Philosophie contre intelligence artificielle

Finally, this entire debate about AI raises too the problem of validity criteria – of the debate itself, or its topic –, in more or less general terms, about human beings or the machine. It is to determine what, in a given discourse, is about facts, mythologies, ideologies. One of the main computerization (and especially AI) advantage, consists in explicitly dealing with these questions. We also can think the habit of measuring computers performances, purely technically speaking, has a lot to do with it. Elsewhere, AI adversaries themselves are happy enough when they think they can turn back these criteria against it, without noticing they validate, at least, these methodological results:

“Minsky once suggested that in evaluating the programs presented in his book one might ask five questions: 1. Why were these particular problems selected? 2. How do these programs work? 3. What are their limitations? 4. What do the programs actually achieve? 5. How can they be extended to larger domains of competence?” [NOTE 150](Dreyfus, p. 131).

“At M.I.T., a grant proposal from this period [NOTE 151] distinguishes between ‘no holds-barred, special purpose, domain-dependent work’ and ‘no-tricks basic study’. And it seems to be generally accepted that every program we discussed in Phase III, and, indeed, the whole micro-world concept, was in this straight forward sense, a trick.” (Dreyfus, p. 27).

According to the principle of balance, if we accept the M.I.T. or Minsky’s criteria, a large number of philosophical works fit for the trashcan too. We wonders thus what are the philosophical criteria making the Turing’s intelligence test to be refused (Dreyfus, pp. 72-73). Indeed, the project of the Turing’s test is to evaluate the machine ability to simulate human behavior. From this point of view, it has at least the advantage to provide an independent control. Although originally, it was not about intelligence (but gender), it still is about assessment of performances for human beings and machine. But even in this intelligence area, in what the current philosophers would have a monopoly of validation, or even of competence (which would rather be psychologist’s one)? One cannot argue any more against the DP men or computer scientists’ excess of pretension, because it seems, on the contrary, that the people most motivated by AI are again too modest:

“H. Dreyfus […] notices the results already obtained are not so brilliant. It is necessary to admit that it is true, and if I had written this book after being subjected to an injection of truth drug, I would have most possibly been even harder than him, including about my own works.” (Jacques Pitrat, “Discussion,” in Dreyfus’ French edition, p. 434).

In general, the evaluation is relative to calculable results, but then it is precisely a function of AI to organize or to spread calculability. Any partial result is total successes from this point of view.

Traditionally, the first criterion of validity of a discourse depends on internal coherence necessity (consistency) studied by traditional and modern logic. Its principle allows considering communication as grounded on controllable formal criteria. Obviously, logic is a knowledge it is indispensable to salvage. Elsewhere it is already the case in the very wiring of microprocessors. The most interesting point of data processing seems precisely the approach in terms of encoding, in which control procedures are integrated from the very beginning, from the binary level until the instructions one:

“A programming language should allow proving that a program written by using it is correct. It should therefore be defined formally in a perfectly clear way.” (René Moreau, Ainsi naquit l’informatique [The Computer Comes of Age], p. 188, my translation).

So, AI has the merit to constitute an explicit formal framework for philosophical debates themselves. As much it has been previously admitted that it embodied the realization of classic philosophy. Data processing has a riddle value of reasoning, as Weizenbaum naturally admitted it:

“One of the most cogent reasons for using computers is to expose holes in our thinking. Computers are merciless critics in this respect.” (Weizenbaum, p. 65).

Duly noted! Here is certainly the real cause of the resistance to data processing from the Weizenbaum’s intellectual clientele. Would one want otherwise express the debates and the problematic by making them uselessly complicated and undecidable, like “most philosophical theses” (Searle, p. 30)? AI is therefore the correct (well-formed and current) language in which problems are raised today. And this definition of language should please a Heideggerian. But, for the creed followers, an expression like “to dwell in language” only means to use a literary, religious or poetical language. Nevertheless, if this phenomenological phrase were valid, the DP men or computer scientists, scientists of natural or human sciences, would dwell too in their own language.

In the worst case, the existence of this debate about AI will have had, at least, the advantage to develop an interest to artificial and natural languages. And one of the results is precisely the generalization of these control procedures of consistency, conceptually or discursively speaking. Elsewhere, the linguistic or semantics competence, mythicised by the intuitionists, is widely identifiable to the mastering of these (grammatical or semantic) procedures of control, we have reminded in the previous chapter.

It is also obvious that rationality (the long chains of reasons) is opposed to intuition by its orderly characteristic, and this amounts to denying the Gestaltist’s understanding Zen flash. The converse literary/intuitionist attitude often consists in resorting to abracadabras or occult causes. But, it is necessary to confess that the intuitionist attitude against the formal criterion can be often explained by too many restrictive criteria (streetlight principle style) that declares inadmissible everything which isn’t already formalized. In order not to be confined to the restrictive tautology of the scientific discourse already controlled alone, it is possible to consider internal validity as a judicial procedure that simply defines the conditions of admissibility. Data processing simply allows to increase procedures of control (especially for a written text), more easily than in relationships between human beings, in which a finicky control is judged unbearable. We can understand this way the expert systems reasoning justification mechanisms, allowing controlling how they reach a conclusion.

The entire debate could have rest on this initial confusion of genre between literature and rational construction. But the solution cannot obviously consist in getting rid of logic. We know Piaget, who admits its limitation too, has very well expressed the necessity of a formal level:

“It should not be such a mental meaning except in relying on practical operations […]. From this point of view, formal logic is not an adequate description of any thought as experienced: formal operations exclusively constitute the structure of the final equilibrium, to which practical operations tend.” (Piaget, La psychologie de l’intelligence, p. 160).

Formalism alone cannot be synonymous with scientificity, contrary to epistemological contemporaries habits. The same author points it out again:

“Because of its ‘simplified’ characteristic, no axiomatics can either claim to ground, or especially to replace the corresponding experimental science.” (Piaget, La psychologie de l’intelligence, p. 35).

It is these limitations of formalisms that precisely involve the combinatory problems: the possibles are too numerous if they are not limited by the real. We could admit, from this point of view alone, the Weizenbaum’s objection to formal work (p. 116). He thinks the practical programmer behavior enjoys a too big liberty of creation of potential worlds, until what he calls compulsiveness, with a principle of corruption possibility proportional to the power extend. But this point of view seems distorted by a partisan mind. On the one hand, it has already been noticed that bugs (programming mistakes provoking program crashes) are sufficiently frequent not to give an omnipotence feeling. On the other hand, about virtual universe themselves, the critique does apply as well to literature [NOTE 152], to music and to (contemporary) arts. Their esthetics is elsewhere the result of the constructivist Kantian epistemology, which should have to be the natural addressee of this critique. His (usual) contradiction lies in the fact that furthermore Weizenbaum glorifies the artist’s, or human being’s, absolute liberty, art for art, or science for science (in the beginning and the end of his book). While the formal model interest is precisely to test the limitations of theories by the study of their ultimate developments, as nevertheless again Weizenbaum has emphasized himself:

“As Patrick Suppes remarked, ‘The attempt to characterize exactly models of an empirical theory almost inevitably yields a more precise and clearer understanding of the exact character of a theory. The emptiness and shallowness of many classical theories in the social sciences is well brought out by the attempt to formulate in any exact fashion what constitutes a model of the theory. This king of theories which mainly consist of insightful remarks and heuristic slogans will not be amenable to this treatment.’ [NOTE 153]” (Weizenbaum, pp. 152-153).

Without having any scientist illusions about formalisms, Piaget also knows how to recognize the competence allowed by the acquisition of abstraction [NOTE 154]. He especially brought out the formal limitation of intuition (in the famous example of children tempting to evaluate the capacity of glasses of different shape). Piaget explicit then the intuition limitations not allowing handling abstract concepts representing the real world:

“Insofar as [intuitive thought] imitates real actions by vivid mental experiences, it clashes against this obstacle that actually (for instance) we cannot know how to build two necklaces with the same elements, while, insofar as [operative thought] proceeds by internalized actions becoming entirely reversible, nothing prevents it to make simultaneously two hypotheses and to compare them.” (Piaget, La psychologie de l’intelligence, p. 144, my translation).

The phenomenological claiming of the abstraction or formalization refusal can simply amount to learning refusal. Unless it simply concerns a feeling of world disenchantment. Through the apology of intuition, phenomenology, or Bergsonism, thus only models the pre-conceptual infantile stage.

The other pole of the traditional dichotomous limitation of criteria of validity concerns the discourse external correspondence (relevance or objectivity). But AI adversaries adopt about this a radical standpoint. The mere possibility of any objective correspondence seems denied by the refusal to consider knowledge and language in their function of representation of the world, as we already have seen in Dreyfus. But he is not the only one:

“At its simplest, the rationalist view accepts the existence of an objective reality, made up of things bearing properties and entering into relations. […] Thinking is a process of manipulating representations. This naive ontology and epistemology is one of the central issues for both Maturana and Heidegger. Neither of them accepts the existence of ‘things’ that are the bearers of properties independently of interpretation.” (Winograd & Flores, p. 73).

We do not see very well in what these interpretations are not representations. But we observe also a tendency to the negation of the world itself. Nevertheless, it would be insulting to phenomenology not to grant its desperate quest of inking in the world. But oddly enough, it clearly takes a stand of denying this capability to computers or to robots by referring to the body. In fact, phenomenology considers the case as over and done its partisans can satirize (with an academic irony):

“Few philosophers adhere to the naive view that one can assume the presence of an objective reality in which objects and their properties are ‘simply there’.” (Winograd & Flores, p. 18).

