by Paul Ducay in Philitt.fr (in French)
Bruno Bérard holds a doctorate from the École Pratique des Hautes Études (Religions and Systems of Thought). He is the author of essays on metaphysics and political philosophy, including Jean Borella: la révolution métaphysique, après Galilée, Kant, Marx, Freud, Derrida (2006), Métaphysique du paradoxe (2019) and La Démocratie du futur (2022), published by L’Harmattan. With theologian Johannes Hoff, whose research at Cambridge University’s Von Hügel Institute focuses on the anthropological challenges of digital transformation, he publishes Conversations avec ChatGPT sur l’Homme, le Monde, Dieu et l’Intelligence Artificielle : Intelligence ou raison artificielle ? (Conversations with ChatGPT on Man, the World, God and Artificial Intelligence: Artificial Intelligence or reason?), L’Harmattan. In it, he reveals that the conventional term AI is built on a profound misunderstanding: the modern confusion of comprehensive intelligence with calculative reason.
Raphaël Enthoven’s recent best-selling essay “L’Esprit artificiel” (The Artificial Mind) represented the first, long-awaited reaction of contemporary man to the irruption of artificial intelligence into everyday life by the ChatGPT conversational agent developed by the American firm OpenAI. This spontaneous reaction, however, does not go beyond the thinking of the romantic man, who believes it is enough to oppose the outrageous rationalization of existence with the rights of sentiment. He rightly opposes machines to the fact that, as Johannes Hoff explains, “computers don’t scratch their heads because the enigmas of the world take them to the limits of the conceivable”, that “computers don‘t dream of a future where no one has ever been”. Indeed, if we confine ourselves to the “modern definition of intelligence”, the critical witness of this technological progress notes how this human notion of intelligence still includes aspects too extensive to be adequately applied to AI: in particular, Bruno Bérard lists “the generation of consciousness, volitional autonomy and affective behavior”. It would never seem reasonable to the lucid romantic to attribute these lines by Musset to any machine:
If only daydreaming were still possible!
And if the sleepwalker, extending his hand,
Didn’t always find nature inflexible
Who strikes his forehead against a pillar of bronze.
In the absence of incarnation, the cognition of the computer system will never be subject to the tragic alternation of the human being whose life oscillates between the dream and the test of reality, which pulls him out of his dream to confront him with new truths: for not only does AI not seek truth, but probability, but truth can never be for it the gracious fruit of a test. So, before being intelligent, we must first object that AI is artificial.
The modern reversal of intelligence
Yet doubt remains. These phenomenological precautions, however necessary and justified they may be, do not provide a lasting solution to the apparently highly paradoxical hypothesis of an embodiment of the machine. In the book’s exergue, researcher Sarah Spierkermann, who since 2009 has chaired the Institute for Information Systems and Society at Vienna’s University of Economics and Business, in fact defines “AI system” as “a virtual and/or physical integrated computer system, capable of independently executing a wide range of cognitive functions […] based (…) on unstructured, content-rich data sets”[1]. Now, given the progressive sophistication of myoelectric prostheses that reconstitute the nervous sensations of amputated organs, it is by no means forbidden, under these conditions, to envisage the insertion of such a computer system within a physical system that would enable the AI to access the states of consciousness of an embodied living being. Faced with this eventuality, however hypothetical, reference to the phenomenon of the flesh is no longer sufficient to establish a clear distinction between human and robot. In fact, Bruno Bérard’s compelling reflections have the merit of deploying a critique of AI that no longer confines itself to the conditions of manifestation of intelligence, to which the phenomenology of the flesh is limited, but extends to the very essence of intelligence considered in itself and distinct from other essences, for if it is true that “nothing is in intelligence that was not first in the senses”, as Aristotle taught, we must also note with Leibniz: “if not intelligence itself”.
