker's shop, as it happens, and suddenly I hear 'good evening' — but just as I've just told you, 'good evening' — I turn around and I see a little mynah bird in a cage which, with its little beak, looks me straight
52:08 in the eyes and repeats to me 'good evening.' You know, I've never seen a parrot — that's it — happy and content, but this was 'good evening,' calmly, you see. So it's not that we lack the organs — language isn't just about being equipped to speak, because we can prove there are animals that aren't even equipped, but it's not enough to have this equipment. The proof, and Descartes gives it, is deaf-mutes from birth, who can't speak in the sonic sense, isn't it, in the auditory field, but who will find a substitute for this communication by which one can transmit thought, and not just express instincts, isn't it, with conventional signs — and the CON of convention proves indeed that it presupposes a relation between men, and not that things pre-existed our birth, isn't it, like heredity. And since we're there, I'll hand the floor back to Christophe afterward, I'll end with the example of someone quite forgotten today, Helen Keller — does that ring a bell, Christophe?
53:19 Well, Christophe, you're lucky to have your youth — if you had my white hair, you'd have heard of her, but it's been a long time since people talk about her. But she's someone I recommend you look up on Wikipedia, because she's someone who interests a lot of research. So, I say she's a little American girl who was born deaf, mute, and blind — the poor kid, it happens — but you must have heard about her, maybe forgetting — yes, but not the name — I no longer remembered the name — there you go, there you go, and you'll tell me — and who ended up a university professor, and who wrote books, that's what's extraordinary. So what happened? I'll tell you — well, I'll tell you as if I were the one decreeing it — no, let me tell you the most important thing for me, as I sum it up: it's love that intervened. Let me explain: she had had a maid, a servant at home, who had this brilliant idea — each time, since she was a little animal, deaf, mute, blind from birth, it wasn't just a grunt like that when she asked for water — one time the servant, instead of giving her water the way you satisfy an animal that shows it's hungry, took her finger
54:30 because that's all she had left, touch — that's where you see one must be a materialist, the sensible is unavoidable, at least one sense must remain — she took her finger and made her trace a circle in the sink, on the bottom of the sink, a circle with her finger, and every time she asked for water she did that, and that's how the kid gained access to the symbolic function. It started with water, with the word water, then with other things. And I tell you, in the end, you see, it's the love and attention of a servant that made this girl, who was set to remain, wasn't she, on the margins of humanity, end up with abilities that I probably don't have. There, that path — I give you back the floor, Christophe. Yes, but regarding the example, I don't remember the name, but it's true that what you're saying is really a topic that's very current, about empathy, the empathic relationship, etc. For example, to take an example about medical diagnostic tools, it's the fact that even if we have — to go back a bit to Descartes — even if we can have very sophisticated machines, etc., it doesn't yet manage
Empathy, otherness, the Turing test, and the Chinese room 55:40 I still say 'yet' — to reproduce, all the same, the empathic relationship between the doctor and the patient. I think we already discussed this last time, and I think there are several things I noted along the way that also resonated — there's the aspect, already I think there's an important notion too, that among human beings there's also the question of otherness, which we perhaps don't necessarily have with robots that all look alike, because that's also part of it a bit — look, when there are science fiction films, you see a whole row of robots that all look alike, and well, even for us, in a sense, that's a bit frightening. And you were also talking about the Nazi period, etc., and I think there's also an author, a philosopher, who is a bit the father — the father of ethics — who also spoke of the notions of the face, Emmanuel Levinas, who said precisely
56:51 there, and who spoke precisely of otherness, and that ultimately otherness — I remember, for instance, 'the self before the other is infinitely responsible' — which had a lot — well, the face is a metaphor for the other, and it has a relationship too — there you go, and one — yes, and which is asymmetrical, the relation is asymmetrical with the other — I remember, I had noted, I must not expect reciprocity. So why am I bringing this up? Because with AI systems, well, we can also get lost in our senses, we can get lost because we have an illusion of reciprocity. An example — this is doing a lot of damage, I believe, in Japan, myself — it's all — because we're afraid to commit, etc. — it's all these, I don't remember the name, somewhat android systems where in fact it's about having a wife or a husband at home, etc., with an illusion of reciprocity, but it's
58:01 and that it's a trap all the same, set by the designers of the machine — so there I go back to some reflections I'd had with Judith, because this plays on addiction mechanisms, right, in terms of reward and frustration. And so on this first point, I think it's also important to see, as you were saying about the universal, it's really the defect — not the defect, there's a difficulty — that's why we're not going to head, I think, toward strong AI. I recall, strong AI means the more or less full replacement of the human being, meaning we'd have a humanoid robot that could act like a human being, but there's something that's still very difficult in the field of research, and which echoes what you were saying about the universal: it's learning common sense. Now, learning common sense is something that's universal, common, that's linked — there you go — and that's really a very great difficulty.
