Philosophie contre intelligence artificielle

A kind of unhealthy atmosphere currently happens in the epistemological discourse whose philosophical archaisms can be the cause. Paradoxically, some philosophers see science only in mathematics, or dissections. The philosophers’ pseudo-scientific paralogisms demonstrate their fetishism about formal reasoning they challenge in the same time. While their own discourse rather pertains of archaistic tautologies: 17th century ideas with 20th century words. It can be thought as an outlook of science from the literary point of view, which handles symbolic items without understanding them. Or which absolutely wants to enjoy the credibility, or rather the prestige, of sciences that these writers disparage, while giving themselves up to the most absurd reductionism.

But the consequence of such a reduction would be acceptable under a strict condition: philosophy has to be reduced to neurosciences. And actually, it is the current standpoint of some research workers to consider problems concerning consciousness as only valid at the neuronal process level. It is elsewhere possible that they are not wrong in large measure, following the reduction possibility of philosophy in its components. But that doesn’t mean that neurosciences are the only component. Anyway, particular sciences simply make meaningless the philosophy label.

For the philosophical point of view, and the layman’s one, most the time, science has a terrorist aspect. It is true that epistemological classics references to (especially quantum) physics, or to mathematics, often pertains to imposture or pedantry. This phenomenon has been emphasized by Searle, who derives a conclusion oriented to its prejudices:

“’Science’ has become something of an honorific term […]. A good rule of thumb to keep in mind is that anything that calls itself ‘science’ possibly isn’t for example, Christian science, or military science, and possibly even cognitive science or social science. […] The rival picture I want to suggest is this: what we are all aiming at in intellectual disciplines is knowledge and understanding.” (Searle, p. 11).

Even if the premise is unacceptable, the good dispositions of the conclusion can only be applaud. After all, an extremely general definition would satisfy anybody, and would for good avoid the eternal false issues. Specifically, the most narrowly specialized research has always a bits and pieces aspect before being great intimidating and perfect systems. But Searle keeps his good intentions only on the very short distance of a paragraph:

“Some disciplines are more systematic than others, and we might want to reserve the word ‘science’ for them.” (Searle, p. 11).

Everything has to be done again! Escaping the gravitation of former doctrines is difficult. Such a more exclusive, demarcationist, idea, requires asking the classic question of knowing when physics, for instance, has become a science. Or conversely, of knowing when unscientific residues have disappeared. This topic, everybody knows, is a (quantitatively) important part of the epistemological production. And the authentic answer is: “Never, of course!” If the mythical, pre-scientist, ideological, social, etc. aspects are considered by turns, a former period has always a remainder to eliminate. Such a definition is essentially retroactive or anachronistic, and is grounded on the infantile attempt for differentiation from predecessors. When science simply makes progresses by two ways: cumulative contribution and old assets reform too, and it will always be so.

It is unlikely that any scientist has ever raised the science definition question before doing it, contrary to the classic philosophical problematic. Even Dreyfus adopts this approach (for natural sciences), most possibly to benefit the science honorary usage by association, while he challenges formalization in human sciences. On the contrary, a more trite definition of science can make us accept the results acquired in AI: what is feasible can be done first, what is automatable can be automated. The approach is a more empirical science. The correct paradigm [NOTE 43] of science or research is rather most possibly medicine, at least because anybody understands what is medical research. Thus, this weaker definition allows anybody to estimate a science in progress.

This trivialization, by identifying rationality to discursivity (or to calculability which derives from), eliminates the possibility of paradoxes or anti-rationalist paralogisms. The idea of science as reasoning in a broad sense is actually largely accepted (cf. Weizenbaum, pp. 14-15), even if this term can be associated to possible negative value judgements. This approach of scientificity in term of discourse analysis seems strengthened by sociologists works: about the scientific discourse in Latour & Woolgar’s La vie de laboratoire, La production de faits scientifiques [Laboratory life: The Construction of Scientific Facts], or even about the rationality of nonscientific discourse, in Raymond Boudon’s L’art de se persuader des idées douteuses, fragiles ou fausses [The Art of Self-Persuasion: The Social Explanation of False Beliefs]. However, it is not indifferent that science is only a more elaborate form of argument processes. These processes allow managing human conflicts, when demonstration ways are available. The general method consists therefore in considering critiques as constraints to take in account of.

Weizenbaum is right to consider a famous anecdote as a critique of science, because the possible shortcomings of the scientific method can also be responsible of a pathetic blindness:

“There is a well-known joke that may help clarify this point […] ‘Why, if you lost the keys over here, are you looking for them under the streetlights?’ The drunk answers, ‘because the light is so much better here’.” (Weizenbaum, p. 127).

Indeed, the scientific principle consists in analyzing a phenomenon with available and already known ways, i.e. these that are already thrown light on. It is what could called the lamppost principle. But then it simply concerns a perversion of the academic principle, whose formula would rather be:

i) Not to look only what is clarified, measurable, but to look if it is measurable.

ii) If measurement is impossible, discuss and make compromises, and everybody can measure what it yields to the other.

iii) Explain a phenomenon by using to full advantage already known principles, what allows distinguishing science (or its teaching) from research.

In the very previous passage, Weizenbaum (pp. 121-126) came to criticize the recourse to magic. It should have therefore been glad of this occult principles or ineffable avoidance. This is precisely about such occult principles resort that the academic principle simply asks to try first to explain phenomena with known principles. But it is correct that this method, by definition, has the limitation of making more difficult the discovery of a new explanatory principle. It can even postpone or forbid this discovery. The best example is precisely the Newton one, who integrated explanatory principles formerly belonging to the occult sciences field against the Cartesian Mechanism [NOTE 44].