But we have to admit that the obviousness generality is rather a rhetorical turn of phrase, because the claiming of the contrary doesn’t seem bothering, as soon as it is about valorizing as laboriously one’s own beliefs, by satirizing against “many sophisticated linguists and philosophers” (Winograd & Flores, p. 61) sharing these reactionary objectivist ideas [NOTE 155]. Actually, the problem maybe lies simply in the confusion between object and word. In a given language, an object is not always referred at by the same word, and reciprocally a word doesn’t always refer to the same object. But that supposes a fussy idea, and especially false one, of discourse analysis. Isn’t it enough to grant a minimum of validity to the common language? And that can lead us to accept a realistic correspondence, against useless attempts of labored overcoding (Winograd & Flores, pp. 48-49) in which linguists and philosophers of language ended by getting bogged down in glosses and quibble. As much this supple correspondence was admitted (idem, p. 17) and that it is admitted later that nouns do refer to objects, and verbs to actions (idem, p. 174). After all, if the word “dog” doesn’t allow distinguishing every dog, it allows distinguishing dogs from cats. This is said in jargon: “A language […] is constrained by the nature of the world to group a certain set of objects or properties [NOTE 156] together under whichever names it uses.” (idem, p. 61). The relativist way of understanding arbitrary of sign, encouraged by some linguists, appears to me as a misinterpretation. The chair on which I am sitting, and its characteristics which I use, seem to be said real. We guess what reductionism, grounded on a doubtful interpretation of quantum physics, allows thinking something else. We notice elsewhere that the “ontological problem” is as naively admitted as real!

Anyone still doubting of the correspondence of words with things, can remind the scene of the famous movie, Miracle in Alabama, about the life of the born blind young girl:

The partisans of the negation of the real are free to gouge their own eyes to see the difference. We can origin this subjectivist questioning in the philosophico-esthetics glosses about color classifications. As Dreyfus (pp. 175-176) reminds us this debate by wondering how to program the distinction of colors. But, besides the fact that linguistic differences (also reminded on that occasion) do not forbid a discrimination within a culture, colors are perfectly internationally calibrated. It is easily controllable by ordering a car, red for instance. Would an artist, a philosopher, or a linguist, to whom a green car is delivered, would admit that, with a certain light, it has a red tonality all right?

The definition of all the acknowledged elements of any concept already exists in professional nomenclatures (like graphic technicians’ colorimetric calibration). The Encyclopedia model, a Leibniz’s project, has been realized by 18th century philosophers who weren’t scared by cataloging. And the classification of knowledge has no reason not to produce some structuring. Anglo-Saxons people benefit since 1852 of the Roget’s Thesaurus already proposing such a classification by notions, which can allow, it is true, treasure hunts appearing infinite [NOTE 157].

The problem of realism is reduced today to the exploitation of the patiently accumulated knowledge, subject to acknowledgement of other people’s work! The constraint of expertise or knowledge elicitation is satisfied then. It suffices to the supporters of AI to read the Dreyfus’ “review of difficulties” as “hints” (p. 37) already available (for ‘well posing a problem is solving it’). Simple difference between pessimism and optimism.

One of the motivations of the objectivation refusal is also the meaning shift about the term object, which would want that who is objectified is magically transformed into some thing, a little like in the legendary case in which a photograph imprisons the savage’s soul in exotic stories. Dreyfus even sees the western philosophical tradition of these last twenty centuries as the transformation of a human being into an object. We wonder however about this offhand synthesis of all the western philosophy, in which mechanism (nevertheless recent) seems to sum up any trends. It depends too what meaning is given to the term object: Heidegger’s Nazi friends have had a more radical way to deny human objects.

The term objectivity and its derived very obviously refer, contrary to this mystical pun: to relevance in general, to operationalization or pragmatism. The universe is limited by contexts, but it remains the objection of possible worlds, which derives of the interpretations infinity. We already have seen the risks of formal combinatory. It is therefore simpler to operate a reduction by means of an economic criterion: the real. The real can therefore be economically defined by the context as conventional, verifiable by observation.

Oddly enough, while phenomenology often tends to empiricism, it becomes formalistic while it opposes to science (positivism, techniques, etc.), when not simply reductionist. Actually, the philosophical refusal [NOTE 158] of observation, or induction, is equivalent to legitimate criticizing against excessive generalizations. But rejecting any observation method, for some practical mistake example, represents an excessive generalization too. This negation of induction is absurd anyway, because the epistemologists’ works are grounded on the positive science history, and more precisely on the (only) one known by most scientists or epistemologists, physics (or, possibly, biology).

Philosophers of the beginning of the 20th century had noticed that the Aristotelian/Stoic’s syllogism, model of deduction, rested on empirical data (besides the fact that such reasoning to exist before being formalized). Indeed, the fact that men are mortal, or that Socrates is a man, rest on observation and classification. The deduction becomes therefore a system of relationships between observations, a means of to classify empirical statements. This famous syllogism could therefore be rephrased, with the involuntary participation of a current adversary of empiricism: ‘All men have been mortal. Popper is a man. Therefore, Popper is possibly mortal’. Would this refusal of induction be grounded on the truth of this Freudian’s thesis, ‘finally, nobody believes to one’s own death’ [NOTE 159]? Actually, this opposition to induction only challenges the use of paradigms or typical situations, beloved of Dreyfus. For empirical experiences cannot be absolutely convincing (according to the classical or current epistemology).

The fundamental problem is rather the permanent experimental control of the correspondence of our inferences. In general, so that deductions remain valid, it is especially necessary not to forget what are the inductive operations that have led to these conclusions. Let’s notice that is the only ambition, since Bacon, of positivism:

“The real induction, ‘interpreting’ experience by lingering over for long and by not elevating to axioms except by a slow progression […] this interpretation of experience represents the very science, while ulterior deductions only aim applications of scientific knowledge to practical uses.” (R. Blanché, L’induction scientifique et les lois naturelles [Scientific Induction and Natural Laws], p. 7).

A non-controversial definition of positivism can be expressed simply by the creation of theory from observation. But, contrary to what its critique supposes, observation is often insufficient. Sociologists invite us judiciously “to abandon the contempt for social inquiry,” they consider as a: “thought habit inherited from the philosophical origins of sociology. […] A lot sociologists think as if their knowledge of the detail of functioning of their society was already sufficient.” (R. Ghiglione, B. Mathalon, Les enquêtes sociologiques [The sociological inquiries], p. 95). We can guess that philosophy, or those finding here a valorization, does not feel concerned. The refusal of a representation of the world simply opposes to what, in the epistemological tradition was called: “to save appearances.” This program should have satisfied phenomenology.

We also knows the other opposition to objectivity in constructivists, but their construction of reality, beloved of some sociologists, is not a kind of plot, or a fascism of language like for Barthes. Actually, in their works, this construction of reality seems to characterize only others people’s ideas. This is a regression to the accusation of ideology, in the controversial meaning, in the continuity of Marxism and its scientific pretensions. Their critique concerns especially the ideas contradicting their own prejudices, or merely their information, limited on principle. Another version of this construction does not contradict objectivation. It grounds the construction of personality, the relation to the world and to others, both as objective. This constructivism is implemented by the Piagetian’s work:

“A world without objects is such a universe there is no systematic differentiation between external and subjective realities, a world consequently ‘adualist’ (J. Mr. Baldwin). By this very fact, this world will be centered on own action, the subject remaining all the more dominated by his egocentric standpoint he is unconscious of himself.” (Piaget, La psychologie de l’intelligence, p. 122).

This methodological dualism is actually opposed to the extremist version of the Kantian subjectivism by the recognition of the exteriority of the world. Oddly, this reference to objectivity appears too in the book about Turing (directly concerning our matter, as founder of AI). His biographer incidentally recalls an anti-totalitarian purpose of the Orwell’s book 1984:

We see that Orwell identifies the presence of this constructivism in dictatorships, in that they do not want to admit an exteriority to the human being arbitrary power. This humanism (rather 18th idea) had simply replaced the arbitrary of God. And the phenomenological standpoint can be characterized too as a formalization of the only egocentric stage of intelligence, despite its opposition to rationalism claiming the subject’s autonomy.

In the usual epistemology, as we already have seen, the problem of the truth is only solved by the opposition between internal validity (consistency, coherence) and external validity (objectivity, relevance). But this traditional resolution precisely leaves a lot to be desired, because despite the erudite satisfaction of the conformist repetition of academic tradition, many other criteria of validity can be put in light. Precisely, phenomenology can be understood as a discourse essentially accounting for the criterion of subjectivity. But this option has shaped up in a German philosophical context, contradictorily marked by the taste for totalities, residue of the Hegelian tradition. To that is added the Kantian subjectivism, in which the subject imposes its categories to the world. It resulted a contesting of the notion of truth, by some classic solutions (rhetoric, relativism, ideology), or more modern ones (sociology of sciences, constructivism, phenomenology). This had simply led to the oblivion or the negation of any other possible criteria. These excesses are revealed sometimes, in some contradictions, or when AI adversaries oppose one another. We can understand this way the critique of the programmers’ abstraction/subjectivity by Weizenbaum, or his condemnation of magic, he identifies to be the refusal of randomness. Oddly enough, he assimilates the programmer to the follower of magical explanation:

“The compulsive gambler believes himself to be in control of a magical world to which only few men are given entrance. ‘He believes,’ writes Bergler, ‘Fate has singled him out… and communicates with him by means of small signs indicating approval and reproach’. [NOTE 160]” (Weizenbaum, p. 123).

We already knows that this critique should rather be intended for the famous Searle’s mental causation, which can be interpreted as the phenomenological version of magic. The mentalists’ internal signs are not a sufficing criterion of certainty in the relationships to the world [NOTE 161]. We can incidentally notice the Marxists’ omni-determinism has this difficulty too (under the expression: “Nothing happens at random”).

It was told the French corsair Surcouf hailed by an English adversary reproaching him to fight for the money and not for honor, retorted that everyone fights for what he is short of. Besides the corsair’s wit, this story indicates that problems concerning some statements have sometimes as a solution an indication about the speaker.

The legitimate concern about the observation of subjectivity by AI adversaries is, unfortunately for them, biased by their own epistemological situation (idealistic schema converted to the real). The holistic, innatist and fixist assumption (cf. Piaget, La psychologie de l’intelligence, pp. 23-27) rather had a tendency to deny body, what the ideal opponent invented by Dreyfus has never done, because usually, modeling cannot obviously deny the modeled object. From this starting point, the desperate effort of phenomenology in tempting to reintroduce the body has obviously nothing to see with the conceptual work of an empiricist, or even a reasonable constructivist.