59:11 Because we manage very well — and that's also what I wanted to say — what's rather striking, or not, but what's striking is that we see there's a continuum, as you presented so well, between Descartes and the reflections we're having now on AI, because you illustrated it very well, and there was also this aspect of the universal, there was also the question of how to distinguish a system of so-called intelligent robots from a human being. So we presented last time the famous Turing test, meaning we'd take the robot and the human being, put them in two separate rooms, and by asking questions, according to Turing, we could distinguish the robot from the human — but we can also get lost in our senses. And then there was also the fact of a counter-example by someone named Searle
1:00:21 and it's called — the Chinese room, where in fact you can imagine you're in a room, you're given instructions in Chinese, and on the other side out comes a translation in English. So you might say, well, the system is intelligent, it has learned to speak English and to translate — but the trick is that one could say that inside, in Searle's argument, it's rote learning, because there was a book that made a correspondence between Chinese characters and English. There, another — I think last time we also talked about the first myths — not the first myths, the utopias of AI in the years around '57, of reproducing human reasoning, etc. And what I wanted to talk about, because it also echoes a bit what Descartes said, the mechanistic side — for example, there's a project that
1:01:31 I don't remember exactly the date, but for example in 2013 there was a huge project, which is still ongoing but has been reoriented, called the Human Brain Project, whose mission, whose objective, was to reproduce the functioning of the brain — we have lots of data, we have computing power in the tera-, or soon exa-, range, there you go — and so it was said, and what's rather — there was funding of over 1 billion, and we find again, which is rather funny, we're still in the same thing, meaning we have more computing power, so we're going to say we'll reproduce the mechanism of the human brain simply
1:02:41 with a computer, with supervised machine learning, with rote learning — remember the stories about how I teach it to recognize a cat, I teach it to tell a dog from a cat, but it's by rote — and this got reoriented in fact toward more of the cognitive sciences, in fact we add knowledge, we add — and classical learning algorithms aren't effective, but we still launched into this project because there's something perhaps, I'd say, timeless, from a human point of view, that is, with machines we can reproduce, we can perhaps put ourselves a bit in the place of the creator, if one exists. So the last thing is that maybe we'll talk about it afterward, about consciousness — is it, yes — so what did I also want to say, I find
Mathematization of the world and the First Meditation 1:03:51 Yes, when you spoke of course of Galileo, but we won't get to it — it was also, as you said precisely, the beginning of the mathematization of the world, that is, we really stepped aside from the sensible world. For instance, it means observation actually refuted the laws — so there, it's the mathematical laws, for instance concerning falling bodies, as against Aristotle — or if we take, say, a lead ball and a feather — the lead ball — in principle you'd think one would go faster, and it's because afterward there was — science is counter-intuitive, right, but sometimes we're forced to come back to the sensible side too — but then we'd get back to Plato's analogy of the line, but coming back to it after having formulated hypotheses in order to verify, refute, validate, or confirm a hypothesis — that's not at all the same as walking one's dog thinking it's enough