In a sense, the phenomenological standpoint can be explained in reaction to the previous discourse the ‘epistemological break’ and other ‘breaks from common sense’. Psychologist Skinner is a providential example, speaking about “disastrous results of common sense in the management of the human behavior” (Weizenbaum, p. 245). But it was an enough widespread opinion in the research of this time, and Weizenbaum has not gone to a lot of trouble in busting the Forrester’s systemic jargon (idem, p 246). But the evocation of thoughts of “Plato, Spinoza, Hume, Mill, Gandhi and so many others” (Weizenbaum, p. 246) against Forrester is very clumsy. The argument from authority appears to be a paradoxical form of the illustration of the validity of common sense, because traditionally, philosophers opposed at it, at least as much to poetical reason, to which Weizenbaum refers like a panacea, in the Heidegger’s following.

It would be elsewhere possible to admit that the Forrester’s systemic method support ineffable, due the human being lack of self-understanding hindsight. It is the explicit opinion Winograd & Flores: “We are always within the situation, and to throw light on it is a task that is never entirely completed.” (p. 29). The alternate Forrester’s quotation meaning is of course that old ways are insufficient. But it is/was an extremely widely spread philosophical opinion too. And elsewhere, the phenomenology founders and follower jargon looks a lot like the systems analysis’ one in its pretension of breaking with common sense.

It has been observed too that many philosophers criticize common sense by tackling its hidden premises. Although, the difference between science and common sense is not in the fact of referring to causality (cf. Searle, p. 71). Generally, the AI adversary’s arguments are contradictory on this point, especially because AI is characterized by the attempt of modeling common knowledge, precisely on the same level than algorithmic and formalized knowledge. The whole debate is therefore in the skepticism, more or less dogmatic, of philosophical trends like phenomenology, concerning the “incredibly difficult” task (Dreyfus, p. 3) of common knowledge representation.

This perpetually oriented thought seems to forget that physics is also usually considered as arduous. In both cases, the problem first consists in finding operative definition of abysmal questions. And, without the rigor of formalization, it is easy to claim to be the common sense interpreter. It is possible too that the difficulty in conceiving a common sense modeling is only about the reasoning organization (the rationalist philosophers’ long causal chains). Rationalist practice precisely opposes to the metaphorical or symbolic thought of poets (and advertising or propagandists) who adopt shock formulae, images, slogans. It is this last idea which justifies an overall perception, ineffable, which resists with disgust to laborious demonstrations (these that expert systems try to mechanize [NOTE 45])! The reasoning sequence is the real meaning of artificial or natural intelligence, which seems perceived by the popular thought. But this alternative is disqualified. Weizenbaum want to wipe the slate clean of both scientific and popular categories, in their “too simplistic notion of intelligence” (p. 203). He regresses then to the former framework of religious or irrationalist models he is fond of. The regression consists in the fact that science is precisely equivalent to the enigmas and incompatibilities resolution of the previous mythical systems.

Anyway, the study of common sense has to begin with the methodical collection of the learned knowledge, and forms of observable reasonings, including false or incomplete ones. (Conversely, the logic theoretical approach is normative or corrective, in the moralistic tradition of philosophy.) This collection of common expertises embodies the continuation of the encyclopedic tradition, of which Dreyfus seems to question the project opposed to corporate secrets, to monopolies [NOTE 46]. And computer can be the very mean reminded by Leibniz to realize this collection and formalization. Once again, the explicit reference to this formalization of the “most important observations and turns o skill in all sort of trades and professions” (Dreyfus, p. 70) is originally available for the philosopher. But the backward-looking blindness remains, most possibly to help this practice retain this “confused trashing around” (idem) look Plato thought they had. It is possible that it concerns simply a former style pregnancy, due to academic conditioning by this period. This can explain the ontology/poetry topic, the return to “meaning of Being,” etc.

The formalization of existent practices problem is explicitly raised too by Dreyfus about medical expert systems. The same lack of understanding can be observed while criteria are manually introduced in MYCIN, most famous program of its kind. Dreyfus claims that: “The knowledge engineers have wisely not even tried to reduce this intuitive aspect of the diagnosis to heuristics rules” (Dreyfus, p. 310, note 71). But the specialist’s opinion about the severity of illnesses is not grounded on intuition. Estimate is grounded on observation rules, elsewhere easily formulable: first, obviously, the patient’s suffering or the importance of lesions; and others can easily be imagined, like the number of cases, or deaths, or the cost of treatment, personnel or collective, etc. More, direct interactions with the patient are not forbidden. An expert system could thus prevent from human mistakes. That would also avoid myths or illusions, sometimes grounded on the best intentions: in order to minimize, to comfort, or to reassure oneself, like in the following example:

“During a long week, nevertheless, Nicolas complained of his stomach. A scanner finally revealed its plexus were being compressed by a big sphere. A pain felt as dreadful for adult. Immediately, physicians delivered him from his hell. Then, instructive experience, they asked him to estimate, on a short ruler graduated from 1 to 10, the intensity of his pain. Nicolas estimated it at 8, real schema in its case. But, submitted to the same test, the nurse that followed him and cherished him for a long time, made a mistake: she indicated only 3!” (“La douleur des enfants [The children’s pain],” L’express, 15/3/85, p. 62).

Let’s note too that, today already, a simple databases does represent – by definition – a knowledge processing system. The improvement of database searching languages like SQL (Structured Query Language), or document handling, can already allow an access to knowledge in relatively natural language. The result validation criterion of the access and structuring of knowledge is therefore always definable in term of continuous progress. Knowledge is always what is already there and not what is missing. The main interest of artificial intelligence is here the not encoded expertise elicitation, since others are available, or can be made directly available in databases.

On the growing of knowledge topic, it is obvious that it is enough to record a new statement. But a recognition capacity of its novelty is also necessary, like in the philosophical necessity of AI acknowledgement. And the knowledge general principle remains experience accumulation. Now computer methods can be represented as the outcome and the unity of observation techniques in human organizations, as well as a formalism theorization (logic, language theory, and calculability). The paradox of the AI adversaries’ standpoint is they refuse alternately both approaches, by lying sometimes on one, sometimes on the other.

