Ever since the foundations of this scientific field, so full of intriguing prospects, were laid in 1956, we have witnessed periodic flare-ups of enthusiasm, which regularly deferred to a more or less near future the promise that “machines will one day behave like human beings.” Given that the word “intelligent” is now applied to a great many technological applications, it may be worthwhile to return to these questions and reflect on them, today more than ever.

To restore this debate to its proper horizon, it is worth asking: what is thought? What is the relationship between thought and language? Where does consciousness reside and, above all, what is the mind? These questions, which blur into and intertwine with the whole history of philosophy, have long engaged and tormented scholars across many disciplines without ever yielding a definitive answer, and they show just how inadequate our understanding of human intelligence still is. After centuries of reflection and impassioned but inconclusive discussion, the question remains the same: what is it that makes the human being what it is? We have repeatedly seen, on the one hand, the assertion of a rigorously materialistic view of the human mind, underpinned by a mechanistic conception of mental events, and, on the other, spiritualistic interpretations that have sought to bring out the specific and ineradicable traits of the human intellect, while still today the attempt continues to define, or even to reproduce, the brain’s infinite capacities. The contemporary debate has restated the questions set out above in the light of the advances made in the scientific field, renewing the challenge thrown down to humankind and to its identity. A challenge that, as has been observed, has its roots in the modern age, an epoch of humiliations for the human being: from the cosmological humiliation brought about by the Copernican system, to the biological one handed down to us by Darwinism, to the psychological one produced by Freud’s discovery of the unconscious (Alici, 1990, p. 39).

Artificial intelligence, a subject invoked for many years now in support of a materialist philosophical anthropology, is the latest attempt, in chronological order, to call into question the primacy of the human being as person; yet it has also long been one of the most evocative and significant horizons of the science and technology of our time. In this technological challenge we may therefore glimpse an opportunity for the human being to spell out what is specific to his own intelligence, and thus to arrive at a deeper awareness of human nature and of knowledge. A further question has been raised: why does artificial intelligence have philosophical relevance? (Agazzi, 1991, p. 1) At bottom, it is a scientific discipline oriented towards technological applications, with no particular philosophical interest of its own. And yet that is precisely the point: those who theorise the realisation of an artificial intelligence are convinced that the human capacity to reason can be reproduced, and confidently claim that in the very near future they will be able to build intelligent machines. It is precisely this use of the adjective intelligent, normally reserved for human beings, that prompts an inquiry, a closer examination, a philosophical question, as well as a clarification of terms. There is, moreover, good reason to reflect on these matters, since it is the very function of philosophy that is being called into question; for the world of AI, indeed, the knowledge that counts is that which comes from positive empirical research, concerned exclusively with describing and explaining the world and its processes in an objective manner, resting solely, or chiefly, on a behaviourist psychology and stripping philosophy of any foundation of a cognitive nature. The issue thus takes on decisive importance, since what is at stake is not machines but the human being and its essence, and, ultimately, the very value of philosophy.

The project of thinking machines

History is full of attempts, more or less strange and mysterious, to reproduce human intelligence. To avoid getting lost along a winding path that, should we wish to trace our steps backwards, would take us from Paracelsus’s homunculus all the way to Egypt in 200 BC, we shall take the modern age as our point of departure, not least because it is precisely to this epoch that the first idea of a thought capable of unfolding within a machine belongs.

“We can define, that is, determine what we mean by this word reason, when we count it among the faculties of the mind. For reason, in this sense, is nothing but a reckoning, that is, an adding and subtracting of the consequences, of general names agreed upon for marking and expressing our thoughts: marking them, when we reckon them for ourselves; expressing them, when we demonstrate and explain our reckonings to other men.” Perhaps, in the modern era, the beginning of the project may be traced precisely to the moment when Hobbes, setting his own materialist ideas against the Cartesian res cogitans, formulated in the mid-seventeenth century the first expression of the computational view of thought. Convinced, moreover, as is well known, of the need to constitute an artificial body, the State, the English philosopher was firmly persuaded that to reason was nothing other than to calculate; and, embracing the idea that reality was essentially mathematical, he held that thought rested solely on symbolic operations. The following three centuries saw Hobbes’s original philosophical intuition develop until it became one of the most debated and fascinating philosophical knots in contemporary scientific and technological knowledge. Beyond this, three episodes are regarded as pivotal to the progress of the so-called thinking machines: Leibniz’s machine for arithmetical calculation, the calculating engine conceived in the nineteenth century by the English mathematician Babbage, and the definition, worked out by Alan Turing in 1936, of an ideal reasoning machine characterised by logical omnipotence (Pratt, 1990, pp. 9-16).