The subjectivist concerns of phenomenology does constitute a human knowledge, which it is necessary to salvage. The phenomenon is usually elementary: it consists, for instance, in observing prosody or vocabulary markers indicating the social or geographical origin. It is precisely the case for Eliza Doolittle, the George Bernard Shaw’s Pygmalion heroine, used as model by the Weizenbaum’s ELIZA.

This Surcouf’s principle, according to which a statement doesn’t provide only information on the topic, but also on the speaker subject, and more generally on the context, is perfectly codable in an artificial or natural intelligence framework. This would be the link with some formal models allowing to analyze sentences of the type: “I think,” “I believe,” “I want…,” expressing the point of view of the subject on the world, or one’s real life experience, without denying an objective referent. But we are not bound to be limited to this kind of statements, which are formalizable anyway, by the old or modern (modal, deontic, fuzzy, etc.) logic.

We know too that, traditionally, the human sciences methodology, or the epistemology formal one, depreciate subjective testimonies by invoking testimony errors or optic illusions. But the counterexamples to this carefulness exist too. In law or history, testimonies are actually accepted. And one is found in very bad society of negationists, or quack attorneys, in persisting in formalism, or in researching mistakes in the drafting (of a close procedure) [NOTE 162]. A book about the World’s War I studied this testimonies validity from numerous letters and literary work relating, in the nineteen twenties, souvenirs of war:

“In spite of this amazing diversity, we can observe that their testimonies, each with its individual touch, shape a unique portrait of the war, in which everything harmonizes with. The lack of contradiction is a fact so new, and however so obvious, that it obliges us to revise ideas that has imposed to us the classic experience of the unavoidable dissimilarity of depositions of accident witnesses. An accident only last some seconds and human abilities cannot record phases fleeting like cinematography. Every witness instinctively, and following his own nature, completes the strings of rapid phases whose several ones escaped to him. […] But a battle, the life into the action, is not an ephemeral accident. The lasting was sufficient to allow the adaptation of senses and intelligence, the correction of the mistake of the day before by the clearer experience of the following day. Instead of rapid phases, everyone different, there was the monotonous and almost identical repetition of stormy days […].” (Jean Norton-Cru, Du témoignage [On Testimony], p. 29).

The necessity of recording testimonies, despite their widely known weaknesses, is still obvious. AI precisely allows considering it with the expert system category, which can practice the elicitation of daily expertise. We already have seen that a business information system is implemented this way. From this point of view, any computerization is already artificial intelligence by any other name, but truly working daily. Even the so-called refusal of testimonies or of empiricism simply forgets to consider established knowledge, or informal conversations accompanying the pure experimental work. Winograd and Flores are right to emphasize them, but they obviously go to the other extreme of the formalization refusal.

Mentalism regresses then to the traditional method of philosophy consisting in informing the world. Intentionality seems to be the idealistic answer, even solipsist, to the theory of action, as we can explicitly notice it in Searle:

“Visual and auditory experiences, tactile sensations, hunger, thirst, and sexual desire, are all caused by brain processes and they are realized in the structure of the brain, and they are all intentional phenomena.” (Searle, p. 24).

The philosophical intention immediately jump from the fact of having an object to the confuse claiming of being caused by the brain. This idealistic idea is again more apparent below, as approaching parapsychology: “When you have a thought, brain activity is actually going on.” (Searle, p. 25). But if we can raise our arm when we want it (if not hindered), or to give an order to a human being or a trained animal, we cannot move an object via telepathy, neither change the weather. This characterizes this subjectivism is the non-distinction between these different states of the resistance of matter to our will. Differences of nature between material bodies, living ones (animals, other human beings, himself), can be characterized by differences of means of modification of the world: action on the matter, communication with other living being, intention with our own body. We can see now that phenomenology simply proceeds by generalization of subjectivity to the relationships to the whole world. This is the origin of parapsychology. Mentalism is therefore understandable as an explanation exclusively valorizing subjectivity, or as inducing confusion by using the same lexicon.

Despite the Winograd & Flores’ explicit claiming of resistance to solipsism, the reference to the only subjective framework obviously induces a drifting to this pathology. Especially when they dispute (p. 30) the language informational validity for the sake of subjectivity of interpretations or the uniqueness of situations. Actually, this theory (coming from linguistics) excessively generalizes the analysis of deictics (element relative to the situation, like: “here, now, this, that…,” pronouns, etc.) to all the elements of the discourse. This theory artificially cancels the fact that a context is generally known by speakers. In the debate of AI, it plans to hide this context to the machine to reproach it not to guess (a human being cannot tell any more what means: “I give him that here”). The reality is simply grounded on the accumulation of indices, and of former knowledge of who has to understand a statement or who has to interpret one of its fragments. The model is then the archaeological research one (without reference for the usage made by Foucault, whose theory fell in the same indeterminist and constructivist derive, with some belated regrets).

In subjectivist considerations of this kind, any reality seems to disappear in aid of the only mental state if one grants too much importance to the subject or to one’s supposed mental states. Finally, the biological reductionism does ends to the solipsist hypothesis claimed to be avoided: “From this [Maturana’s] perspective, there is no difference between perception and hallucination. […] From the standpoint of the nervous system, it is not a relevant, or even possible, distinction.” (Winograd & Flores, p. 43).

A system of representation of the cognitive activity allowing differentiating these two states should rather be adopted. It is certainly the demonstration of the reductionism limitations, and of the inanity of the scientist pretension of biologists (as Maturana used as a credential by these authors) acting as overnight sociologists or linguists, because finally, the cultural or personal subjectivity is actually getting bogged down in its contradictions. Due to fact that everything could be considered as relative, it does not obviously derive that everything is true, and especially not that everything is absolutely true! It can simply mean that everything is more or less true, and that it happens to be some means of verification or increasing guarantees.

The criterion of subjectivity, as complement of formal and objectivity ones, could not be considered as the only alternative. The practical method of decision-making, or simply of action, can be opposed to abstract reason by limiting the subjective liberty by the existence of others’ opinion, by cultural habits, and the burden of tradition. Contrary to the very idea of breaking down with tradition, intersubjectivity does characterize sciences themselves. Science is equivalent to a consensus [NOTE 163]. The principle of experimental or formal demonstration is simply, from this point of view, identical to the capacity to convince. Otherwise, it is sufficient to impose any opinion – what was traditionally the dogmatic method of religions or dictatorships.

The principle of all belief is equivalent to the inter-subjective criterion of validity. We can admit that the situation of belief is permanent, as claimed by Peirce: we only replace beliefs by other beliefs, whatever knowledge it is. The purpose can be, as he asserted, to find again the initial state of unity. The important sociological consequence is that an intersubjective unit is necessary to have a collective belief, since collectivity is a prerequisite to debate, and to language itself. But the surest means to produce sociologically belief is by never contradicting anyone. As Searle said innocently: “If I am right, what I say should seem obviously right” (p. 62). The criterion of validation of such an approach seems solely grounded on the self-hypnotic ability or on conformity to prejudices – holistic here. This is the only way of analyzing the amazing revelation below:

“Language, as consensual domain, is a patterning of ‘mutual orienting behavior,’ not a collection of mechanisms in a ‘language user’ or a ‘semantic’ coupling between linguistic behavior and non-linguistic perturbations experienced by the organisms. […] In using language, we are not transmitting information or describing an external universe, but are creating a cooperative domain of interactions.” (Winograd & Flores, pp. 49-50).

In the intellectual process of jargon, we can see at work the same type of redefinitions of the world than those specific to conditioning methods of sects. It is precisely a case of a language becoming a “mutual orienting behavior.” These totalitarian trends, unfortunately, often exist in the research area and among intellectuals in general. It is the foundation of the ideological process, which can be consider as a perverse effect of this status, and which is well enough described by the constructivist theory. In this last quotation, it doesn’t concern an external reality, but an autobiographical confession all right. Professional or amateur psychoanalysts can appreciate that some discourses, more elaborate than an acte manqué or a slip, are so revealing. It is a fact this discourse does seem to formalize an objective reality, even if the matter in question is the delirious subject.

Thanks to the elaboration of this intersubjective criterion, we can actually understand the Winograd & Flores’ standpoint. They take themselves the trouble to specify how their linguistic constructivism doesn’t end into solipsism that would be an obvious risk outside this collective framework. Their solution (borrowed to Maturana) is in “a consensual domain – a domain that exists for a social community. Reality is not objective, but neither individual.” (Winograd & Flores, p. 51). But these statements (and those following this passage), on the famous Sapir-Whorf’s hypothesis mode, finally appear tautological (a different culture is a different culture, the solipsism doesn’t exist when we communicate). Our two authors simply forget to consider the possibility of a collective solipsism. They can summon Maturana, Gadamer, Heidegger and Habermas, to root the meaning of language in a social tradition. The framework of the debate is clearly weakened if, as the last one said: “In order to distinguish true from false statements, I make reference to the judgement of all others […].” (idem, p. 62). The intersubjective criterion considered like the alpha and the omega doesn’t allow distinguishing truth from false, but only orthodoxy.

The choice of the language act linguistic theory (idem, pp. 35, 143, etc.) seems less appropriate than information theory for our topic. We can consider that information precisely gives what Maturana calls “mutual orienting behavior,” as it provides nontrivial data about the environment, or about speakers. If there were no information transmission, language would not exist. Thus communication is at the basis of recognition of others involving the social commitment in question, and lying on “a precise correspondence between combinations of words and the structure of commitments listened to in a conversation” (idem, p. 159). These relationships should be analyzed by a computer when it concerns sentences as: “John kills Peter,” “Peter kills John,” or simply “John promises Peter to come.”