In the cumulative schema, the knowledge problem could first be expressed very simply by the fact that elements of knowledge exist already, and it is first enough to collect them. And precisely, the artificial intelligence project is often a simple available knowledge elicitation. Weizenbaum (pp. 237-238) precisely expresses strong regrets about a computerization of a geographical documentation, in which case it was decided not to consider former data. It is necessary to notice that this situation concerns the normal written production too or human experience in general. But this is precisely what AI aimed avoiding, contrary to the traditional computerization, which does the most urgent first for cost reasons. Natural language processing aims the elicitation of former or not formalized expertise. This former-texts elicitation remains a constraint of the AI contractual specifications. Otherwise, it is necessary to admit a total opacity of any otherness in time or in space.

Contrary to the decadence thesis, it is not necessary to discuss the fact that there is actually more available knowledge. Whatever the considered period is, the current science hasn’t ceased to accumulate knowledge. And children of today know by simple transmission what was ignored in the past. This cumulative point of view is what defines human condition. Intelligence as innovation is marginal. The possible heresy would be here the resurgence of the quasi-ontological myth of the decadence that was the norm until the 18th century (for instance, it was believed that the world – or national – population was decreasing). The negation of cumulative idea can lie in the excessive generalization of real, but incomplete, or unchecked data. Nevertheless, it is possible, punctually, to lose some knowledge that remain theoretically available:

i) Rare illnesses. Physicians of developed countries neglect sometimes to relate observed symptoms to illnesses rare or absent in countries provided with medical technology. That becomes a professional fault when travels allow easily contamination. Medical expert systems could at least be used as memorandum or security. They can also be used as tools of epidemiology vigil, if results are centralized.

ii) Household competence. The end of the domestic roles strict sexual division has obviously cause/induced the loss of specialization benefits. The culinary preparation industrialization has the same effect. Here also, a recipe can be reduced to the algorithm definition. The problem is not the innateness claimed to justify this roles definition, but in the popularization that seems to depreciate specialties and mysteries going with.

iii) Modern techniques mastery. Products or industrial machines, whose instruction for use or maintenance is unknown of their users, constitute certainly a current challenge to the cognitive environment mastery. Education or child curiosity allows the familiarity, but the risk lies in the quickness of changes. It is now obvious that nobody can claim to possess a general environment mastery, which can be available only on instructions for use easily accessible. The usefulness of the multimedia computers generalization is obvious then.

Let’s note, by opposition at cumulative theory, the libraries practice consisting in regularly purifying the stock, by the operation called weeding. This operation can seem in contradiction with the library mission that could be to preserve the whole. But the number of books published each year already made this a difficult task in 1902 (according to the President of Harvard, C. W. Eliot). Some explicit selection criteria formalize this operation:

“H. A. Bert, in “One weeding” (1942) […] identifies seven selection criteria: 1) material state. 2) incorrect (or exceeded) information. 3) to replace by a new and best edition. 4) to replace by another recent book. 5) outdated matter. 6) no longer adequate to the public (modification of the library public). 7) ordering mistake.” (BPI, Dossier technique n° 5, Le désherbage: élimination et renouvellement des collections en bibliothèque [Technical File n° 5, Weeding: Elimination and Renewal of Libraries Collections], p. 37).

This weeding operation takes place too in personal libraries or knowledge. And it is easy to imagine at least potential weeding: new knowledge eliminates gradually the oldest or the most mediocre ones. More generally, in the image of the world built by a person, or a group, the fact that formalism is implicit, or simply oral, emphasizes again this phenomenon, because an oral knowledge has its own memorization methods (see McLuhan). Paradoxically therefore, it is computerization that will allow a full memorization of intellectual works. Facilities provided by the new storage capacities can allow it.

The claiming of cumulative possibility doesn’t exclude the necessity to avoid repetition. This non-repetition topic, dear – among others – to Karl Popper, is obviously justified by the elementary logical principle (a & a = a). But the anthropological, ontogenic, and literary reality, or history of sciences itself – through quarrels of precedence –, admit reinventions. Concerning the propositions, the second part of the previous logical one demands the fact in question to be recorded, because in general the second statement has first to undergo regular transformations to be reduced to the identity. Concerning people, a rediscovering isn’t so silly or naive that it could be concluded from logical calculation:

i) On the one hand, it can actually concern a discovery of a new ways to obtain the same result. The comparison of these ways can possibly reveal a lesser cost. This is not indifferent in the human activity in general and in the industrial one especially.

ii) On the other hand, a rediscovery represents the best part of the cognitive learning. A permanent knowledge repetition would not prepare to autonomy. It is reasonable to think that the capacity of research doesn’t have to be grounded on the only repetition, even if it is eminently necessary, on the communication between human beings account.

In this connection, it is easy to remind the episode of the famous early AI research naivete: the discovery of a new proof of a mathematical theorem announced by W. R. Ashby for the Gelernter’s program (see the paragraph Referencialism). Then, what was significant for the computer is that this demonstration had not been programmed, what would constitute a rediscovery by a human being. This philosophical refusal of rediscovery grounded on the referencialist bias explains the rejection of enthusiasm or the lack of acknowledgement of intellectual work in progress for research workers (or of learning for youths)

The ignorant or specialized philosophical discourse develops, no matter what it can be said, a kind of anti-scientific leitmotiv which seems to do instead of thinking, or act as guarantee of it! The backward-looking attitude more or less bucolic or ecological is present too, possibly lying on actually documented researches (especially on Mumford’s works), but often distorted to the very root by archaic prejudices: the clock replacing the sun or the crowing of the cock “for signals for sleep and rising, and so on” (Weizenbaum, p. 25). Usually this attitude is an error of analysis. The fact to be regulated by the sun, the cock, the clock, or hunger, are conscious or social controls (more than technical ones). The opposition between natural and cultural doesn’t challenge in anything the fact that individual depends on this constraints since its birth and has to adapt to them. On the contrary, a scientific and abstracted encoding can allow the correction of an aberrant cultural norm, which itself is simply a stage of our knowledge of the world. The backward-looking attitude is obscurantist by the refusal of bringing a natural phenomenon to light and of updating a behavior supposed to adapt to it. And it is contradictory too in its evocation of natural to describe a cultural phenomenon. This backward-looking attitude can be challenged too when it appears explicitly in a so absolute manner:

“It is important to realize that this newly created reality [by modern science] was and remains an impoverished version of the older one, for it rests on a rejection of those direct experiences that formed the basis for, and indeed constituted, the old reality.” (Weizenbaum, p. 25).