A mathematician and philosopher on the one hand, a diplomat by profession on the other, Leibniz, a precursor and, in a certain sense, a founder of formal logic, was among the first to entertain the idea that a machine might perform intellectual tasks. Starting from a reflection on the nature of logic, convinced of language’s inadequacy in mirroring the structure of the world and aware of the possibility of mechanising reasoning, he encouraged the attempt to construct a kind of symbolic language that would use combinations of elementary propositions (with antecedents in the combinatorial logic of Petrus Ramus and in the Ars Magna of Ramon Llull, and with a continuation, at least in part, in the Principia Mathematica of Russell and Whitehead), a language that could thereby represent every concept and so win the palm of “universal writing” (characteristica universalis) and hence of universal knowledge.

Later, more than a hundred and fifty years ago, the brilliant English mathematician Charles Babbage, after an attempt he subsequently abandoned with the so-called difference engine, designed a calculating machine of exceptional size, whose task was to have been to free man from the most laborious calculations. This machine, known as the analytical engine, was to have the power to combine generic symbols together and to embody a new algebra, conceived as the science of general reasoning. Alongside him, Ada Augusta Byron, Countess of Lovelace and daughter of George Byron, intuited that the potential of the analytical engine might exploit the formalisation of logic to build a symbolic system, one that would create a language capable of expressing all the laws governing the relation between any two things.

“Can machines think?” Alan Turing was already asking in a celebrated article (Turing, 1950) that appeared in the journal Mind. In that contribution he argued for the operational dimension of thought and encouraged the attempt to build a machine endowed with powers of scope equal to those of a human brain. His project, first proposed in 1936, was thus the first genuinely concrete idea of building a thinking machine. The English mathematician, besides making fundamental contributions to the development of the first electronic computers, set out the first mathematically precise theory of computation, one that in fact encompasses some surprising discoveries about the capabilities of computers (Turing machines). The so-called Turing test, then, went on to be used as the basic paradigm for the design of artificial intelligence.

A seminar held at Dartmouth College during the summer of 1956 finally marked the official beginning of studies in artificial intelligence, launching a sizeable group of scholars from various disciplines towards the idea that there existed an objective and rigorous way of explaining the human mind. This came about partly through the creation of Logic Theorist (1956), a program capable of automatically proving certain mathematical theorems; of General Problem Solver (1958), a now celebrated problem-solver; and of Perceptron (1960), the first device built to simulate the workings, no longer merely logical but also physical, of a neuron. The goal accompanying the project was, in essence, openly declared: to bring machines to learn. During the 1960s Hilary Putnam, introducing the concept of functional isomorphism, laid the foundations for a logical and psychological justification of the mind-machine analogy. A few years later the neo-functionalists Dennett and Hofstadter, the latter the author of a hugely successful manifesto of AI (Hofstadter, 1979), would definitively enunciate the principle that equates the mind-body relationship in the human being with the software-hardware relationship in the computer, together with the ensuing analogy, recalled above, coined by Putnam.