1:05:01 to watch apples fall in order to find the law of universal gravitation — that's the difference. Since we don't have much time left before the questions, I'd like, using what you've just said, to make two points. First, about otherness — that's an important moment in what you said. I recall Rimbaud: there's no one here, there's someone. How can I know that this something in front of me, speaking to me, is not merely a something but a someone? I don't know how to be sure of it. I recall — I'm not deciding, but I recall — since we spoke of Descartes, that in the First Meditation, the one where Descartes goes to the paroxysm of doubt, he hasn't yet found the cogito, he hasn't yet found the bottom of his plunge into doubt — that's it, he's touched bottom, can he come back to the surface? No, he's in the process, to his astonishment, of seeing everything collapse that he thought, like everyman, was certain — for instance, that I am seated here by the fire, writing, and how can I deny
1:06:12 that these hands here — I quote him verbatim — and this body, are mine. You can see he's looking at what's closest to him, his body — there's nothing more sensibly close, for each of us, than our body. How can I deny that, unless I compare myself to those madmen — and there he speaks of the insane, whose brain is so clouded by black vapors of bile — I'll spare him that reduction of the explanations of madness — that they think they're kings when they're stark naked, or that they think they're stark naked when — well, whatever, that they have a body of glass. But these are madmen, and I'd be no less mad if I judged myself by their example. And there you see how Foucault could immediately pounce on him and say, you see, it took the age of reason for man to define himself as reason and exclude madmen from humanity — not at all, because right there he puts a full stop, Descartes, and takes a step in depth by humbling himself — that's where he's an immense philosopher. However, I must consider here that I am a man, and that consequently I am accustomed to sleep, and to represent to myself in my dreams things just as extravagant, and perhaps sometimes even more extravagant, than those madmen do when awake
1:07:22 so that, thinking about it carefully, I see there is no conclusive indication — sensible, that is — that would let me distinguish waking from sleep, so that I am quite astonished by this, as though by thunder, and this thought is almost capable of persuading me that I am dreaming. It's an extraordinary text — he's essentially saying: well, you see, your grandmother, there she is right in front of you — yes, but in a dream too, I sometimes see my grandmother, and yet, alas, she's been gone a long time. So the sensible criterion isn't enough to say there's someone there. So where am I heading with this? It's that, in relation to objects, they despair of seeing anything certain assert itself, but suddenly, well, they form the hypothesis that he's not alone, but that there is — and this is the famous hypothesis — an evil genius. There, we're no longer in the realm of somethings — as powerful as God, but unlike God, malicious, employing all his industry to deceive me and give me illusions, isn't he — and it's only once he's made this hypothesis, which brings him
1:08:32 to the paroxysm of doubt, that he reaches bottom and can finally have the first validated certainty: 'let him deceive me as much as he likes, he will never manage to make me be nothing, so long as I think I am something. So that, all things being maturely and carefully considered, one must conclude, and hold as constant, that this proposition, I am, I exist, is necessarily true every time I pronounce it, or that I conceive it in my mind' — we always forget to read what follows — maybe it doesn't hold when I don't say it in words, we'll come back to it. So I wanted to recall this, and maybe — well yes, it would be time to give Pierre-Alain — there's just a bit left — that's fine — I think we can give the floor to the audience if there are questions. Can you hear me?