Like in the case of the John Dewey’s “[putting of] some portion of an apparently stable world in peril [by every thinker]” (idem, p. 26), such an imbalance is created, on the contrary, to increase rationality, i.e. stability, or to solve a problem. This Rousseauist standpoint was wrongly supposing that the world is stable. Actually, it would rather be the research of a new balance that disturbs, if it can be said, an old disorder. More, when rational means precisely have replaced irrational ones, it is ridiculous to assert that peoples’ problems come from their too big rationality (Weizenbaum, p. 221)! A real problem isn’t solved with less science. The only reservation of its best application still demands a more important one. Like most of the anti-scientific discourses, it seems that the problem lies in the differentiation of the current reason from the potential one.

The difficulty of the question of science today is simply in the fact that the world is on principle more foreseeable than it was only one or two centuries ago. Contrary to what claim its opponents, science represents the only current knowledge means. It is no longer about to suppose a common reference which is not scientific, or in competition with science. Common sense is merely equivalent to weak-science, under an over-lenient or relativist form.

When Polanyi dares to assimilate science to magic due to fact that “Any contradiction between a particular scientific notion and the facts of experience will be explained by other scientific notions.” (in Weizenbaum, p. 126), he doesn’t contradict science but actually justify a kind of monism, or an encompassing cultural framework, an epistemological end of history. There is no other reason that Reason.

This scientific monism cause/induces however the Weizenbaum’s resistance (p. 16). But the statement of the too big authority of science is very little relevant. On the one hand, because the philosophical or popular resistance to science is well known; on the other hand, because the humanity’s majority is still deprived of scientific prospect; finally, because the reason is truly the only stance in which discourse can criticize itself, contrary to the traditionalist dogmatism. But our author judges correctly the AI specialists’ intentions, maybe the simple naives enthusiastic ones, as a call to rationality, but he still judges this reason as only instrumental conversely to the “authentic human rationality” (Weizenbaum, p. 253). Indeed, AI represents a call to reason, and the top of western philosophy. And reason is instrumental because there is nothing like non-instrumental rationality – a reason which fails is useless and harmful [NOTE 47], the more if it doesn’t record refutations (in it and not of it). Weizenbaum cannot have any more justifications to refuse problem solving than Lysenkoism to refuse genetics. Partisans of reason have to choose the reason side.

In addition, and this is not only an easy irony: what would be the legitimacy of the Weizenbaum’s speech without science and, especially in his case, without technology? Has this kind of elusive discourse playing on all counts any value? On the contrary, it can be analyzed as the proof that there is not any other knowledge than scientific ones.

The knowledge general principle consists, what else can anybody say, to increase the exact propositions about the world. The Berkeley’s immaterialist false solution has already been contradicted by the French psychologist Zazzo in Le devenir de l’intelligence [Intelligence Becoming] who solves for good Heideggerian resistances, denounced from the very origin, in 1947:

“Beyond real, what is found, is real again – richer –, beyond reason, it is again reason – larger and deeper.” (in Cuvillier, Les courants irrationalistes dans la philosophie contemporaine [Irrationalist Trends in Current Philosophy], p. 80)

Actually, it cannot be excluded that knowledge produced by phenomenology could be recycled. But that will ask a patience and a capacity of vigilance against drifting that only a machine will be able to show.

Whether the reference is science or reasoning, an aspect of the critique addressed to artificial intelligence concerns the lack of rigor of its partisans’ discourse. Dreyfus repeats the Chomsky’s reproach about “a natural but unfortunate tendency to ‘extrapolate’ from the thimbleful of knowledge that has been attained in careful experimental work and rigorous data processing, to issues of much wider significance and of great social concern” (in Dreyfus, p. 79), whose himself has made of it a specialty. But it can be admitted that scientific papers would profitably also explain their limit. The beginnings of AI have had their ups and downs, for instance in the results of chess programs. Dreyfus (pp. 83, 96) does not fail to emphasize the sudden discretion of research workers in a difficult position or disappointed.

But it has to be noticed that this critique could as well aim philosophers for all kind of topics. The critique itself would suppose that a refutation criterion is used rather than a provisional neurobiologist skepticism. Now then, a quality of AI, or even of data processing in general, is precisely in allowing to validate the theories relevance by measuring performances, even by non-expert end users. It remains that the best critique, in appearance, consists in showing cases in which predictions have been denied, like Herbert Simon’s ones in 1957, concerning GPS (Dreyfus, p. 81).

But the rigor demand requires too that predictions be not simply postponed. Dreyfus notes himself that the third prediction about computer psychological modeling “has been partially fulfilled” (p. 164). But both others (about chess or mathematical theorem demonstration) are not obviously impossible to satisfy with a bigger extension, especially because of the effective progresses observed since the sixties. It is possible too that some fields that Dreyfus neglected at that time like cybernetics, considered as sterile (Dreyfus, p. 130), were only dozy. The result of these researches is currently neuronal systems – which can elsewhere be used too against the so-called traditional AI, by research workers of these new fields.