The revival of the topic of manipulation or the deception is grounded too on this community negation of the real. But the possibility of deception already existed before computers, partially because of its absence, because discernment, too, should be automated (kind of semantic anti-virus software), and its distribution then could be generalized to any individual. This was certainly not previously the case when each individual usually depended, besides its own discernment and its personal experience, upon local community authority mechanisms. The phenomenon of rumors is entirely grounded on possible distortions by human mediators!

Anyway, the cultural obviousness of intersubjective agreement doesn’t exclude the machine itself. From this point of view, “the creation of a consensual domain of behavior between linguistically interacting systems through the development of a cooperative domain of interaction” [NOTE 164] (Winograd & Flores, p. 50) only generalizes one of the language effective characteristics, determining inter-understanding, to any natural or artificial communication agent, when this very theory was supposedly demonstrating the opposite.

We already have seen that the problem of the argument from authority is often badly stated, as indicated by its quite peculiar usage by the DREYFY method, in which it becomes inverted in a justification of one’s own opinion (Lazarsfeld’s principle). Weizenbaum obeys this mechanism too (on a less subtle way than Dreyfus does), when he has to justify his anti-scientists opinions (Weizenbaum, p. 11). Its argument amounts in short to: experts think like Weizenbaum, therefore they are right – or even Weizenbaum is right to think like all the authorities who think like him! Anyone perceives immediately too a reservation to the authority principle in case of “they differ among themselves and even disagree on many vital questions” (idem). The uncertainty deriving from these contradictions is nevertheless important to decide what is the legitimate authority [NOTE 165]. In this question of authorities conflict, we are driven to historical or geographical, cultural or ideological relativism. But the intersubjective criterion of validity is usually limited to the authority of tradition.

However, if the sociologist of knowledge has to observe the excessive usage of the argument from authority, he has to admit its obvious validity too. The argument from authority is on principle receivable as it is preferable to see a competent person. And the current crisis of science, or crisis of truth, is first a crisis of the scientific authority, whose symptoms are the end of the churchy legend of the great scientists, and the end of the myth of saving progress.

The computer actually could have been considered as a candidate to this return of authority. From this point of view, the only reproaches addressed to data processing seems to lie in the too large confidence that we risk to grant it. Weizenbaum (pp. 40-41), is surprised indeed that our confidence in the machine is undermined only after several mistakes. But one doesn’t see where the problem is. The judgement about an individual would be the same that about a machine: one will consider his/its competence called into question only after some repeated mistakes. It is therefore demonstrated that what we consider in both cases, it is a calculator – i.e. the assimilation of a mechanism, an algorithm of calculation. More, the lack of self-confidence, or the abdication of criticism, is to be attributed to the human subject alone. They both exist in a human bureaucracy too. The problem raised by Weizenbaum can be reduced, more specifically, to the question of the bureaucracy’s political prerogative, or to the question of the rationality of any power in general. The confidence granted to data processing, even if he thinks it disproportionate, is not different of the one granted to any governmental specialty (like quantum physics), as he finally confesses himself (p. 247).

From this point of view, his references to AI, contradictorily to his own pretensions, are therefore only a metaphor: the refusal of technocracy one. A metaphorical use of his own computer field is equivalent to the way of raising a problem for a lay person in politics. As the political processing of the questions precisely belongs to the fields in which his target, Herbert Simon, is one expert: i.e. science of organizations, problem solving or decision in state of uncertainty.

More generally, the same problem of research program justification exists too concerning quantum physics or any scientific practice submitted to the democratic representation authority – and to public opinion. This functional credulity of the decision-makers is not new neither undue. Earlier, kings do trust astrologers. And current managers do rely on pseudo- sciences in order to judge human beings. In France, 93% recruiting officers use graphology (Cf. French monthly Science et vie [Science and Life], March 1993)!

More, if Weizenbaum (pp. 27-29) asserts we can do without computers, he also forgets to describe some bureaucratic consequences: the number of necessary civil servants for replacing computers would possibly generate an undesirable type of organization [NOTE 166]. As a computerization advantage is to solve problems of organization too. Organizational problems appear then with a measurable form in terms of system performances. The lack of automation even could be a technical explanation of the planning communist ambition, applied to the former USSR, in the absence of adequate means. The failure of leftist utopias would derive, from this point of view, of mere voluntarist anticipation. Thus, reproaches addressed to bureaucrats did not concern their malice. This was not acquired by an observation in term of human behavior, due to classic psychologization or moralizing guilt. While in the intersubjective framework, the problem raised by the authority legitimacy is especially about an adaptation to a possibly unhealthy environment.

An important reservation to the authority of expertise is conformism too, which can be called the good pupil’s principle in the academic context. It consists in repeating like a good boy/girl what is expected, and to orient all one’s intelligence to satisfy this expectation. And the weakness of the educator consists in being caught to the docility. Piaget notices it concerning logic of children:

“Constraints of the child’s circle would not suffice to generate a logic of his/her mind, even if truths they impose are rational in their content: to repeat right ideas, even in believing they come from oneself, is not logical reasoning.” (Piaget, La psychologie de l’intelligence, p. 173, my translation).

The good pupil’s principle only reveals the normative behavior of who recognizes a disciple. And it is most possibly too one of the limitation of pedagogy. This confusion between reference and reverence is a Weizenbaum’s specialty, most possibly due to his reverential reference, in which the academic conformism sometimes rival to religious illusions:

“I am acutely aware, for example, that there is nothing I say in this book that has not been said better, certainly more eloquently, by others. But as my friends continued to point out to me, it seemed important to say these things again and again. And as Lewis Mumford often remarked, it sometimes matters that a member of the scientific establishment say some things that humanists have been shouting for ages.” (Weizenbaum, p. X).

To say what your friends say is not a criterion of truth but only a conformity one, perhaps of sectarianism if, like here, you do not compare several points of view, and if you admit turning over common places. His big expositions on Mumford’s beloved topics amounts to the disciplinary abuse of mutual quotations or of sending the lift back. Indeed, it is important not to forget that such a conformism reigns in research too, in which it was not indeed expected. Simon noticed, having badly assessed people’s interest about AI:

“The only mistake we made was to overestimate of how this field was going to fascinate people and trap them into working on it. We just couldn’t understand how anybody could stay out of it. […] There are possibly more timid people in the world, even in science, than one likes to believe, people who like to do things in well-structured environments where there already is a paradigm to work in.” (Pamela McCorduck, Machines who think, p. 190).

The mistake possibly lies on the scientists’ reciprocally overestimating. They often forget their hyper-specialization doesn’t give them an overall vision. Those, like Simon, possessing such a vision, naively generalize their own case. We can observe too an odd academic behavior consisting in knowing to be ultra-specialist, while simultaneously astonishing about ignorance, in one’s own private area, from other ultra-specialists of different fields. Latour and Woolgar report on this anomaly (without lingering at) in their observations of research workers’ conversations (La vie de laboratoire [Laboratory life: The Construction of Scientific Facts], p. 75). We can observe then academic members’ behavior consisting in having an extremely reserved attitude in front of factual data they ignored! This should encourage upgrading of epistemological popularization, or interdisciplinarity necessity. This resemblance with common sense really characterizes the protection of prejudices (possibly called paradigms). That could have been a very consequence of a methodological practice, correct on principle, as another author stated:

“It is necessary indeed to begin by limiting and by delimiting a problem in order to study it scientifically. But it is necessary not to forget that such a demarcating is often arbitrary because it is caused by constraints of research or because it is linked to a momentary interest of such research worker or group of research workers. But once a problem is delimited and raised, other research workers rush into it without questioning the arbitrary of this demarcating, contributing to fix the problem frontiers. Such a fixing leads to limit the very problematic.” (Pierre Moessinger, La psychologie morale [Moral Psychology], pp. 124-125, my translation).

The consequence of this conformism is that scientific credibility is granted on nonscientific criteria, like membership to the same team, but more frequently on ideological friendships. Each ideological camp has its own experts who are only organic intellectuals then. The problem is obviously not liberty of opinion, but the ability of contesting nonstandard facts, or more formally, the epistemological incapacity to integrate them.

We can observe then, in research, or in arguments with methodological pretension, an adjective being used (in French) as indicator of this conformism tendency: “serious research workers” (we can recognize the reverential cliché referring to “very serious” dailies like Le monde, The Times, etc.)! We can guess the authority conferred to these characters whose professional status remains to be defined. Of course, we find this criterion in our so serious authors:

“Would any set of syntactic rules, however detailed, and any computing speed, and any size dictionary suffice to produce high-quality translations? Every serious worker now agrees that the answer to this question is simply ‘no’.” (Weizenbaum, p. 186).

It should have been preferable to begin with them, and not by the comical research workers. Both are easy to identify, whether or not they are in our side. The DREYFY method make the process more sophisticated by using rallied authorities of “two of most respected and best informed workers in the field of automatic language translation” (Dreyfus, p. 107) against the very competence entitling them. These serious research workers obviously obey the consensus criterion, general or sectarian. This leaves a choice between normal science and coterie, or perhaps opportunism described by some sociologists of sciences.

The sectarian conformism can induce more specifically the production of ritual credo. We have often noticed (sometimes to question it, of course) that phenomenology was the continuation of theology by other means. It indeed opposes on principle to any positive processing, not to say rational, obviously supposing analysis. This stylistics hesitates between pedagogical paraphrase and (auto) hypnosis. Of course, a credo would not be a real credo without a blasphemy possibility. Here, Minsky is the guilty one, against intuition and holistic grace:

“The habit of our culture has always been to suppose that intelligence resides in some separated crystalline element, call it consciousness, apprehension, insight, gestalt, or what you will but this is merely to confound naming the problem with solving it.” [NOTE 167] (quoted in Dreyfus, p. 210).

The unfortunate Minsky does commit a blasphemy which condemns him, fundamentally because he dared to pronounce the holy words without having immediately been enlightened. The fact Minsky has stated the verbalist mechanism underlying this standpoint would be infinitely more serious if followers of this belief were able to understand the importance of his analysis of the magical and philosophical process. This magical belief is the entire intuitionist standpoint, because it is a revealing sign: partisans of intuitive comprehension are precisely those whose cognitive mechanism is characterized by a production of abracadabras. And it validates partially their theory to the description of their own pathology.