Indeed, it is the very nature of intelligence that has been lost in modernity: “the 1956 appellation (of mathematician and computer scientist John McCarthy), ‘Artificial Intelligence’, is very much in the air of modern times”. This appellation is the result of a confusion between reason and intelligence that characterizes the entire anthropology established by modern philosophers. In this regard, Bruno Bérard takes up the decisive analyses of metaphysical anthropology by the last great Neoplatonic philosopher of the contemporary era, Jean Borella, who, in La Charité profanée (reissued under the title Amour et vérité in 2011), traces the genesis of this “rationalist reduction” in two stages. The first moment was the confusion of intellectuality and rationality by René Descartes who, in the Latin text of his second Meditation, established a pure equivalence between “intellect” (intellectus) and “reason” (ratio), in contrast to the earlier philosophical tradition which, from St. Augustine to St. Thomas Aquinas, had almost always distinguished them. The confusion of the two faculties was followed by their inversion. As Bruno Bérard sums up, “inverting these two faculties, reason and intelligence, is the work of Kant, who placed reason at the summit of the cognitive faculties, denying the possibility of intellective intuition”. In Kant’s view, the superior faculty of the intellect becomes “understanding” (Verstand, intellectus), inferior and subordinate to reason (Vernunft), which is placed at the top of the hierarchy, without obviously attributing to it the intellect’s ancient capacity to intuitively contemplate the essences and first principles of being and knowing. “From confusion to negative inversion, this is the path taken by Western thought”[2]: the deconstruction of metaphysics by modern humanism is paradoxically what has made possible this “defamation of man” (S. Spierkermann) by technological civilization.
Computers and Plato’s oblivion
So it’s hardly surprising that the technicians and consumers of modern technology should attribute to this form of computer system, endowed with organizing independence, the misleading epithet of “intelligence”. But however convenient, this epithet is no less “misleading”, since it confuses two rigorously distinct faculties: on the one hand, “reason is a power of calculation [and] reasoning under the aegis of logic”, while, on the other, “intelligence is the faculty of understanding these calculations and reasonings”. The difference between reasoning and understanding (or intelliger) is given to us by St. Thomas Aquinas, who, in the Summa contre les Gentils (I, 57, §4), “subtly distinguishes between the very act of reasoning, which consists in ‘passing from principles to conclusions’, and the ‘judgment on an argument’, which consists in ‘looking (inspicere) at how the conclusion follows the premises, considering them both together'”[3]. But the computational model of the computer system in no way incorporates such a “gaze”, such a contemplation of the necessity that links the premises of true and not merely probable reasoning.
The confusion of intellectuality and rationality, so characteristic of the dominant interpretation of the nature and powers of AI, is thus the result of an oversight of Plato[4]. Just as we have mistakenly taken the democratization of information for a democratization of knowledge, forgetting the great lesson of Theaetetus, where Plato explained how it is not enough to possess information, even accurate information, to possess knowledge about it – for it still remains to know how to justify and demonstrate it rigorously – so we have forgotten this “immemorial distinction, formulated by Plato (Republic, VI, 511d-e): on the one hand, hypothetico-deductive knowledge, the discursive reasoning (dianoia) of reason (ratio), and, on the other, knowledge by intellectual intuition (noèsis) operated by intelligence (noûs, intellectus)”.
It’s clear, then, that the problems posed by the qualification of AI require us to return to the forgotten principles of traditional metaphysics, and in particular to rediscover the nature and implications of intellectual intuition. Certainly, we must be wary of reducing “contemporary ‘transformative’ technologies” to “statistical parrots”, as Johannes Hoff warns, because “they don’t just reproduce pre-established data, but incorporate a certain level of randomness that may surprise us”. In particular, notes Bruno Bérard, these technologies have “sufficient sophistication to enable recursive improvement, at least in the form of a self-learning function”. However, “recognizing faces or words, winning strategic games, automating cars, simulating military operations, organizing complex data and so on. But “recognizing human speech or organizing complex data are all matters of programming, calculation and automated reasoning“. However, “recognizing human speech or organizing complex data” is strictly not the same as “understanding human speech or interpreting complex data”. The rediscovery of theories of knowledge that are far more complete and unambiguous than those established since the Cartesian confusion therefore forces clarity: the letter “I” in “AI”, which stands for “intelligence”, “should de jure be replaced by an ‘R’ for ‘reason'”.