In general, it can be concluded that AI extrapolations are not inevitably baseless. First, on principle, a research worker extrapolation is grounded on likely hypotheses. In philosophy, historically, extrapolations are rather called utopias. The empirical argument doesn’t hold either, because observable successes can be opposed to actually observed AI extrapolations failures. But especially, extrapolations are justified by the fact they do not propose qualitative leaps, contrary to philosophers’ statements, because computerization already model human behaviors or intelligence. To admit that, it is enough not to depreciate the most elementary skills

To deny extrapolations, Dreyfus stigmatizes too what Bar-Hillel calls the “fallacy of the successful first steps” (cf. Dreyfus, p 129). It is about claiming that possible failures are only provisional, or that the “thimbleful of knowledge that has been attained in careful experimental work” (idem, p. 79), of which Chomsky was speaking about, can only grow. And it is a fact that this argument represents a declarations (and not a results) analysis of AI protagonists, in most of the area mentioned in its brief history.

However, the rejection of this standpoint (considering that “these [narrow] programs are steps toward ways to handle knowledge” [NOTE 48], in Dreyfus, p. 146), is exclusively grounded on the holistic prejudice. Actually, it is pretty obvious that any simple program allows handling knowledge, and that its very principle allows to increase the knowledge of this handling. It is very obvious that the sophism is in the holistic side which seem to refuse any breakdown, and any cumulative approach: first steps are obviously necessary for any (new) field. Like for a child, the sophism would consist in claiming that it is necessary to learn how to run before to know how to walk. Or as could again say a Zen aphorism: a last step before a first one. The problem is here the unacknowledgement of partial results, and the negation of their compounding capacity, on the pretext of the “wholistic” myth (Dreyfus, pp. 53-54), because the role, and the practice, of research is not to propose an overall solution. In this analytic refusal, the more general resistance to science and to its applications can be perceived. But in medical research, for instance, it seems that research workers – and sick persons – are satisfied of these specific progresses from experimental method, often enough purely and simply systematic. The panacea research, how satisfactory for mind can be its outlook, is not conceivable as research program.

Phenomenology being inspired by Gestaltism has adopted the flash understanding idea, and innate actualization one. It can easily be understood that progresses are never the sign of the final success (cf. Bar-Hillel in Dreyfus, p. 147) from this point of view. But, if it happens to be something like the myth of the first step, it is even more difficult to do the next one without doing the very first. It would be actually easier to begin with the second. But in science, it is often two different persons who do these two operations, what can precisely explain the rejection or the notorious lack of acknowledgement of which founders are the victims.

A good example of this first steps notion tendentious excess is precisely given by Dreyfus when he reminds ironically the prehistory of the spatial conquest:

“Feigenbaum and Feldman […] define progress very carefully as ‘displacement toward the ultimate goal.’ [NOTE 49] According to this definition, the first man to climb a tree could claim tangible progress toward reaching the moon. Rather then climbing blindly, it is better to look where one is going.” (Dreyfus, p. 100).

To what Jacques Pitrat answers in the afterword of the French edition, taking the trouble to consider his adversary’s argument with patience:

“But this man was a genius! The first thing to do was of course to climb in a tree. That hasn’t yielded anything and it has been necessary to try something else. But it would have shown a very complicated mind if he had decided from the beginning that it was necessary to build a three-stage rocket. […] A great deal of works are indeed criticizable, but […] we can criticize them only because they have been done.” (Jacques Pitrat, “Discussion,” in Dreyfus’ French edition, p. 435).

The example in question is about the experimental method too, which is opposed to the ruling pretension of philosophical introspection one. Weizenbaum, who emphasizes the risks of the first one for programmers, forgets to warn about the dogmatic risk of the second one. Is it necessary therefore to believe, according to the Gestalt theory, that our venerable ancestor should have already had an intuition, or have known everything from time immemorial, like philosophers who inform future? The myth of the first step opposite is equivalent to the myth of philosophical hypothesis that would be supposed to enlighten research, because in any research, there is inevitably a first step, sometimes to a dead-end, that some number of which is the fate of true researches.

The prehistoric man has therefore been able to verify that he did not reach the moon. This justifies the classic principle according to which a waking idiot (or naive) is always going further than a siting philosopher, except if the philosopher can decree rules which forbid the movement. Such judgements focused on scientific progress are easy a posteriori. It seems that this is a professional philosopher bias too.

Obviously, the prospective can also represent a deceptive advertising, like in some immoderate declarations bordering on swindling. But it especially concerns marketing problems of software companies, or news sensationalism, like in both examples quoted by Dreyfus (pp. 79-80). These legal characterizations can also concern scientific research in its struggle for credits. A literary illustration of this criminal practice (concerning precisely data processing and… automatic generation of political clichés!), existed, as soon as 1964, in the excellent Robert Escarpit’s novel, Le littératron! But it would be a tendentious comprehension to only see it as a critique of AI, rather than a more general critique of scientific or political habits.

It would maybe be necessary to describe this swindling characterization as the contingent expression of a myth birth for public at large, like in the case of the Newell, Shaw and Simon’s chess machine (Dreyfus, p. 83), because, contrary to its condemnation, the human way to appropriate a new phenomenon, consists mainly in linking it to other traditional myths (here: puppets, robots, machine men, superior intelligences,…), or by amalgamating a technical knowledge to fiction. It can also concern a simple fashion. Weizenbaum (pp. 27 and 33), emphasizes this computerization tendency to follow fashion, with the business or scientific projects over-equipment, without automation necessity, for tasks that could have be undertaken manually at lower cost.

These critiques against extrapolations, possibly legitimate, do not have to inhibit the reciprocity of rigor demands, because this AI opponent doesn’t deny himself extrapolations, sometimes obviously like below (but elsewhere often slier):

“Recent work in neurophysiology has suggested new mechanisms which might confirm the Gestaltist’s intuition […]. While still nothing definite is known about how the brain ‘processes information,’ […] while model based on the properties of optical hologram look perhaps more promising.” (Dreyfus, p. 20).

“Neurophysiology offers, admittedly speculative, accounts of such similarity, but these are holographic not information processing models.” (Dreyfus, p. 25). etc.