But the fideist strategy lies in demanding the recitation of a credo, under an always more orthodox form. The softened version of anathema is rectification, as a correcting more or less indulgently a bad-good pupils’ clumsiness (“His computer model of formal program […] leads Neisser to betray his own Gestaltist illustration.” Dreyfus, p. 245). The believer, i.e. who is reciting a catechism, is caught in the act. One does not acknowledge the possible effort of syncretism, and the heresy is condemned. It has to be noticed the failure of the incomplete convert only characterizes his belief level and not the validity of the system to which he is tempting to adhere. The other imperative is the double bind of originality. The good naive, contrary to the bad one, has to rediscover the most extremist ideas of the sect:

“This leads Oettinger to a very phenomenological observation: ‘Perhaps… in perception as well as in conscious scholarly analysis, the phoneme come after the fact, namely… it is constructed, if at all, as a consequence of perception not as a step in the process of perception itself’.” (Dreyfus, p. 239).

But (besides the extrapolations) is his statement an inference from facts or a deduction from an already admitted theory, phenomenology here? Maybe the epistemological state of justification context (contrary to a discovery context) is analyzable in term of self-justification context. And more generally, the valorization of intuition against demonstration or formalization clearly looks a lot like good pupil’s discourses, quietly repeating what their good teacher says. So, computers will be intelligent when they will agree with phenomenologist philosophers.

Besides the psalmody redundancy, the general method of belief is the confusion of fact to prove and of the proof. The mystery is self-sufficient as proof of subjective conviction, of faith. We can recognize elsewhere the tautological style of the protestant minister preaching to an old maids bunch: “I could be doing something else. […] This, I submit, is the source of our unshakable conviction of our own free will.” (Searle, p. 95). And the preacher can conclude in the most blandly possible way, having reassured the metaphysical vertigo of his flock. Elsewhere, the practice that consists in preaching to the converted seems to be general in this environment:

“I could go on to describe the later stages of the socialization of the individual human, […]. But that could be of no help to anyone who is already convinced […].” (Weizenbaum, p. 213).

“We do not attempt to provide a philosophical exposition and critique in which arguments for and against each position are enumerated and weighed. We find more fruitful to present the central points, listening for the relevance to our own concern.” (Winograd & Flores, p. 9).

This funny way of arguing is revealing that it is less about a contradictory debate than to provide (pseudo) arguments to believers. But the conviction is finally vanishing when the principle of uncertainty is reigning. The confession is almost touching:

“I had once hoped that it would possible to prove that there is a limit, an upper bound, on the intelligence machine could achieve […].” (Weizenbaum, p. 206).

The whole Weizenbaum’s book is grounded on this lack of proof. It solely originates in desire of reinforcing a belief and denying any comparison. It thus attributes to others its own insufficiencies when it contradictorily assumes that those starting researches do refuse refutations. We can think that Weizenbaum doesn’t want/can go further himself because he has reached his own limitations, as a lay person, like “the computers expert who knows nothing but computers (the Fach Idiot as the German call such a person) [who] can derive no broad intellectual nourishment from his expertise” (Weizenbaum, p. 160). But the principle of modesty can sometimes play tricks when you confess in a same time: “I am professionally trained only in computer science, which is to say (in all seriousness) that I am extremely poorly educated” (Weizenbaum, p. 8). What characterizes this author is precisely not to be serious, in wanting to have the phenomenologist Polanyi’s cake and to eat the Stalinian Bukharin’s one [NOTE 168].

In addition to numerous contributions to applied epistemology, data processing finally allows to solve the practical manifestation of the argument from authority. We already have seen the professional philosopher characteristic can be purely and simply reduced to a referencialist indication. For the indisputable necessity of using or identifying references, amount to document processing obviousness. It is therefore sufficient to index any statement with his author’s name. That allows too reckoning credibility indication, by concordance. One can even add an inter-quotation structure. More, stocking and processing capacities theoretically allow (from heighties), for an average institutional library, to work on the whole available reference on a topic. And before long, by the distribution of techniques already existent (in 1994), it will be possible to own the quasi totality of human knowledge in a personal library [NOTE 169]!

Let’s notice that these techniques secure intellectual property too, at least by an exact indexation. But then authoritative intellectual practices can be precisely understood as attempting of securing property by institutional, dogmatic or bureaucratic means. This is the meaning of bureaucratic structure, or corporatistic and sectarian ones (especially by the esotericism of jargons). Is this institutional situation of property in public services what makes this other AI opponent emphasize?

“The antagonism between computerists’ values and security constraints […] is a professional version of a more fundamental opposition between the free data processing, in the free circulation of the information meaning, and economic liberalism. When information becomes a merchandise, it has to be protected. […] Some people are attached to the norm [of liberty] they become partisan of delinquency, as much over-lenient or liberty of the information defenders.” (Philippe Breton, “Une contradiction insoluble entre deux systèmes de valeur [An insoluble contradiction between two systems of values],” Le Monde Informatique [The Computer World], March 1993 15th, pp. 50-51, my translation).

His analysis is obviously false since the computer business is precisely a case in which intellectual property has produced huge fortunes. Bill Gates,’ Microsoft CEO, has become the first one in the USA within less than 15 years. It is as obvious as there is no contradiction between economic liberalism and the free circulation of information. Industrial property is precisely defined by the publicity of patents. A large-scale distribution can lower production costs too, and prices for customers then. Conversely, rarity leads to an information and culture concentration in elite’s hands – who can deplore mass distribution then! If the question of piracy is indeed a problem, it can be solved by various means. Elsewhere, it has to be noticed that book is also hackable (plagiarism, photocopy). Plagiarism as source of creation exists too as explicit in the literary or in fine arts fields. An author even plea for its apology (under the rubric “Ideas” of the daily newspaper Liberation, concerning the case of one Jacques Attali’s books, Verbatim). And a Parisian painting exhibition in Louvre, called Copier-Créer [Copy-Create] (30 April-25 July 1993), sums up this topic by the right to inspiration, or by the resumption of eternal topics. This, incidentally, relativize too the innovative pretensions of literature, or human creativity itself. The Breton’s book, The computer tribe is itself a repeating of Weizenbaum’s arguments, with a Marxist rather than religious background.

Theses phenomenological theses can also be understood by bringing a pragmatic criterion to light explicitly or implicitly. If we first consider a classically materialistic definition of this criterion, we can see its absence only generate chatter. The traditional cut from reality, in philosopher, seems to make him ignore that all what he say can be practical. And it can be observed linguists themselves have an unfortunate habit of being caught by formal games totally cut off from language practice. The research of pseudo-ambiguities is the best example. Nevertheless, in the Dreyfusian lineage of the cognitive function of the body, Searle reminds us very precisely this operative element which partially justifies the critique of classic philosophy which “has persistently emphasized knowing as more important than doing.” (Searle, p. 62).

The pragmatic criterion belongs to these authors’ philosophical references, what is generally the case in the Anglo-Saxon framework. If it is good to remind that Peirce, the founder of pragmatism, had himself created ironically a metaphysicians association, it is also necessary to emphasize that his solution, in reaction to the European philosophical context too much theoretical, does solve the problem arose by phenomenology, because abstracted concepts are scientific only as long they are operative, technical. That characterizes the science of this century. Theories which do not want to be reduced to such a pragmatic processing are ‘empty’ or ‘shallow’ theories that Weizenbaum was spoken about (pp.152-153).

This pragmatic criterion represents an answer to the Sapir-Whorfian relativists too (Dreyfus, Weizenbaum…) by providing a kind of dynamic framework, centered on realization rather than needs (like the functionalists). That allows correcting the culturalist tendency of always imagining too big cultural differences. This last phenomenon maybe resulting from a simple intellectualist bias consisting in characterizing a material or cultural element by its specific difference, forgetting its resemblances.

We could also analyze pragmatism as the real origin of the refusal of formalization or representation by phenomenology. Classic problems of knowledge as reality reflection can find a solution by an action on the world. We can better understand the phenomenological drifting of the Marxist partisans of a philosophy of praxis. They materialize dialectic darkness of their former philosophy denying the validity of representations (while often remaining into the stage of a demystifying ideologies reframing). We will notice too that, for certain intellectuals, the lack of operativeness rather results from their own jargon: “We can have a structure that reflects the history of our interactions in a medium, but that medium is not composed of “things” that are knowable.” (Winograd & Flores, p. 50). For it is sufficient using a more descriptive language to allow a first operationalization: “A mental experience is the reproduction in thought, not of reality, but of actions or operations focusing on it.” (Piaget, La psychologie de l’intelligence, p. 101). If we don’t throw the baby with the water of the bath, as Marxists used to say, nothing prevents of considering that language uses categories. As they allow us speaking about things we have not the direct experience. We can also enrich ourselves with others’ experiences, thanks to communication, and not to the Heideggerian-Marxist trick of negating representations while maintaining the avant-gardist intellectual role.

“We could not escape being immersed in a tradition, but with an adequate language we could orient ourselves differently and, perhaps, from the new perspective, generate a new tradition.” [NOTE 170].” (Winograd & Flores, p. 40).

What can really mean the term pragmatism overused by certain people? On this very point, AI can precisely be understood as an operationalization of philosophy by defining understanding in the framework of a Peircian operative theory? But Dreyfus, as we already have seen about understanding (pp. 12-14), and Weizenbaum (p. 157) do not accept its pragmatic association with a procedure, which is quite better than its etymology (like the comprehension one: “to take with”). Besides the fact Dreyfus always considers peoples as excellent when they agree with him, the excellent Peirce (who does with me) considers this kind of operative conditions as sufficient for any definition (his successors, James and Morris, put intention back). However, we can understand the philosophers’ anti-formalist confusion, because the pragmatic criterion does not need having a preliminary formal expression. It expresses a theorization before the term. But although inexplicit, this formalization by action is not less than real. This judicious Searle’s remark shows it very well:

“I am not saying that in order to have the institution of money people have to have that very word or some exact synonym in their vocabulary. Rather, they must have certain thoughts and attitudes about something in order that it counts as money and these thoughts and attitudes are part of the very definition of money.” (Searle, p. 78).