The inevitable loss of progress
An understanding of the unintelligent yet ratiocinating nature of Artificial Reason (AR) should in turn enable a rational, rather than fascinated, use of this set of tools, which cannot precisely overstep its function as a tool. According to Bruno Bérard, their “misuse”, either by “the user”, or by its character as an “imperfectly mastered technology”, or by the “combination of both”, could induce potentially very serious risks, at a time when the development of “mental energy [accumulated] by mankind” since the “entry into service of the IBM automatic sequence calculator or Mark I” in 1944 is now tending to reach “the level of the most destructive mechanical energy (atomic bomb)”. That’s why the only kind of use for Artificial Reason, notes Johannes Hoff, is “more exploratory and dialogical”, using its systems as “collaborative tools that enhance our own intelligence rather than simply answering pre-formulated questions”. Otherwise, the development of information will be inversely proportional to the knowledge and prudence of a humanity increasingly enslaved to its own tools, in which it continues to externalize its specific capacities to act and reason.
An ancient Chinese meditation shows, however, that the form of use is not enough to avoid the problem. While the former cannot exceed a certain limit (“for example, no one will want to or be able to eat five meals a day”), “non-vital needs, on the other hand, seem indefinite, exceeding what the earth can provide”. Now, in the 4th century BC, the sage Zhuang Zhou (chap. XII, 11) uses the parable of a Gardener exhausted by constantly going down to the bottom of a well to fill his pitcher with water, to whom Zigong offers to give him a “chadouf”: “a machine carved from wood with a heavy back and a light front”, “drawing water as you lift your arm”. To this proposal, the wise Gardener counters with an answer that has lost none of its topicality (translated by J.-L. Lafitte):
I have heard my master say that he who uses artifice works with artifice, that he who thinks with artifice loses his purity, that he who loses his purity loses his peace of mind, and that the Way does not support him who has lost his peace of mind. It’s not that I don’t know the advantages of this machine, but I’d be ashamed to use it (Zhuangzi, XII, 11).
Zhuang Zhou explains, in the words of Bruno Bérard, “that the quantitative benefit is not everything – essentially proposing to create new avidities – and that we must also consider that the tool transforms the man who appropriates it, as well as society as a whole”. Quantitative progress can be accompanied by qualitative regression, just as a qualitative benefit in one respect can be accompanied by a qualitative loss in another: here again, we need only think of Plato, for whom writing was indeed a collective remedy against forgetting, but a personal poison for memory. And yet, what do we lose by delving deeper into the gulf between the unbridled complexity of technology and the public’s increasingly tenuous knowledge of its machines? Doesn’t technical development in the modern world increasingly consolidate the mechanism of oppression that Simone Weil identified in the divorce between “those who think” and “those who execute”?
[1] See her paper in Metafysikos at https://metafysikos.com/en/on-the-difference-between-artificial-intelligence-and-human-intelligence-and-the-ethical-implications-of-confusing-artificial-intelligence-and-human-intelligence/.
[2] Jean Borella, Amour et vérité. La voie chrétienne de la charité, chap. VII, 3, 1: “Intellect et raison”, L’Harmattan, coll. Théôria, Paris, 2011, p. 112.
[3] Michel Nodé-Langlois, “L’intuitivité de l’intellect selon Thomas d’Aquin”, Revue Thomiste, No. 100, 2000, p. 199, n. 118.
[4] Cf. https://philitt.fr/2023/12/07/lintelligence-artificielle-loubli-de-lart-de-la-memoire/.