We can observe that this kind of vague statements weakens the argument they are supposed to sustain, especially when someone put forward this very principle to others. In general, discourse devices are useful to develop hypothetical negations or to be grounded on technological abracadabras like holography, without the lesser evidence, except the first step principle, grounded on indices – what seems to forget the overall vision principle. The Searle’s argument can be easily opposed to this last holographic abracadabra:

“The temptation is always to think that the solution to our problems must wait on some as yet uncreated technological wonder.” (Searle, p. 30).

And when confessions appear sometimes, euphemistically it is true, they do not go until to admit neither the consequences, nor the opposite references:

“Of course, it is still possible that the Gestaltists went too far in trying to assimilate thought to the same sort of concrete, holistic, processes they found necessary to account for perception.” (Dreyfus, p. 20).

This “possibility” is indeed considered, for a long time, as a certainty in the psychological circle. Thus, Piaget devotes pages 68-75 of his book, The Psychology of Intelligence, to the already old critiques of Gestalt psychology. He draws attention to generalization excess about these totality structurings “to perception, motricity and elementary functions, reasoning itself and especially syllogism” (Piaget, p. 61, my translation). Dreyfus who uses this very Piagetian’s work as a reference to strengthen his opinion, is very careful not to reproduce his reservations, present a little everywhere in the book. The overall reading magically becomes very selective! Dreyfus, in its confusion of references, sometimes subjectivist, sometimes holistic, selects obviously the ones reinforcing him. This is precisely the very ideological approach, as biased as partial (always according to the Lazarsfeld’s principle). The contradiction is double when it is about to justify holism:

“Piaget […] has come to a Gestaltist conclusion: ‘ […] A single operation could not be an operation because the peculiarity of operations is that they form systems. Here we may well protest against logical atomism… a grievous hindrance to the psychology of thought.” (Dreyfus, p. 253).

Although, individual exists, even if he has been fathered, and is part of a long chain. Followers of the body – which is always individual – could be coherent on this point. If Piaget adopts partially the overall Gestaltist point of view, and of course the body operative idea, he distinguishes a formalization stage too. And especially, he stands in general in a dynamic model of learning, in which he criticizes the Gestaltist fixism. He picks out the mistakes about the constants postulated by the Gestalt theory (cf. Piaget, La psychologie de l’intelligence, pp. 70, 114, 121). More, it is strongly possible that this holistic framework is itself the consequent of this philosophical context influence, neo-Hegelian or Marxist, on Piaget. He however holds himself aloof in a perfectly explicit way:

“Analysis which always consists in translating facts into total “field” is the only legitimate way, an atomistic reduction always twisting the unity of real. [¼] But [¼] the language of totality is only a mode of description, and the existence of overall structures demands an explication which isn’t included in the totality itself.” (Piaget, op. cit., p. 69).

More again, the overall notion is precisely made operational in the systemic approach questioned by AI opponents. If the reference to a totality has to be considered, as the hologram image indicates, while preserving the advantage of analysis, it is not necessary to see in the operation a magical solution grounded on intuition. The best image is more simply, or more traditionally, the reconstitution of a whole skeleton from only one bone. The very reference to totality is grounded on the accumulated knowledge.

To the first step principle (which anticipates on results), this trivialization principle can be opposed. The trivialization method is quite simple. When AI adversaries admit all the same some tangible results, they are minimized with condescension, or characterized by their limitations. This condemns them by definition, according to the holistic principle: ‘Everything which is not all is nothing’. This method is even applied to Zadeh, the father of fuzzy logic, although used to solve the binary problems of traditional logic (Dreyfus, pp. 322-323, note 80) and to express undetermined quantities in natural language. But it was precisely this kind of inability: “’between 3 and 5 miles away’ or ‘somewhere near the orbiter’” (Winograd & Flores, p. 85) AI was generally criticized for. On the same pages 126-127 (about Zadeh) of the Dreyfus’ book, a Wittgenstein’s labored gloss is supposed to solve many crucial philosophical problems (like the resemblance of same family members!). Even if, according to Dreyfus himself, Wittgenstein “continues the above remarks rather ambiguously” (p. 126). But in the philosophical context, fuzziness can be allowed! That does look like a submission to hackneyed phrases.

We see the trivialization principle proceeds by minimization (often contradictory) because the “The very fact that such achievement deserve to be applauded itself testifies to how utterly primitive is our current knowledge about the human mind.” (Weizenbaum, p. 198). Despite this shallowness, knowledge about human mind make progress by experiences of this kind. And to stop these works won’t make them progress. Especially when there is a tendency to deny the very progress of knowledge possibility (what is empirically absurd), to claim an overall comprehension? The all‑or‑nothing option (Dreyfus, pp. 8-9) can only apply a part of the knowledge of the world. Actually, it would be most possibly necessary to make validate these partial results by philosophers. This can produce some howler, significant of these verbal or cognitive mannerisms:

“Perhaps this is a good place to say that though I am not optimistic about the overall research project of cognitivism, […] I certainly do not want to discourage anyone from trying to prove me wrong.” (Searle, p. 53).

If someone can prove that Searle is wrong, I do not see why he would need his permission (nevertheless, I thank him for his warm support). This is where leads dogmatism in its mentalist tendency, a bit too much grounded on the fantasy of controlling others.

Finally, it is the fable of the tortoise and the hare (to please Weizenbaum, who likes children tales), which is shown in the first steps refusal, like in the case of the critique of micro-worlds: “This belief in local success and gradual generalization.” (Dreyfus, p. 124), because confidence is obviously established by local successes, in the same way the trouble has most possibly begun to take a hold in holists by their repeated failures. As for their generalization idea (except the fact that no one can see what else than local stuff could be generalized), is most possibly grounded on the normative prejudice of philosophers, who would want to inform it by overall and innate ideas.