In the 1980s the advent of computational linguistics, that is, the study of computers’ capacity to understand natural language, and the mechanisation of predicate logic opened the way to logic programming and to the rise of the so-called expert systems: devices created to solve problems within a specific domain of knowledge, initially used in the diagnosis of illnesses and designed to imitate typically human reasoning processes. They represent the first real attempt to create intelligent machines. Let us look briefly at the idea underlying them. An expert system is a program that seeks to solve a problem using two kinds of knowledge: the conceptual and theoretical knowledge that comes from books, and the knowledge that comes from experience in applying reasoning methods to solve problems. On this basis, an expert system is made up of two distinct parts, a knowledge base and an inference engine. The former consists of a representation, within the computer, of the available knowledge; the latter is an algorithm grounded in the reasoning method, capable of identifying solutions to given problems. A trivial example makes the point clear: if I know the connection between illness and fever (knowledge base) and I know that Marco has a fever, I can deduce that Marco is ill, according to the most classic of syllogisms.

The philosophical foundation of AI: the functionalism of Hilary Putnam

Speaking of intelligent machines, one certainly cannot overlook the thought of Hilary Putnam, the so-called advocatus diaboli of the mind-body problem, the man who, at least initially, maintained that it was possible, from a logical standpoint, to grant civil rights to a Turing machine. The American philosopher’s early convictions about the mind and the nature of psychology, which led him to sketch out an original theory of his own, functionalism, may by now be regarded as the philosophical foundation of artificial intelligence.

Putnam’s functionalism led a tortuous and difficult existence, to the point of being almost entirely disavowed by its own author. Having written, together with Paul Oppenheim, one of the most decidedly hyper-reductionist manifestos of the contemporary period, Putnam advanced, at the beginning of the 1960s, the hypothesis that a computer, or rather, a Turing machine, is an appropriate model of the human mind; for the American philosopher, a Turing machine is an abstract machine, physically realisable in a virtually infinite number of different ways (Putnam, 1987 (1), p. 402). In this period Putnam shows himself decidedly critical of those who do not believe in the operational dimension of thought, arguing that whether robots are conscious can be neither affirmed nor denied: to assert that a robot is alive and conscious may be false, but it is not a contradiction, and leaning towards one hypothesis or the other is the fruit not of a discovery but of a decision. The early Putnam thus asserts the existence of an evident analogy between human beings and that species of finite-state automata that are Turing machines, between the physical mental states of the human being and the structural logical states of the machine. On the strength of what Eccles and Popper sarcastically labelled a promissory materialism, he holds that, given the growing speed of social and technological transformations, it is entirely possible that one day robots will exist and declare: we are alive, we are conscious (Putnam, 1987 (1), p. 425). Some functionalist element is already present at this stage, since Putnam, against the identity theory, advances the hypothesis that a robot and a human being have the same psychology, in the sense that both can obey the same psychological laws, meaning by this that physically different structures can obey the same psychological theory. Here a distinction must be drawn between type-type identity and token-token identity, since functionalism criticises the former in order to affirm the latter: by identity of types is meant a kinship that refers to the essence of the terms in question, whereas identity of tokens refers to function, a distinction, therefore, between being and acting, a distinction decisive for functionalism. All this led Putnam to make claims that have disoriented and disconcerted more than one scholar, as when he asserted that today no knowledge in our possession is, strictly speaking, incompatible with the hypothesis that we are all Turing machines, even though certain things known to us make the hypothesis rather implausible; or when he maintained that “it seems preferable to extend the concept of consciousness in such a way that robots too are conscious, since a discrimination based on the softness or hardness of the parts making up the body of a synthetic organism seems to me every bit as foolish as discriminatory treatment of human beings based on the colour of their skin” (Putnam, 1987 (1), p. 443).

Towards the end of the 1960s, however, he moves towards recognising an autonomy of the mental and towards a rehabilitation of psychology, both by criticising traditional materialism and physicalism and by taking a negative view of his own earlier assertions. On the question of identifying the psychological states of a human being with the logical states of a Turing machine: for the mental states of a human being to be similar to those of a Turing machine, they would necessarily have to be instantaneous, which is obviously impossible for states such as jealousy, love or competitiveness, since these depend on a great deal of information, on learned habits and facts, and on social relationships.