But, even from a pragmatic point of view, formalization of a new theory represents a different stage. It is necessary to record this reality to escape puzzles. In the absence of any recognition of formal statement validity, there is nothing else left as mysterious empathic capacities of comprehension. Weizenbaum can satirize (p. 158) against the atheist scientism calling “mystics” those who do not have them.

As we already have seen, the idea that human being is formalizable is not different on principle than the fact nature is too. Once again, the philosophical influence confuses the problem statement in the questioning form with the negation of a solution when discovered. Even if cognitivist formalizations are incomplete, they only claim behavior is formalizable. Otherwise, it seems that mystery must remain mysterious. Very classically then, ineffable is also a way of distorting pragmatism. It is sufficient to present as an impossibility what is, either suitable for no matter who, or the consequence of a training:

“No one who has not rowed can see an oar as truly an oar. The way someone who has never played one sees the violin simply not the same, by very far, as the way a violinist sees it.” (Weizenbaum, p. 18).

This idea, providing an acceptable common sense distinction, excludes any possibility of communication. The tradition of philosophical subject isolation into dereliction (anomie for sociologists) is found back with satisfaction, while it essentially characterizes the comprehension limitations reached by philosophers. The philosophical infinity put operativeness itself in an infinite framework (“for there to be a formal theory of pragmatics, one would have to have a theory of all human knowledge.” Dreyfus, p. 198). The refusal of pragmatism can finally indicate the resistance in front of its unavoidable consequence. Its acceptance would legitimate algorithmic principles that are its practical realization. If this implementation in data processing offends too much one’s convictions, it is enough to recognize an algorithm in every recipe (“You will need two eggs, a book of flour, a sachet of yeast, etc.”). But scientificity cannot look neither too simplistic, nor too digital, to please philosophers.

Actually, operationalization is a philosophical contribution which, by definition, does characterize computer science. We can find it in the classic procedural/declarative opposition. The procedural – algorithmic –, includes the introduction of temporality (a order relation) characterizing any theory of action. We can assimilate pragmatics to the procedural. By opposition, classic philosophy is rather equivalent to declarative programming, consisting in giving definitions. The paradox lies in the fact that classic programming is equivalent to modern phenomenological philosophy, while artificial intelligence would be equivalent to classic philosophy. A reconciliation could take place in object programming which is a mix of both types of programming.

It is however possible to accept the non-operative terms (like intuition). They characterize simply then, at best, some more or less ad hoc heuristic supports. More possibly, these terms only express inhibitory functions (slogans, confusions, paradoxes, etc.), which it is necessary to formalize or express too. They are the operationalization of non-operativeness. As action is not always possible, contrary to (Marxist or American) voluntarism.

The formal criterion can also amount to the technical capacity criterion, as formal calculability is often questioned for the sake of practical incalculability. For instance, in the case of chess, all moves are theoretically foreseeable, but impossible to calculate in a reasonable time. On this principle, when Dreyfus has written his book (1972, and 1979 for the second edition), a large number of problems could be reduced to technical limitation of contemporary computers. Several times, he especially insisted on the large number of data necessary to deal with in formalizing common knowledge or in problem solving in the case of GPS (Dreyfus, p. 94), and especially in the case of the natural language processing:

“The definition of eight hundred and fifty words comprise far more information than can be modeled in the core of today’s computers…” [NOTE 171] (Dreyfus, p. 145).

These failures do seemed, in 1967, caused by the machine capacities, while today the capacity and the quickness of hardware are infinitely superior (see Introduction). According to Dreyfus himself (pp. 217-218) although grounding his reservations on semantic aspects, the increase of means could allow to solve problems of access to data in databases.

Because of these technical limitations, in the seventies, and all the more so in the fifties, we can even notice that research workers in computer science, and AI, were rather obliged to spend their time in creating tricks to improve quickness of processing, or to lower memory occupancy. The only interest of these steps is that these result are always available. It is same for the facilitation provided by the new programming languages:

“Gelernter’s high school geometry theorem-proving program, which took several work years to design in the late fifties, could be done in a few days by a graduate student in the late sixties, thanks to a high-level programming language called PLANNER, designed by Carl Hewitt at M.I.T..” (Pamela McCorduck, Machines who think, note p. 193).

A current computer allows an improvement of research, not only in AI as such, but in many other sciences, especially human sciences. We could give an idea of the progress realized in the world of computer in representing it by the disposal of a miniaturized accelerator of particle for any student in physics. Currently, the access to such an equipment is often not available for all research workers. And its cost make them spend their time to find technical tricks for manipulations [NOTE 172].

In the practical history of AI research field, we can emphasize too the weakness of the mobilization, or the lacking of researches follow up reminded by Jacques Perriault (French “Preface,” in Dreyfus, p. XVII). This widely explains some provisional failures that AI adversaries consider as final. The poor number of early research workers, the pioneers’ difficulties, did not allow programs follow up. The lack of recruitment, partially caused by Dreyfus’ and Bar-Hillel’s criticizing, did not allow exceeding these initial difficulties. It would have been necessary to consider that the practical situation is far more extraordinary, and far more banal:

“We sometimes need to reminding that for all their comparative simplicity, nothing like the computer programs of the mid-fifties had ever existed in the world.” (Pamela McCorduck, Machines who think, p. 193, my italics).

Jacques Pitrat, in the discussion of the Dreyfus’ book French edition (p. 436), added that the bad quality of implementation can be explained too by the necessity of quickness in some applied researches. But a reason of some failures can be the controversy on AI itself too. The trivialization could foster research workers in AI to take the exact opposite, through researches on prestigious topics, but out of reach. Minsky (quoted in Dreyfus, p. 149) had already noticed this possibility. From this point of view, the mistake only consists in the too big initial ambitions. For instance, research workers should have better tried to automate first checkers, rather than chess game.

Critiques like philosophers’ ones could dispute the very principle of research in any areas this way. The slow AI advances reminded by Anthony G. Oettinger in the “Preface” of the Dreyfus’ French edition (p. XX) doesn’t enable estimating results, but refers to demands or hopes. In the first edition of his book, Dreyfus (pp. 1-2), emphasized that by the end of a decade, the optimistic predictions of AI partisans were far from being realized. But claiming that the outcomes are only general or trivial doesn’t concern only this research field. One can tell the same concerning research against cancer or AIDS for instance. Nevertheless, fundings have not been interrupted, neither disputed. It is the same for any applied research, whose fruitfulness can be estimated only with specific criteria:

“Any innovation has a high failure probability. In industry, one estimates that a new product out of four succeeds … profits have to compensate losses of the three others. In oil prospecting, the rate is one productive well out of ten ‘dry’ drillings. In pharmaceutical research a ratio of 1 out of 12 [… is found]..”(Edmond A. Lisle, “L’entreprise de recherche en sciences humaines et ses problèmes de gestion,” Social Sciences Information, 12 (1), p. 163 [“The research business in human sciences and its management problems”], my translation).

In the same way, the assumptions study should have been considered in term of fruitfulness as description (relevance), or applications (productivity). The principle amounts to the biblical criterion, beloved of Weizenbaum, according to which “the tree is known by its fruits” [NOTE 173]. One could possibly, according an empiricist principle, being interested only in outcomes, whatever assumptions are. And this outcomes are precisely the demonstration of the critical validity of AI, contrary to the philosophical criterion grounded most possibly on the principle Kantian according to which it is the thought that counts. But the philosophical tradition, by its depreciation of technique, as allowing to deny outcomes, only lefts critique of assumptions possible:

“I want to make absolutely clear from the outset that what I am criticizing is the implicit and explicit philosophical assumptions of Simon an Minsky and their co-workers, not their technical work.” (Dreyfus, p. 87).

This academic hypocrisy allows no matter what. It’s be rather necessary to control outcomes, because the reactive attitude of AI adversaries, in term of (formal) critique of assumptions, can essentially be reduced to the characterization of limitations, constraints, pragmatism, operationalization. It the more reasonable point of view can be admitted as calling for the responsibility of a definition of limitations (cf. Dreyfus, p. 79). But we already have seen that limitation can be defined into an applied researches standpoint. The real always reminds its constraints.

We can be inspired too by the fundamental medical principle of indications and contraindications. It is essential to resist to the holistic illusion of panacea for having discovered one active ingredient. Although in medical practice, this tendency is actually observed too, it derives that, if this definition of limitations is a good method constraint, it is illegitimate, at the same time, to demand research workers the opposite on the holistic mode.

What about philosophy anyway, as for its promises and outcomes. We already know it resulted in no longer giving answers, only in order to raise questions! As an ironic French May 1968 slogan said: “Enough acts, words!.” One could even notice, according to Heidegger, it has bluntly forgotten the meaning of Being during two thousand years. And after having claimed that millennial efforts of philosophy were useless, and that it was necessary to start all over again from the beginning, he enlisted himself to the Nazi party in 1933!

Obviously, the possibility of debating, correcting the dogmatism or ritualism limitations, has some perverse effects too. The recourse to anathemas (elsewhere as residues of the dogmatism) and fake-debates are not excluded. And they often appear in AI adversaries. The use of almost no matter what argument claiming to demonstrate anything one way or the other emerge sometimes, even to translators:

“When a writing instrument with a ball at the end was introduced, the end was nonetheless called a point (not a tip), and the instrument a ball-point pen (not a pencil), presumably because it was crucial to the users that the mark this instrument made could not be erased.” (Dreyfus, p. 124) Translators’ note: “This demonstration is attractive… but it is a lot less convincing in French, in which one can say as readily ‘ball-point pen’ as ‘ball-point pencil’.” (in Dreyfus’ French edition, p. 82, my translation).