Slip-ups of the piano-playing robot and the chess champion's defeat 1:09:29 Yes, how — can I hear myself? Good, can you hear me on the control room side? Yes, so I'll come back to you just to say you have a bit of time if you want to read other texts under the article Christophe had prepared, or otherwise you can see — I don't know if Christophe, do you see them, or Dominique — ah yes, I hadn't put up the comments — yes, yes, so you can look at the comments, there you go, I hadn't seen that — there we go, he's established — our old companion, well, our companion of long standing, sorry — who, with a lot of humor, and even a member of parliament, can utter words — well, and unfortunately for some groups of MPs, we don't see otherness — but that's good to add, it must go together, actually, and they're not the least talkative. There was a question also about the piano-playing robot — yes, no, it's more — I can read
1:10:40 one day the piano-playing robot will also have its sorrows, its flaws, its imperfections — technological perfection could also play imperfect passages, touching mistakes. Mistakes — yes, that's possible, that a robot could make a mistake at the piano, but that they'd be mistakes due to its sensibility, that would surprise me. I'll take the example, in jazz — probably not the only one to say this, but of a pianist I've thought about for 40 years, I'm thinking of Bud Powell, a great companion of Charlie Parker and Dizzy Gillespie — I don't know if it happened to him to fumble — that's how it is — but to fumble because he was doing controlled skids at 250 an hour with his right hand — so there, it's a fabulous risk-taking, such that he fumbles, and in a way he fumbles at the moment when there's an inner tension — I mean, it's always the shadow of a moral world, what we hear with our senses in man is the production of a moral world. The machine can very well make mistakes, but they're contingent mistakes, and not ones that have an intended, necessary aesthetic meaning that fits into the unity of the work — I'm cutting you off, sorry — no, me too
1:11:51 And then, one might also wonder, if we teach the machine to reproduce imperfections perfectly, are they still imperfections if it manages to reproduce them perfectly? Yes, absolutely, they're no longer imperfections — often, actually, mistakes are harder to execute than — what were you going to write? Yes, and then it's true there's also the bug aspect, because sometimes there are bugs — I don't know if I told you this too, the famous example, if we talked about it last time, of the chess game where Kasparov precisely had gotten lost in his own perceptions, he'd taken a machine bug for a fabulous move, or a move he wasn't expecting, because he had started learning how the algorithm of the machine worked, and that destabilized him and he lost. But there, that's — but you'd had a very nice exchange with Philippe Moreau from our association, who is an excellent chess player by the way, and who said, of course the machine beats the champion, but it's not because it has
1:13:02 more skill, it's rather that the champion is thinking about all sorts of things at that moment — emotional things, nerves, whatever you like, I mean — in short, it's the human surplus that can be defeated by the machine, that's it — to echo what you were saying earlier, I come back to the analog parameters — exactly — and we could also raise the question, would we be happy to see robots run the 100 meters, I don't know, there you go, that's it — no, no, but there you go, in other words the machine will beat the champion because that's all it does, play chess, it was designed to be a chess champion, agreed. Whereas there's something else in the head of the man playing chess besides chess, and fortunately, and in his gut too, not just in his head, and in his heart — there you go. Yes, so I see a lot of questions revolving around people who still have the illusion of believing, you see, that with progress there are machines that will eventually manage to do everything that we're currently denying to the animal, for instance
Logical, not ontological, break: from signal to sign 1:14:13 you see. So, there was also a question — you know, we went past it, but there was one — I think it was rather funny too — there are pianists who play like automatons, so why not the reverse. Yes, yes, yes, yes, wait, so I'd seen that somewhere, someone said, well, for instance, this, to say that one can absolutely accept that one day the machine might acquire properties that until now we thought exclusively human — I'm thinking of our very loyal friend Redrom — thought, consciousness, meaning, speech. So I understand very well that this objection would be justified if I were suggesting, and Christophe too, that there's suddenly some mysterious break between the animal kind and the human kind, between the world of nature and the world of the mind, isn't there, a discontinuity that would border on idealism, or rather
1:15:23 that would border on more than just border on it. But that's not at all what I'm saying — there is a material continuity, I'm the first to say so, I'm glad someone says somewhere further on that in order to have a soul, there has to be a body, isn't there, of course — but it's not an ontological break that I'm making, in the Cartesian style, precisely, that's not what I'm doing, it's a logical break, I mean, and properly speaking the gap is language. I mean, and I go back to our distinction between signal and sign — an animal, I keep using the same example because I've noticed it's clarified things for a lot of people, so I go back to it — when an animal in the savanna smells smoke, that's a sensation, it's a signal to it that there's a danger of bush fire, isn't there, that there's fire not far off, because in its here and its now it's not in the universal, it doesn't speak, there's the sensation, which is always here and now — this, here, now, that's what sensation is. And in the sensation it has, the smoke has always been consecutive to, or linked, somewhere in the environment