Putnam had in fact already introduced into the debate of the time on the mind-body problem the first considerations foreshadowing that working hypothesis and that epistemological-psychological framework which functionalism was attempting to spell out within contemporary philosophy of mind. The traditional philosophies of mind, as is well known, could be divided into two areas: dualism and monism, where by dualism is meant that current of thought for which human reality is of two orders, mental and physical, whereas for monism that reality is essentially made up of a single substance. Until the period preceding the birth of functionalism at Putnam’s hands, the doctrines that seemed most widely credited were dualism of Cartesian origin on the one side and materialism on the other. For the dualist of Cartesian stamp, then, the mind is a non-physical substance; for the materialist, the mental is not distinct from the physical, and all mental states, processes, properties and operations are identical to physical states, processes, properties and operations. Among the materialists, the behaviourists hold that the mental can be eliminated by reference to environmental stimuli and behavioural responses, while the proponents of the identity theory affirm that mental causes exist and are to be identified with neurophysiological events in the brain (Fodor, 1981, p. 100). Gradually, the inadequacy of both positions (the dualist and the materialist) came to be recognised, given the difficulties concealed in their theoretical formulation: dualism is incapable of accounting for mental causation (Descartes’s pineal gland), since it is by no means clear how a non-physical mind could give rise to physical effects; materialism, especially in its behaviourist form, encounters difficulty in defining both the mind-body interaction and certain rather important concepts such as those of consciousness and intentionality, since beliefs, expectations and items of knowledge have something more than a merely behavioural qualification. As has been observed, indeed, while every behaviour is also an activity, there may be activities that do not translate into any specific, physically observable behaviour; consider, for example, “a man lying on a bed, immersed in a perfectly tranquil sleep, and the same person who, in the identical position, instead of sleeping is thinking, eyes closed, about something” (Agazzi, 1967, p. 7). The identity view too, offered as an alternative to logical behaviourism, runs into complications, with an overly deterministic conception of the human mind. It was recalled above that for every identity theorist every mental phenomenon is nothing other than a specific neurophysiological event. But while sensations seem fairly easily traceable to physical events/processes, how can I compare to physical events such mental phenomena as beliefs, intentions, hopes? Moreover, if a surgeon were to examine, with instruments however sophisticated, the brain (or the CNS) of a man who is choosing or a man who is believing, he would never find the choice or the belief there as such. Which by no means implies that psychic phenomena bear no relation to specific neurophysiological processes.

On the ruins of these complications, Putnam gives rise to a philosophy of mind that is neither dualist nor materialist, one that goes by the name of functionalism and that holds that the psychology of a system does not depend on what it is made of, but on the way in which its constituent parts are assembled; functionalists, in short, maintain that artificial intelligence functions in exactly the same way as authentic intelligence. Putnam’s reductionist stance is thus abandoned to make room for an affirmation of the functional organisation of mental states; this original formulation by the American philosopher rests on the concept of functional isomorphism: two systems are functionally isomorphic if there exists a correspondence between the states of the one and the states of the other that preserves their functional relations, and this holds even for systems physically different from one another. Indeed, strictly speaking, a Turing machine need not even be a physical system: anything capable of passing through the succession of states over time can be a Turing machine. In essence, then, it is of no importance that the physical realisations differ, since a computer built with one type of valves, wires and relays finds itself, during its computing operations, in a chemical and physical state different from that of another computer made with a different type of valves, wires and relays, yet the functional description can be the same. Putnam now distinguishes between an entity and its behaviour.

More recently, Putnam himself has admitted that the computational analogy between man and machine fails to provide adequate answers to important questions concerning the nature of mental states such as belief, preference, reasoning, rationality and knowledge, undertaking a careful self-criticism of his own original convictions; Putnam’s change of heart has not, however, stopped the advocates of AI, who draw precisely on his concept of functionalism to build their own theories of artificial thought.