The rationalizing delirium, seeking to justify the origin of notions, forgets here the arbitrary of the sign. But the university rhetoric is hardly any better, even in its lucidity. The understatement “a lot less convincing” means exactly a proof of the contrary. Is it reasonable, even as a formal clause, to talk about a “attractive demonstration,” who risks to slip the oxymoron to academic sycophancy (never contradict a colleague, or a client in the translators’ case)?

Winograd and Flores are right to insist on conversations, if we exclude their tendency of considering them as a bit too informal, or even as tending to telepathic communication. We can most possibly recognize here the romantic origin influence of phenomenology, linking communication only to cultural communion, in order to deny Enlightenment rationality. This explains this demagogic idea of common sense, and its contradictions.

Common sense is admissible all right, first as fundamental origin, by definition, and as dialogue of specialists with the lay person. Elsewhere, democracy precisely means that any individual opinion is receivable. Especially by the fact that the society is made of citizens having the right to speak, including in the scientific area. The resistance to this idea simply characterizes an archaic idea of science precisely comparing science to the argument from authority. Scientific practice sometimes still looks too much like the former physicians one. Medicine having only recently improved the dialogue with patients.

The specialist’s contempt for the lay person (intellectualism) is as ominous as the lay person’s one for the specialist (anti-intellectualism). Even when a criticism is not justified, or a question is badly stated, they constitute a problem to which it is necessary to answer. There is no such thing as a lay person’s non-conceptual world, but only a formalization incapacity – of which phenomenology has made, elsewhere, a virtue out of necessity. The problems met by the lay persons are universal problems. It is the reason why anybody has to face them. If these problems are already solved, the so-called specialist still needs to be able to produce an explanation. His responsibility would have to be to re-establish questions in their canonical form. This would suppose that he has this competence.

Elsewhere, for questions about “mental substances” or social ones, the corpus of answers is less systematic. It is therefore often possible that a naive person, or a literary work (and therefore a poetical one too, but of course not preferentially), provide correct answers or reinvent a former but more or less neglected one. And it is possible too a naive person can be more relevant than a scholar, like the maid Nicole, from the Molière’s Bourgeois gentilhomme, in the fencing lesson, who touched in tierce without passing through quarte. But on pain of falling in the Rousseauist archaism, an individual’s common sense, or the small boy from The Emperor’s New Clothes, ought to be under the same possibility of assessment than science. With no doubt, this other Rousseauist mistake motivates the scientists’ resistance! But it made them throw the baby research worker with the water of the vulgar thought bath. Or more seriously, to prefer a dogmatic to an empiricist, or a docile parrot to a naive and eclectic dilettante.

It is most possibly according to this principle that, more generally and explicitly, Dreyfus considers that the burden of proof isn’t on him concerning the nonexistence of “complete laws of behavior” (p. 193). But contrary to what he asserted, there is always an equality of burden of proof. And his process, consisting in stating a different hypothesis to believe to cancel the previous, is frankly puerile. Then, as it is rather about adopting a dogma – here holism –, at best, both hypotheses are equivalent. As much that even when some new researches deny a former hypothesis, this one would have simply been a legitimate state of knowledge, in the light of this former knowledge.

In the same way, the fact that a dogma can be sometimes and locally confirmed doesn’t remove his dogmatic quality similar to astrological predictions: if two magi make a prediction, one the sun and the other the rain, one of them is right and can boast of for his clientele, without possessing a scientific validity. It is elsewhere the only true reason of the Popperian thought against empiricism or pragmatism, as proof of validity. But the only case in which this is working is a binary alternative, as R. Boudon has shown in L’art de se persuader des idées douteuses, fragiles ou fausses [The Art of Self-Persuasion: The Social Explanation of False Beliefs].

Dreyfus, as for he, has a very philosophical way to record refutations. His way consists in relativizing, then to use this operation to dig his adversaries, who jus had score one point, in a bigger hole, like when the Greenblatt’s program Mac Hack defeated Dreyfus at chess (Dreyfus, pp. 84-85). One guesses “the glee” of the computer community, nonetheless human. But it is false to say like Dreyfus he had only done “a correct report of the state of the art at the time (1965)” (Dreyfus, note 45, p. 317). At the very most could he say he spoke of performances of the chess masters. But he repeat, without criticizing, the unambiguous remarks of one of these grand masters and he considers the next losses of the machine as normal “given the limitations of digital computers this is just what one would expect” (Dreyfus, p. 85). The progress recorded by Dreyfus in the second edition additions of its work ought to surprise him. And this even represents, this time, a refutation, not only of the phenomenological thesis, but also of the intrinsic limitations of the assumptions which motivate it.

Many years later, on august 31st 1994, an improved version of Chess Genius 2 ($ 200 software on a standard computer, with a 16 Mo. memory and a microprocessor Pentium 90 MHz), eliminated the world champion Kasparov in the first or second round of the INTEL Grand Prix in London. It was a blitz game (in limited time), in which the computer is favored, because it calculates 100.000 moves by second. Nevertheless, this eats into the myth of the absolute superiority, for ontological reasons, of human being’s intelligence on the machine’s one. Criticizing persons will be satisfied to notice the same computer has been defeated since. They generally add it is the proof of the superiority of the human brain on microprocessors. This superiority most possibly consists in the capacity to record one-sided evidences [NOTE 174].

About the substance of our debate, how to understand the critique of assumptions in a context of democratized science? If that means there is always assumptions, it simply concerns the rule of discussion, or wording, demanding an agreement on premises. Is there an assumptions analysis which is not a pedantic and labored implicit reminding (if common) on methodological pretexts? Subjective and intersubjective criteria have some limitations which can be overcome in the general framework of debate (with formal criteria), but without decisive priority of any thesis over another. The day can be carried by the recourse to objectivity (or empirical) criteria.

The democracy is grounded in managing the situation in which people do not agree. The archaism about it is the utopia of the perfect consensus and coming before debate. The reality of this utopia is realized in dictatorship. It is the by rights ignorance of the necessity of free debate, which creates the scandal of democratic practice. And it is the de facto ignorance, which creates the surprise, including among scientists, who then reproduce the dogmatic tradition and the argument from authority. The splitting of their market, concerning disciplinary specialization can explain both the opposition to AI, and the riposte. Nevertheless in democratic countries, it is obvious that in science like elsewhere, everybody is free to criticize any thesis. But then, this liberty of debate still seems to be a problem. As in its time, AI criticizing were very badly welcomed, as Dreyfus reminds it:

“My report is called ‘sinister,’ ‘dishonest,’ ‘hilariously funny,’ and an ‘incredible misrepresentation of history.’ […] Seymour Papert of M.I.T. […] ‘protest vehemently against Dreyfus with any good. To state that you can associate yourself with one of his conclusion is unprincipled’.” (Dreyfus, pp. 86-87).

The exaggerated reaction, pressures (quoted p. 307, note 6), besides the deplorable absence of being confronted with direct criticizing in the scientific fields, can however be explained by the emotion in front of a totally negativist discourse (not only negative). One can, with an academic diplomacy, admit that Dreyfus proposes some possible critiques, which one would possibly to take account of. But it is also true he proposes a so dogmatic philosophical standpoint that it becomes grotesque, because his negation of AI contributions or performance most often amount to misinterpretation on facts, or self-contradiction on his own demands.

This computer scientists’ reaction of defense most possibly happened because of the absence of dialogue, at this time, between them and philosophers. But the situation has changed since, precisely because or thanks to AI which, by speaking about human beings or consciousness, has transgressed disciplinary monopoly. In that, AI had realized interdisciplinarity, mythical until then. The Dreyfus’ fierce attack, like the reactions above, rather stems from corporatism then. This explanation appears more reasonable than the following psychologization:

“In fact, artificial intelligence is the least self-critical field on the scientific scene. There must be a reason why these intelligent men almost unanimously minimize or fail to recognize their difficulties, and continue dogmatically to assert their faith in progress. […] “ (Dreyfus, p. 151).

This cast of mind doesn’t concern only computer scientists (one guesses this accuser figure hasn’t a big interest). But the final statement seems to misuse the Popperian (or pre-Popperian rather) criterion, because, although Dreyfus dogmatically denies the validity of partial successes, he could understand that those considering them as important consider them as strengthening. But that would contravene to his own assumptions. Even Searle does justice to AI according this refutation criterion:

“As a philosopher, I like all these claims for a simple reason. Unlike most philosophical theses [sic], they are reasonably clear, and they admit of a simple and decisive refutation.” (Searle, p. 30).

We are not obliged either to follow blindly the Popperian method. It is true that the principle of verification to which it tackles is not sufficient, because it leads to dogmatism, if used all alone. It therefore simply suffices, to get anybody to agree, of admitting the real scientific practice studies as much conditions of truth of hypothesis than conditions of falseness. But it is also true Popper refuses the logical positivist approach, which he even believes having refuted (Unended Quest, p. 88). We can notice that is equivalent to the indications and contra-indication study in pharmacy, according to the epistemological model of medicine, more empirical than the physics one, beloved by Popper. A theory is rather defined by its limitations (which integrate the Popperian’s standpoint) than by the (Platonist) old formalist principle of demarcation.

We already have raised that the quotations used by the DREYFY technique invalidate the lack of self-criticism, when he annexes some AI community members’ ones. It is conversely necessary to notice that the self-criticisms principle is not the rule either on the philosophical planet. The Heidegger’s philosophical self-criticism, concerning his Nazi commitment is still expected! Finally, it is rather necessary to admit it concerns an extremely frequent practice in the university galaxy. As confirmed by the ingenuous considerations below:

“It’s impossible to deny whoever the right of criticism. In practice, however, this right has to be conquered. […] Ethically speaking, it is a tricky question – Does anyone criticize, especially in public, a colleague’s work?.” (Daniel Andler, “Preface,” in Dreyfus’s French edition, p. XII).