This trivialization is found in Weizenbaum, and in Searle too, refusing any scientificity to AI. Searle comes to disparage programs universality only because “from a mathematical point of view, anything whatever can be described […] as instantiating or implementing a computer program” (Searle, p. 36). His confusion rests on the fact that triteness doesn’t indicate falseness, but obviousness. And obviousness represents a logical proof that simply makes reference to the shortness in the chain of deductions. In the specific case of an encoded language definition as to represent phenomena, the simplicity (or the formal mathematics banality) is not a legitimate reproach. When required conditions are satisfied, it is contradictory to say that such a language is trivial. Elsewhere, while the simplicity is usually, in Searle, a kind of true idea characteristic, it becomes here a sign of triteness, in the same way that in Dreyfus, for the case of banking investment programs (pp. 293-294).

Finally, this trivialization principle is grounded on an epistemological professional bias. When history of sciences is invoked almost solely, any popularized knowledge become vulgar and common sense. Any possessed knowledge becomes epistemologically obsolete. It can be thought it concerns a confusion that assimilates science to research alone. This confusion benefits from the complicity of research workers, most possibly because they overvalue their own status. Here, philosophers find an associated source of prestige, in the literary valorization tradition through the nobility of the addressed topic, what doesn’t prevent them to invoke common sense when it suits them!

More reasonably, on the contrary, it can be admitted that simple knowledge can ask a long working out, or a more or less laborious acquisition. Even if it is possible that new technologies facilitate life a bit too much for still noticing it. For instance, the necessity of learning how to read hour on a dial, which was a real rite of passage (at the time one’s were offered a watch), is lessened today with digital watches. This ridicules incidentally the humanistic pretensions of analogue systems. So, formalization difficulties can shrewdly remind us these realities.

In the rhetorical minimizing framework, it can be observed that philosophers do not deprive any more of the systematic usage of the quotation marks, especially to express the disapproval or the distance, without having to justify them:

“We must be careful to speak and think of ‘information processing’ in quotation marks when referring to human beings.” (Dreyfus, p. 166).

This fashionable expression: ‘in quotation marks,’ has invaded interpersonal communications, radios, television, movies, newspapers, books [NOTE 50]. Orally, a pause before the brought out expression, replacing the dramatic or quotation marks, can be noticed too. A movement mimicking quotation marks can even mix gesture and words: behavior is affected too.

To pronounce ‘in quotation marks’ characterizes a contamination of oral by written good example. But it seems sometimes that the quotation marks learned usage is ignorant of typographic conventions. Normally, the quotation marks usage concerns quotations (or titles in scientific papers). The confusion concerns maybe underlined words in manuscripts, which classically means italics for typographers. This learned usage ignores the pedagogical or simply communicative necessity too. Indeed, the introduction of a new concept belongs to the category of emphasized words. If the author only targets specialists, the question doesn’t change: either the word (or the meaning) is new and it has to be explained, or the word belongs to the admitted jargon, and a distinction is obviously useless.

Another usage characterizes the current tendency to reproduce the protagonists’ discourses. It does seem indeed that the use and the abuse of quotation marks consists in noticing the different jargon, different dialects mixing, sometimes to dissociate, sometimes for slumming in it. This allows playing on several stylistic registers by holding oneself aloof from popular expressions, which come naturally to mind. Our two jargons amateurs obviously use and misuse of this quotationization facility: “We need to be cautious about words like ‘think’ (even when set off in quotation marks).” (Winograd & Flores, p. 127). Cautiousness should not be concentrating only on words!

Corollary of trivialization, a rhetorical norm of trying to be modest can play some role in scientific or philosophical statements. It can easily be seen at work in lectures, prefaces, even at Nobel Price and other prizes awarding. But this stylistic norm seems established as new epistemological criterion:

“Ross Quillian […is…] the most modest […he…] has made no sweeping promises or claims. This program confirms a general evaluation heuristic already apparent in Samuel’s modesty and success and Simon’s and Gelernter’s claims and setbacks, namely that the value of a program is often inversely proportional to its programmer’s promises and publicity.” (Dreyfus, p. 142).

Of course, the fact in question can simply concern journalistic sensationalism. But most often, the minimization simply indicates an affirmation of a norm, or a conformity demand, in explicitly moral terms. This norm can especially be transformed into an inhibition principle in the common philosophical discourse:

“Philosophers from Plato to Husserl, who uncovered all these problems and more, have carried on serious epistemological research in this area for two thousand years without notable success.” (Dreyfus, p. 36).

Besides the fact that, with a Dreyfus in 500 BC, these researches would not have even begun, the philosophers’ failure doesn’t gives them any particular legitimacy, and especially not the capacity of anticipating the failure of others. On the contrary, it could be supposed this failure could come from the antique negation of technique and from the contempt of naive engineers. These last ones are depositary, on principle, besides their professional expertise, at least of common knowledge, which AI aims to formalize them and to make them operative.

This is the difference between AI and current philosophy, which has abdicated in front of the necessary work for knowledge formalization. This philosophy is pleased in various forms of ineffable invocations, and in religious prohibition of transgressing dogmas that derive from it. This does define an inhibitory function in communication. And this severe control doesn’t exclude contradiction in minimizing the acknowledgement of modesty of those allegedly immodest (Dreyfus, pp. 24, 55) and all of those used by the DREYFY self-critical principle. It can finally be wondered if this modesty as norm cannot be considered as a principle of self-constraint for those who promise little to be little committed. Elsewhere, in order to appear modest, it will be enough to emphasize the number of difficulties, following the other trite principle, “A conquest without peril, is a triumph without glory,” what is another advertising swindling style.

It also can be asked whether the Dreyfusian rejection of the self-congratulating or optimistic AI discourse cannot possibly be limited to the claiming of a cultural or social norm – relative in this case. For instance, the gap between the supposed southern peoples’ exuberance, and the northern (of Europe) coldness reference, adopted as appearance of academic seriousness. But the cliché, even if there is some truth, allows personal differences, and should not represent a stylistic norm in the scientific community.

Then, one may wonders if the so-called modern philosophy is not simply an authoritative demand of a rhetorical use of a tragic, even apocalyptic, style [NOTE 51], which is shown especially in the usual progress or technique refusal. The application of this sifting to data processing, as representative of this technoscience, and all the more so to AI, as ultimate of this triumphant technology, would be therefore an automatism!