The perennial adversary: John Searle

It would be impossible to summarise here all the criticisms levelled at the artificial intelligence project, given that the great debate has produced a genuine rift within contemporary philosophy. As early as 1980, John Searle set in motion the great debate around the mind-body problem, which from that date on has pitted against one another mathematicians, biologists, philosophers, psychologists, epistemologists and AI advocates, all attempting to get to the bottom of the question and so to shed light on the obscure points of the philosophy of mind. Searle himself proposes a distinction between a weak (or scientific) version and a strong (or ideological) version of AI: according to the champions of the first, the computer’s principal value in the study of the mind is that it furnishes us with a very powerful instrument for the formulation of hypotheses; it, indeed, “does not claim to emulate human intelligence, but simply to simulate certain intelligent human behaviours that involve deductive and repetitive tasks” (Basti, 1991, p. 109), thus proving quite rigorous in testing tools and claims concerning the nature of the mind, of intelligence and of psychology; for the supporters of the second, by contrast, the computer is not merely an instrument, rather, suitably programmed, it really is a mind, in the sense that it understands and possesses conscious, aware states: it knows, that is, for example, that it is alive (!).

The principal argument advanced against the strong version of AI arises with Searle precisely from a linguistic standpoint. It is claimed that, while one can conceive of building into a computer program a check on the syntactic correctness of deductions, it does not, by contrast, seem possible to insert into the machine a genuine check of a semantic kind. In essence, the recurring objection to intelligent machines, one that can be turned against ChatGPT as well, is that our thoughts possess a meaning, because a thought refers to something beyond itself, is about something, and reflects that distinctive and exclusive capacity of our brain to relate the human organism to the world: intentionality. It is precisely in virtue of the latter that it seems possible to attribute a meaning, a content, to statements; whereas in machines formulae simply refer to themselves, we, by contrast, do have the faculty of conferring on them a designating power. As recalled above, Putnam himself revised his ideas about the mind-machine analogy, and it is he who argues that the Turing test for reference is no definitive proof; reference is in fact merely an illusion, and everything we can say about things is intimately bound up with our non-verbal transactions with them (Putnam, 1989, p. 17). If an ant were to crawl across a beach and trace in the sand a caricature of Winston Churchill, would the image it had drawn represent Winston Churchill? Obviously not, since the ant has merely, per accidens, traced a line (…) that we can see as an image of Winston Churchill (Putnam, 1989, p. 7). The fact is that there exist rules for entering language and rules for exiting it, whereby we perform actions that are not merely verbal in kind; this does not happen in the machine, and there is no reason to regard a machine’s conversation as anything more than a mere syntactic game that in truth closely resembles intelligent discourse, but no more than the curve traced by the ant resembles a caricature. In the case of the ant, one might say that it would have traced the very same curve even if Churchill had never existed. The machine’s reference to external things is, in fact, shaped by whoever writes the program; a certain causal link between the machine and objects of the real world therefore exists, but it is so weak as to prove wholly insufficient for us to speak of reference. Searle maintains that the computer is a syntax, whereas the mind has something more than a syntax: it has a semantics (…) Minds are semantic, in the sense that they possess something more than a formal structure: they have a content (Searle, 1988, p. 24). He aligns himself, moreover, with those who see in intentionality that feature, bound up with that special kind of observer which is consciousness, that cannot be reproduced by the machine. The thesis that intentionality is the distinctive trait of the mental is the so-called Brentano thesis: all mental phenomena exhibit intentionality, whereas no physical phenomenon is capable of it; hence the mental cannot be reduced to the physical. Along this same line of thought, Agazzi too maintains that the something more which distinguishes the human organism from the machine is denoted by the term “intentionality, by which is meant precisely the fact that in the cognitive activity of the living being there occurs a kind of participation or identification of the subject with respect to the objects which, while remaining themselves, in some way become part of the subject (…) A surrogate for it (one that, however, expresses that meaning only approximately, and points with precision to a level somewhat richer than mere intentionality, since the latter does not of itself entail the awareness of knowing) may be, in current usage, the term consciousness” (Agazzi, 1967, p. 16).