1:16:33 to fire — the animal doesn't need to think or to speak to know there's no smoke without fire, it feels it in its flesh, so to speak. So that's the signal, characterized by the natural consecution of signifier and signified, of fire and smoke. On the other hand, I was taking the example of the word 'lion,' because I know it in several languages, including even Bambara in Mali, isn't there — Löwe in German, lion in English, isn't there, and diara in Bambara — I mean, there's no longer any link at all between the sign and the thing, that's what's called, isn't it, the arbitrariness of the sign among linguists — arbitrary and no longer natural. It's not arbitrary that smoke signals fire, that's a law, it's a decree — it's arbitrary — the proof is, you change countries and, presto, it's no longer the same signifier at all. It's that leap, and that leap didn't fall from the sky at some point, isn't it, it was practical for hunter-gatherers, etc., to move to the verbal stage, and it's progressively that verbal stage that fashioned
1:17:43 that produced the anthropological instance, progressively, more and more precisely, because it's an access to the universal. It didn't fall from the sky, I'm not in some mysticism or vagueness, isn't it, irrational — it's exactly the opposite, on the contrary language is material, you need an eardrum, you need vibrations in the air to perceive it, or even when one is, isn't it, deaf, as Descartes said very well, one has the equivalent with a system of signs, provided that system of signs is common to a whole community where it's agreed upon. In other words, where the relation — and that's the difference with the signal — the decisive relation is not from me to the thing, from smoke to fire, but the relation of me to the other, and there we're in a triangle and no longer binary — it's no longer smoke and fire, that makes two, but there's me who signals the fire to the other, that makes three: me, the other, the fire — the other, let's be courteous, the fire, and me. There you go, and on that note I'd like to answer the question
Feeling, common sense, and the principle of identity 1:18:53 about — maybe before, on the body, I don't remember, it was whether one can distinguish — ah yes, ultimately, in order for there to be consciousness, doesn't one simply need to have a body — there you go — and even without going as far as consciousness, I answer yes — yes, but I mean, even without going as far as consciousness, it's also about decision-making, because there are neurologists — one who's called Antonio Damasio, if I remember — yes, Damasio, there you go — who had shown that anyway, one also has to see that our decision-making often — it doesn't follow reason alone, it's also sometimes the taking into account of feelings, of affect, etc., and that if we didn't have that we'd be incapable of deciding, and we see that we're a bit — sorry to interrupt you, but this one is burning my lips, word for word — Pascal, in one of his — Descartes, one generation later, says this extraordinary thing, you'd think he was paraphrasing Damasio, but with this terse style — all our reasoning
1:20:04 he's an immense mathematician speaking — 'all our reasoning comes down to yielding to feeling.' You can't be more precise. I give you back the floor, because it illustrates your point so well, doesn't it, that's exactly it — and that also allows precisely this taking into account of feelings, because there was also a question, we were talking about common sense — precisely it's that we're not — the human being is not just a rational being, there's also — all the feelings we learn as well, in relation also to experience, etc. We can keep talking, Christophe, because Adibou is asking us this question — good evening everyone, Christophe, can you explain to us what common sense is, thank you in advance — yes, that's a bit it — good evening rather — yes, so that's it, it means that, paradoxically — not paradoxically even, why do I say that — it's very difficult, for example
1:21:14 to do things that seem very simple to us, that seem very simple to a human being — it's very difficult to program them into a machine. Why? Because — well, there's everything, all the coding that we've also inherited from the past, what's sometimes called a form of retention — in fact, there are things that accumulate, from a certain point of view, and that get passed on, and that we don't necessarily manage to formalize — there are a lot of things that are informal, that we can't manage to formalize, and it's very hard to program into machines. Paradoxically, that's actually why sometimes there's a bad example, I find — some people, wanting to popularize machine learning a bit, say, well, machine learning nowadays is like a small child, there you go, it's a small child, and gradually, well, with this
1:22:24 this child-level, by dint of teaching it, well, it's going to become an adult with all the knowledge — but actually, well, that's of course completely false, since a child's education is when children — when there's a child, that's when it acquires, well, all these codes, all this common sense, etc. So this somewhat naive vision that AI now is like a child that's going to progress — well, it's because all the education, all the learning happens, after all, in the first years of a human being's life. So there you go, so common sense, yes, well, it's everything that's hard to — everything that's difficult to formalize, I'd say, in the form of an algorithm, and that's actually why it's difficult, it would require having knowledge that's also shared. So there, there's a lot of work on this, but we're still quite far off, and even, at the mathematical level too, it's going to be a bit, a bit difficult. I hope I've answered.