This amazing refusal of criticism is actually perfectly explainable by the scientific procedure itself. Scientific tribe practices are grounded on the monopoly of expression in each scientific journal. This monopoly naturally induces the apparition of concurrent journals. This is certainly a problem of sociology of science. But it is obvious that it modifies the content of papers, and the general validity of the publication criterion. The only excuse would be to consider them as free expression places, or big yes (group creativity without contradictions phase) specific to brainstorming, only for one sect. But this would remove a lot to their scientific pretension (this term even seems synonymous of absence of debate!).

Like Max Weber seeing in the Protestantism one of the origins of capitalism, thanks to the selection by the environment of the entrepreneurial behavior, one could see in the scientific publication system an image of the protestant sectarisation, by opposition to catholic centralization. The problem of this method, especially in human sciences or philosophy, in which experiment is missing, is that the intersubjective criterion of validity is solely grounded on the ability to find disciples. What difference with sects, theologically speaking [NOTE 175]?

This entire debate around AI could however be reduced to what was called “La lutte pour la vie dans la cité scientifique [The struggle for the life inside the scientific city]” (according to the title of a Gérard Lemaine and Benjamin Mathalon’s paper). This selection of programs or interesting idea obviously never ends. But the fact that a criticism, in the current science, could be perceived as an aggression does raises a problem, as ethical as epistemological. If there is no debate, in accordance with the professional rigor, the permanent selection of disciples amount to collusion, as it often can be observed in reciprocal quotations. The fraud would be consubstantial to the scientific practice then, because the international evaluation is made on these quotations [NOTE 176].

We can actually admit that research workers need, again and always, much money (Weizenbaum, p. 254). But who can pretend that research has to be stopped? The paradigm of medicine, up to its driftings, indicates that it is necessary to control allocations. Any contribution to a research path can delay another more fertile one, but it will often be known only a posteriori. In real time, it can be considered that the arguments belong to defending one’s own cause or to career strategies. At best, we can witness a sharing out, with more or less connivance and insincerity, implicit in some very diplomatic statements:

“Many linguists, for example, Noam Chomsky, believe that enough thinking about language remains to be done to occupy them usefully for yet a little while, and that any effort to convert their present theories into computer models would, if attempted by the people best qualified, be a diversion from their main task.” (Weizenbaum, p. 184).

In the AI history, the linguists’ action on automatic translation budgets happened to be very real. French Dreyfus’ prefacer, Jacques Perriault, very precisely reminds us (p. XVI) the role of Bar-Hillel’s work about the undecidability of grammars in the brutal stop of automatic translation governmental programs of research. This justifies the anxieties of some research workers in AI, although respecting the academic rhetoric of excessive politeness (Jacques Pitrat, “Discussion,” in Dreyfus’ French edition, p. 434), because we can hardly see who else than Dreyfus could be the representative of these “fierce adversaries [of our discipline] who take advantage of any critic papers to prevent us from doing the experiments we want to implement” (Jacques Pitrat, “Discussion,” in Dreyfus’ French edition, p. 434). Indeed, it seems this is the phenomenology’s aiming, or “mind’s intent,” to deny the validity of any other discourse. On this very point, it has adopted the philosophical tradition pretension of ruling the other fields. In practice, it amounts in allocating budgets! And this would not be coherent financing condemned researches, as Dreyfus very explicitly asserts:

“Now, before we invest more time and money on the information-processing level, we should ask whether the protocols of human subjects and the programs so far produced suggest that computer language is appropriate for analyzing human behavior.” (Dreyfus, p. 303).

Because finally, if it was the comparison with human beings which had shocked these authors, it would have been possible to continue researches under the information-processing heading, without focusing on the term intelligence. But as we know now there is no way either of analyzing human behavior even in human sciences, this seems to demonstrate that the AI opponents’ discourse doesn’t only concern the practical validity of these researches. The opposition is complete as soon as some symbolic taboos are broken.

[NOTE 150] Minsky, Semantic Information Processing.

[NOTE 151] Winston and the team of the laboratory of IA of the M.I.T., p. 48, note 24 of chapter 3.

[NOTE 152] A pure formalism, in its literary version could consider the whole intellectual production as already contained in the combinatory of letters and typographic signs corresponding to texts as its result. We find a good presenting of this delirium, inspired by medieval speculations, in the Borges’ short story, The library of Babel, in which each book is one of this combinatory of letters. But then it is obvious that the maximal extension of this combinatory (for a book of half a million signs, with forty relevant signs = 40^500.000 variants) makes difficult any idea of interpreting program. But every introduction of a supplementary constraint limits possibilities too. With a simple lexical criterion the amount is falling quickly, because the number of words is not infinite in a given time. The possibles are limited too with syntactic criteria. Fictions aside, we end by arriving to empirical limitations, i.e. material or observational situations. But obviously, it’d better begin by the last step.

[NOTE 153] P. Suppes, “Meaning and Uses of Models,” in B. H. Kazemier and D. Vuysje, eds., The concept and the Role of the Model in Mathematics and Natural and Social Sciences (N. Y., Gordon and Breach, 1961), p. 172.

[NOTE 154] Formalism, like Piaget indicates, is a way to escape to habits ascendancy, instead of becoming a delirium in which everything is possible.

[NOTE 155] Unless the authors wisely consider the number of them they made change their mind since page 18.

[NOTE 157] This Thesaurus have been translated and adapted only in 1991, for French people … amateur of alphabetical order.

[NOTE 158] Philosophical or mathematical, since we already have seen the formalistic alliance on this topic (see the paragraph Intuition or method).

[NOTE 159] Empirical news, on September 17, 1994, have confirmed the inductive generalization, by the death, at 92, of epistemologist Karl Popper.

[NOTE 160] E. Bergler, The psychology of Gambling, (New York: Hill and Wang, 1957), p. 230.

[NOTE 161] It is however possible that this tendency to resist to fate is objectivable too. It would characterize the subjective state of the protagonist in situation who: i) either, he needs motivations more or less rational in a state of uncertainty. An intuition actually being equivalent to an unconscious observation of real signs (non-existent in games of chance); ii) or he acts in agreement with one’s internal state, characterized by these illusory signs (“to be lucky” for the player) or real ones, which can be, for instance, a state of vigilance.

[NOTE 162] This topic is a permanent argument of fictions, especially Anglo-Saxon one. In the hyper-formalistic local context of the USA, one of the preferential topics is the possibility of communication to others of one’s own experiences. What characterizes again fiction as a highlight of unsolved problems.

[NOTE 163] One can even say that all the authority of natural sciences or mathematics comes first from the importance of the common part. A reciprocal acknowledgement of competences derives from it. Not that debates or quarrels opposing these scientists aren’t harder than in human sciences. But this recognition … possibly reluctantly …, represents the academic validity criterion by consensus (on which insist some epistemologists like Thomas S. Kuhn or Isabelle Stengers). While human sciences, even if their phrases are validated by common references, seem to think as if to reinvent the cheesewire (French idiom) at any time. In these fields, this lack of stable core curriculum can be the cause of the recurrent myth of the Copernican revolution, or the epistemological cut.

[NOTE 164] Maturana, Biology of language, p. 50.

[NOTE 165] Uncertainty, in case of conflict of authority, appeared also in the famous Milgram’s studies. The subject had to punish the human Guinea pigs (simulators) with gradual electrical shocks, if they gave bad answers to questions (raised/stated). When these that gave instructions were not unanimous, the subject tended to be released the obedience.

[NOTE 166] Brazil, the movie according to Orwell’s 1984, has shown a good illustration of what could be a bureaucracy and an ideology of a world without computer.

[NOTE 167] Minsky, Semantic Information Processing, p.27.

[NOTE 168] Besides trying to be modest, the Weizenbaum’s book is grounded on the automatic triggering of a moral judgement program, sincerely using insincere old rhetorical contrivance, the end justifies the means. We can observes these defense mechanisms occur when science or technique produce a new situation which seems to contradict values. And this resistance has occurred for the very scientist producing this new reality.

[NOTE 169] From 1994, only one CD-ROM contains 600 Mo (Mega-bytes), i.e. 600 big books of 1.000.000 characters. Techniques of compression allow to increase this density fourfold. And it happens to be too 10 discs drive (at least). It is therefore possible to access directly 25.000 books, on a personal computer, in a space as big as an element of hi-fi system. And the next DVD-ROM generation has already (in 1998) a twentyfold capacity. Finally of course, the current Internet availability allows to access the whole world databases.

[NOTE 170] Maturana et Varela, Autopoiesis et cognition, 1980, p. 17.

[NOTE 171] Quillian, “Semantic Memory,” in Semantic Information Processing, p. 241.

[NOTE 172] Elsewhere, it happens to be, at the beginning of 1994, the American project of a giant particle accelerator SSC (Super Superconducting Collider), in Texas, were abandoned for budgetary causes. Would it means therefore that research in theoretical physics is impossible?

[NOTE 173] Conversely, what is observed in the academic area, in philosophy or formal sciences, is rather a measure in term of quantity of papers.

[NOTE 174] In the French first edition, I wrote: “I obviously record the fact that Kasparov has defeated back the machine in the beginning of 1996. But he no more claims it will not ever be superior to him, as he formerly did. Today, He only thinks being able to contain it until the year 2000!"

The big event happened on 1997 with his game Vs deeper blue. Of course, as non-ontological thinking, only one victory is not really meaningful (if it could be anyway). And the rules should be modified to allow a human to rest more (albeit it is a machine quality too). But what ontological rationalizations if Kasparov had won?

[NOTE 175] The problem of dogmatism in the real religious framework arises due to the fact that the corpus of reference is considered as nearly intangible (Protestantism, Islam). Furthermore, the specific difference of science is constituted by the idea of confrontation to observation (possibly similar to the Orthodox tradition limiting the interpretation of the text by comparing with nature). One can consider that the catholic system has at least the advantage of comparing the point of views, and even to maintain some control of hierarchy on interpretative qualifications. A catholic science seems to me actually more legitimate than a protestant one: i.e. a universal and developing science is preferable to a dogmatic and sectarian one … although the practice of labels owners could be the opposite of etymologies.

[NOTE 176] This evaluation is of course debatable due to the notorious mutually profitable chumminess reigning in these circles.