Returning to Searle, his most famous argument against AI in its strong version is undoubtedly that of the Chinese room: imagine an individual, shut in a room, who knows no Chinese whatsoever but who is supplied with a set of rules, written in his own language, for manipulating Chinese ideograms. This person would be able to produce Chinese answers to questions posed in Chinese by other people, standing outside the room, who know Chinese perfectly, all the while still understanding absolutely nothing of the meaning of a single ideogram. Now, a computer manipulates linguistic symbols at input and output according to the very same principles as the Chinese room; to attribute to it intelligence, intentionality and semantic powers is to go against the Turing test. Davidson, moreover, maintains that “before a computer can have beliefs, desires, thoughts of any kind that can be recognised as such, it must have at its disposal a great part of the information that forms the natural equipment of each one of us. Until then we can only say that the system represents certain information, or even ends and strategies, but the system certainly cannot be interpreted as having that information, those ends, those strategies” (Davidson, 1989, p. 43). As regards reference, and in particular the machines’ capacity to refer to something external to them, Putnam too, initially fixed in rather reductionist positions in the philosophy of mind, has changed his mind; in his most recent revision of his functionalist hypothesis he affirms that “it seems evident that we cannot attribute to such a device the possibility of reference. It is true, for example, that such a machine can say wonderful things about a landscape; it would not, however, be able to recognise (…) It is a device for producing statements in response to statements: yet none of these statements is linked in any way to the real world; indeed, two machines left to play the imitation game could go on fooling each other forever, even if all the rest of the world were to disappear!” (Putnam, 1989, p. 17).

The supporters of the strong version of AI, appealing to a formalist theory of the mind, maintain instead that intelligence is simply a matter of rule-governed symbol manipulation, thereby dispensing with such traditional concepts as intentionality, consciousness, interiority, and so on. Ultimately, the strong version of AI denies the individual subject the faculty of thought as thinking thought, and thus strips it of any possibility of free self-determination, even though, with its weighty questions about the human being and about the nature of knowledge and of intelligence, the world of AI does offer stimulating avenues of study and inquiry different from the traditional ones, reaffirming the importance of investigating the developments of the cognitive sciences. In spite of this, the computational analogy between man and machine fails to provide adequate answers to important questions concerning the nature of mental states and of concepts such as feeling, thought, meaning, subjectivity, personality, creativity, consciousness, concepts that strike straight at the heart of the philosophy of mind. We have no proof that a machine feels, for example, moved, sad, confident, guilty, altruistic, worried, angry or hopeful, or that it possesses values. And to think that John McCarthy, who, incidentally, was the first to use the term artificial intelligence, used to say that even machines as simple as thermostats can be said to have thoughts and beliefs (it’s hot here, it’s cold here, it’s comfortable here). Not to mention Marvin Minsky, one of the fathers of the so-called intelligent machines, who held that the computers of the next generation, still promissory materialism, Popper would say, will be so intelligent that we shall be lucky if they feel like keeping us around the house with them. All this to make clear just how far apart the convictions of AI’s advocates are from those of its critics.

Further observations and a few reflections on the AI project

The debate over the machine’s capacity to simulate human thought seems by now to be never-ending. Can ChatGPT be the turning point? Are we facing a genuine change of pace where AI is concerned? Can human intelligence actually be reproduced? Or is the question we ought to be asking rather: which parts of our mental work can we entrust to a computer? There is no doubt that the development of AI in its strong version, that of the so-called thinking machines, has enjoyed and still continues to enjoy a moment of great renown, for all its dreadful premise of emulating human thought. The scholars working on neural-network projects, together with a certain amount of further research in support of a materialist view of the human mind, still bear witness to its current lines of inquiry. Beyond all this, it seems fairly evident that the technological challenge posed by AI has radically undermined even the foundations of traditional philosophy; yet the development of the computer passes by way of the human being: living beings act to satisfy intrinsic needs such as survival, whereas the processor is always subordinate to external tasks. Lambertino has observed that “the computer seems to pursue a purpose, an internal logic. In reality, its underlying idea is an objective and not a subjective teleology, since it is a purpose that is necessarily attained in an unaware manner; indeed, its interpretations are anything but subjective, which is to say that its behaviour reveals an objective disposition to process information and knowledge.” Lambertino goes on to stress that “man, unlike the machine, can err, and this not only in knowledge of a moral character (…) but also in knowledge of a scientific character. Indeed, fallibility is something proper to man. A machine, properly speaking, never errs; according to Lambertino “error is to be charged, rather than to the intelligence, to the affective, passional, volitional component, conscious and unconscious, of human action, present in every formulation of a judgement, which denies the intelligence the rigour that research requires (…) Man errs insofar as he is a moral and free subject, not properly insofar as he is an intelligent being. It is, in fact, from an improper use of intelligence that error springs” (Lambertino, 1991).