1:23:37 To the question — but we do have many interesting questions too, yes, but some of them repeat the same problem, and all the better, this problem is necessary. Yes, I think we've more or less covered common sense — the risk of military music versus classical music, yes. In any case I'm very glad about your point that the final decision that comes to conclude a piece of reasoning is not of a rational order, yes, that all our reasoning reduces — I like Pascal's vocabulary, which is very precise — reduces, in the end, at the very last second, to yielding to feeling. There I'd just like — this will perhaps be my own conclusion, I'll speak more afterwards — but it's that all of mathematics is based, at the level of rigor, on the principle of identity or non-contradiction, whichever you prefer — that is, A equals A. Before saying A equals B, A is different from B, A must first be — reflexivity, as they say.
1:24:47 — isn't that right — in mathematics, that A be identical to itself. But this A equals A in turn presupposes that the subject can say "I am I" — that's a formula of Fichte's, from the early nineteenth century — I am I. It's necessary that I am I. You'll tell me it's obvious that I am I. Yes, perhaps for us, but for a schizophrenic it may not be obvious — you see, one has to broaden the debate. A schizophrenic who says "I am Napoleon" — I'm taking a classic image — or Rimbaud, who has nothing of a schizophrenic about him but who has enough lucidity to say "I is an other" — please. And that's also the problem, moreover, of communication between AI and the human: it's that AI is not a subject of that thing either, so one can also get lost in this relationship and then fall into addiction, which is sometimes also what's sought through certain products. So perhaps also something about common sense — I think a good illustration is that there are whole fields on—
1:25:57 what are called human factors. And for example, let's take, say, plants — maybe factories or industrial production, that sort of thing, things like power generation plants — in terms of security, there was sûreté [nuclear/industrial safety] — there used to be a belief that we'd put everything in the form of an algorithm and the operator would just have to follow the path, the points laid out. Well, they realized that when everything goes wrong, what do you do? Well, you have to disconnect the operators from all these algorithms and tell them: here, you have 10 minutes, you have a quarter of an hour, we're giving you nothing more, and it's with this — I could say — common sense that they're going to manage to find a solution. So you see clearly this aspect — one might say, that's always the case, one might say, if we take the field of—
1:27:08 safety — why put humans in it? Because humans are fallible, because humans will make mistakes, etc., so we automate everything and that way we'll be at ease, we'll have optimal safety. But there's always also the need to adapt — the little pebble in the shoe, so to speak — which means that at a certain point, well, things will go outside what we've formalized, and there still needs to be a human brain that also has an irrational side, which will make a decision for reasons we don't quite know why. For example, you saw that pilot — I don't remember who it was — who hadn't landed but had ditched a plane in a river in the United States, etc. In all the piloting systems, perhaps it would have said "don't do that" — he did it, he saved the plane and the people too, which is the most important thing. What did I want to say on this as well — let me take advantage for a moment of your hesitation—
Can the machine give itself an end? 1:28:18 to take the floor again in relation to what I wanted to say, but I just saw a question come up — otherwise I would have remained rather silent given our fellow member Antoine Vatan, who is an economist, a young economist, one of the most brilliant — sorry to make him blush — that I've met in a long time, and with whom, I announce to you in advance, I hope soon to have — for everyone's benefit, if he's willing, though I believe it isn't — so I'm on his side — to have exchanges like the ones we're having now about computing. Now Antoine Vatan asks this very important question: is the machine capable of giving itself an end? In other words, it's another way of putting the question — Antoine will correct me if I'm forcing his thought — is human specificity conceivable if we do not attribute an end to it? It amounts to the same thing. So I'll recall here, to answer his question very very quickly, that while finalism is a cognitive catastrophe when it's imputed—