The computer has, in any case, become an extraordinary complement to human intelligence, to the point of changing, in a certain sense, the way we work and study, and this is not in dispute. Rather than hardening into ideological oversimplifications, it would above all be highly opportune to reflect on the anthropological and social implications that this industrial revolution has produced, and is still producing, in our daily lives. Moreover, the cognitive sciences, born precisely in the wake of research on computers, help us better to understand the workings of the human mind. The mind-machine problem and the consequent admissibility of the theories relating to artificial intelligence remain open problems, even if, as Pratt has said, we still give too much weight to those aspects of our intelligence that best lend themselves to being replaced by the machine, such as skill at quizzes and rapid reasoning; but the retrieval of stored data and formal reasoning are things the machine does well and will do ever better. All this at the expense of those aspects of our intelligence that are considered peripheral and that psychologists call feelings; the capacity to calculate in no way distinguishes human thought, which possesses aspects, bearing on its own intelligence, that resist interpretation as representation. From many quarters, above all from the AI theorists themselves, a new Turing test is being called for, since they too have realised the need for a new and higher definition of intelligence, one that is not a matter of proving theorems or playing chess, still less of producing texts whose meaning the machine does not know. The human being, on the other hand, is obviously body, brain, matter, physicality, but is also, and above all, a person: a person who expresses the irreducible dimension of his own subjectivity and who makes reference to a personal symbolic and cultural system the privileged core for the interpretation of his own being-in-the-world. The AI project is ever timely and full of interest and suggestion, while on the other side of the barricade some reflections on the part of its critics attempt to give a more fitting account of our being human, from time immemorial one of the principal tasks of philosophy.

Bibliography

Agazzi E. Alcune osservazioni sull’intelligenza artificiale, “Rivista di Filosofia Neoscolastica” 59 (1967), pp. 1-34.

Alici L. Il futuro della persona e il dibattito sull’intelligenza artificiale, “Studium” 86, (1990), pp. 37-59.

Basti G. Il rapporto mente-corpo nella filosofia e nella scienza, Bologna 1991.

Davidson D. Rappresentazione e interpretazione, in Mente umana, mente artificiale, ed. R. Viale (Milan 1989), 40-54.

Fodor J. Il problema mente-corpo, “Le Scienze”, 151, (1981), pp. 100-110.

Hofstadter D.R. Gödel, Escher, Bach: An Eternal Golden Braid, New York 1979.

Lambertino A. Quale intelligenza?, in Intelligenza naturale e intelligenza artificiale, ed. S. Biolo, Genoa 1991, pp. 213-218.

Pratt V. Macchine pensanti. L’evoluzione dell’intelligenza artificiale, 1987, It. trans. Bologna 1990.

Putnam H. (1), Mente, linguaggio e realtà (1975; It. trans. Milan 1987).

Putnam H. (2), Il riduzionismo e la natura della psicologia, (1973), now in Progettare la mente: filosofia, psicologia, intelligenza artificiale, ed. J. Haugeland (It. trans. Bologna 1987), 221-235.

Putnam H. Ragione, verità e storia (1981; It. trans. Milan 1989).

Putnam H. Representation and reality (Cambridge 1988).

Searle J.R. Mente, cervello, intelligenza (1984; It. trans. Milan 1988).

Searle J.R. Menti, cervelli e programmi. Un dibattito sull’intelligenza artificiale (1980; It. trans. Milan 1984).

Turing A.M. Macchine calcolatrici e intelligenza (1950), It. trans. in La filosofia degli automi, ed. V. Somenzi and R. Cordeschi (Turin 1986), pp. 157-183.