1:29:29 to nature — that is, when one thinks that in order to explain the movements of the planets, the stars, the galaxies, one must reason in terms of intention, of mysterious will, of divine ordinance, etc. — of course, that's the old system of ancient finalism, in particular Aristotle's — but especially Aristotle, I don't want to repeat the eternal example, isn't it, of the horses that speed up — that's the example Aristotle gives — when they smell the stable, they speed up, and that's why falling bodies also accelerate, because they're getting close to the bottom which is their stable, their home. All that is delirious from a strictly physical point of view — there is no bottom, there is no top, there is no home. So, on the other hand, this justified critique of ancient finalism rests on the fact that man alienates his own finalism, and here I answer Antoine's question — man, creator of thousands of objects, isn't he, since prehistory, to which he assigns an end: a flint to carve, a scraper to scrape, a bow to kill, I mean, a bowl to carry—
1:30:40 water — man has assigned an end to all these objects, and so he falls into the illusion of believing that everything around him, everything, is in his image. In other words, I mean, what is specific to man, namely precisely the ability to project himself into a future — which is what made Marx say, and there's nothing idealist about that, to say the least — what immediately distinguishes the most intelligent architect from the most skillful bee is that the architect has his house in his head before realizing it in the external world. That's what an end is: it's a house I cannot yet inhabit because it exists only in my head — it's a house for engineers, but the workers who would carry it over from essence to existence, if you prefer, aren't there yet. That's how I answer it. So, to carry through to the end the critique of ancient finalism, which was a dead end for science — well, which for a long time was what Bachelard would call an epistemological obstacle — the fact of constantly finding ourselves attributing intentions to things that are inert—
1:31:50 inclinations, impetuses as they used to say in old physics — the same logic obliges us to relocate finality where it actually is, and no longer in its alienated displacement, that is, in man. It's no accident that Kant speaks of the "kingdom of ends" for the intelligible community. There, that doesn't close the debate, but it's to congratulate myself on Antoine Vatan's question, which is a real philosopher's question, if I may say so. Yes, and I suppose too — perhaps another way of rereading the question is: even if the machine is capable of giving itself an end, well yes, but is it conscious of giving itself an end? That's the question. Since you bring up the word "consciousness," I'm a bit annoyed, because people constantly ask "does the machine have consciousness" — it's the mysterious word — there's nothing mysterious about consciousness. Every animal has consciousness. A dog is perfectly conscious that there's a danger in front of it. What is specifically human is not consciousness — thought is not consciousness, it's more than that. Well yes, even an earthworm—
Self-consciousness and the "minuses" that make man 1:33:00 has consciousness. A shark has electrical signals that tell it there's prey — it's conscious, it's not closed in on itself, it has a relation to the outside world, however rudimentary — an electrical signal, but the shark has it, and God knows the shark is one of the oldest animals in creation, among those still existing — well, not "creation," that's a metaphor. Whereas — here we're dealing with truly crucial matters — what's typical of man is not consciousness, it's self-consciousness. Exactly, that's where we see the importance of reflection, of the mirror, of distance — it's a minus that makes man. We must stop looking for it in a plus; it's not a matter of deifying him either — it's a minus that the myth of Prometheus expresses marvelously in Greek mythology, though there are others too. In any case, since we're in the field of Antoine's question, I always insist on these minuses—
1:34:10 that make pluses — whether it's the fact that man doesn't have the badger's claws, doesn't have the tortoise's shell to have a home built into his body, that he cannot fly like the bird or run like the cheetah, that he is inferior to every animal species in each of its specialties, that he has neither the cheetah's speed nor the elephant's strength, that he is entirely naked, has no fur, no shell, and must weave his clothes — it's this minus that forces him to pass beyond his limits through Prometheus's fire, a beautiful metaphor, which is the universal here, which allows him precisely to be transversal. And I was saying that I really like these minuses that make pluses — is it by chance, since I'm speaking to Antoine Vatan who knows Marx very well, much better than many of those who talk about Marx today, who are still trying to figure out how to interpret him — well, he applies him. I don't know, perhaps it would be time, as Marx said, we've spent enough time interpreting the world, perhaps it's time to transform it. Now Marx—
1:35:21 the key to his explanation of the world is precisely this minus that is surplus value — let's not fool ourselves, in "surplus value" [plus-value] there is "plus," yes, but surplus value consists in saying that there is a part of the wage-earner's labor that is not paid, that is withdrawn from him, subtracted from him — it's because he's paid less than what he should be given relative to the value he has created that there is surplus value. I recall in passing — I really like this — one could make a list of these minuses that make pluses — and I'll finish definitively on this with Soria Mehdi, so that's twice now I've answered him this evening because it took me a while to answer him, who didn't understand what a "negative subject" means — well, here we are, right in it: man's privilege is to be able to negate himself, to be able to say to his neighbor at the table "please, help yourself first," whereas everything natural pushes him rather to throw himself on the food, especially when the meal is a bit long. There, I was right in the middle of it — there was something about — I think it needs — about Claude Missy, who almost—
Docile men, arsonist firefighters, and conclusion 1:36:31 to wonder whether, behind certain of these projects around AI, there isn't an unacknowledged fantasy of reproducing artificial men with docile, non-disobedient traits — in fact, there — but I think we also shouldn't fall into the other fantasy. It depends — there, into the other fantasy, because there is a finality nonetheless — I think there's a whole option for AI that can help human well-being, etc., medical AI, etc., but it's like everything, it's a tool that can be used well or used badly, there. As for this story of reproducing artificial men with docile, non-disobedient traits — unfortunately one could also say we didn't need AI to have docile, non-disobedient men before either, so there — but there's a whole field—
1:37:41 as well of — well, sometimes one has to be a bit careful — there's a phrase that Jean-Gabriel Ganascia used, reflecting on the ethics of "pyromaniac firefighters" — in fact sometimes the GAFA companies scare us a bit with an AI that would be uncontrollable, etc., saying "look, us poor humans, we won't be able to fight it, so we need to dope ourselves, transhumanize ourselves, etc., in order to survive in this world" which perhaps won't even be there anymore — at least according to the promoters. Yes, there's that side to it, but of course — but thank you very much for having completed these answers to Antoine's question. I thank everyone who took part in this session, I thank Christophe whose skills we're making use of, I thank Pierre-Alain who manages a bit of a tour de force with all his references, and I bid you until next time, and I'd like to conclude—
1:38:52 just, still, on one small point regarding our audience — don't worry if it seems to go off in all directions, that's our regular Wednesday workshop with a program that continues on the relation to nature, the human sciences. On the other hand, I want to say, there are all these short formats that can be without a subject, without a prior program — the answers to your questions, that's already a short format, isn't it — and I'm very grateful to Pierre-Alain for having already put quite a few of them online. There are still some that have been recorded and are not yet online, so I'm announcing them to you, they'll be appearing shortly. And others — what I've liked to call "analytic reveries," a bit of "freestyles" if you prefer, of reflection — again on short formats that you'll soon see appearing on our YouTube channel. For the loyalty for which we thank you for being so attached — most of you, and increasingly numerous ones, if I'm to believe the figures Pierre-Alain gives me. There, thank you very much Christophe, big thanks, good evening, and goodbye everyone.
1:40:03 Thank you.