This transgression of the trying to be modest taboo reigning in conferences and scientific papers can thus be qualified as it is known that phoniness sometimes and paraphrase often are reigning too. Then, it can be wondered again if this modesty is not a bragging. Modesty is elsewhere inevitably relative to personal qualities, and to real results (themselves being relative). The unbearable gap of those finding something justifies ingenuous enthusiasm. And after all, Salvador Dali is a greater artist than André Breton, the surrealist pope, is.

It is necessary to point out too that this tragic style is opposed not only to progress, but to personal success too. This incidentally allows a recycling of the radical or Marxist leftovers. Their anti-Americanism or anti-conservatism is obviously grounded on the negation of the results measure. As contradictions do not frighten philosophical tradition, phenomenology simultaneously recycles a Christian miserabilism, although stigmatized by Nietzsche.

In this connection, we can notice that the enthusiasm refusal is strange for those, furthermore, supposed to be poetry apologists. To the Dreyfus’ irony accumulation, his stooge Weizenbaum adds contempt for some of his too much enthusiastic students [NOTE 52]:

“A student’s quick climb from a state of complete ignorance about computers to what appears to be a mastery of programming, but is in reality only a very minor plateau, may leave him with a euphoric sense of achievement and a conviction that he has discovered his true calling.” (Weizenbaum, p. 277).

It is true that “programming is almost immediately rewarding” (idem), what could rather be a quality (which reason is precisely the cognitive nature of computer studies). As for a vocation discovery, it seems to be usually grounded on applications of any jobs. The pure research inclination is marginal. It concerns some people surviving academic selection, life circumstances, and academic program aberration (which is not likely to be immediately rewarding). But it is false to assert that programming gives an omnipotence impression, thanks to numerous programming mistakes (bugs) on which ‘compulsive programmers’ (Weizenbaum, pp. 115-121) have toiled whole nights.

Let’s notice too, incidentally, a differentiation criterion between humor and joy. The success and jubilation criterion can precisely provide the key. The hypothesis can be advanced that humor consists in laughing at oneself or at others about failure behaviors; while joy conversely consists in feeling and showing the pleasure of a success situation [NOTE 53]. By opposition, the philosophical tradition, when it still meant something like in the Spinoza’s time, seemed to wonder about joy. Then, the means-end approach seems to represent an answer, which demands obviously the measurement of performances. It is therefore easy, but deceitful, to satirize.

This bad master contempt – for the pupil or computer learning – has to be opposed to the computer scientist’s enthusiasm for his realizations. If Weizenbaum mentions his own emotion before his sleeping child:

“When my children were still little, my wife and I would stand over them as they lay sleeping in their beds. We spoke to each other in silence, rehearsing a scene as old as mankind itself.” (Weizenbaum, p. 201).

Isn’t this emotion applied to the same child who learns, or is the professor blasé in advance? Apparently, because this contempt (at least as discourse) for enthusiastic computer science students is generalized more or less to any fields, and to most students:

“When such students have completed their studies, they are rather like people who have somehow become eloquent in some foreign language, but who, when they attempt to write something in that language, find they have literally nothing of their own to say.” (Weizenbaum, p. 278).

Well, this phenomenon is present elsewhere! And it is met too in the case of everyone own language! Indeed, there is a difference between competence and performance that can be characterized for the beginner as the obsessive concern to first produce correct statements. Everybody can guess that it happens especially with suspicious educators. These ones can therefore be the cause of the lack of originality, and their belief to be strengthened! This judgement does seem to focus more on pedagogical system than on data processing. And it is notorious that professors do not estimate to be responsible of the educative system results, what is maybe the thing to revise.

But it cannot be reproached to students, even goods ones, not to satisfy, during their school years, to the most elaborate research criteria. Students, especially mathematical sciences ones, either cannot be asked to master conclusive arguments to questions still in debate, like computer understanding of natural language, naively justified by its applications in medicine (Weizenbaum, pp. 270-271). In such a case, it concerns an old chestnut answer for people who have not really thought. As most human beings can’t see further than the end of their nose, they generally stop to the first standard justification. Medicine is an excellent one (as it is the very scientific research paradigm). On this question, the problem is therefore not technique, but technicians, because it is not necessary to ask technicians to be (inevitably) theorists. And this is not been theorist to produce such an old chestnut answer. But it is not to be a theoretician any more not to understand this principle, neither not to identify their limited competence.

How to explain the contempt for DP men or computer scientists solving a problem, and considering a methodological extension of this resolution to problem resolution? This contempt is more pathetic than the trivial object in question:

“Newell and Simon sensed the importance of the moment [computers solving such problem as how to get three cannibals and three missionaries across a river…] and jubilantly announced that the era of intelligent machine was at hand.” (Dreyfus, p. 77).

This overall standpoint should have to satisfy (amateur) philosophers, if traditional anti-empiricism did not demand to see (here) the real as principles embodiment. They could have then see abstraction work from vile trivial experience, supposedly beloved of phenomenology. Against a paradoxical rejection of this simplicity, what could say a philosopher in front of an elementary operation as 1 + 1 = 2? “This is not so easy. I am in front of a contradictory situation of uniqueness confronted to another uniqueness. How to perceive a composed unity. etc..” While real calculation practice is grounded on existence of empirical tables allowing complex calculations. A theory of arithmetic actually supposes preliminary existence of these operations. It can be wondered what could satisfy professors for whom any understanding would be judged on the criterion of the genius (as the Pascal of legend who, deprived of books by its father, reinvented mathematics). As much that, unfortunately, our author refuses too the calibrating of performances, as indicated by the classic attacks of the non-scientificity of IQ tests, which would have “so thoroughly muddled the thinking of both scientists and the general public […]” (Weizenbaum, p. 203). Contrary to the usage of these tests in the USA and in the countries they influence, the French experience, with an opposed choice, has caused the same dissatisfaction. In the French system, it cannot even “determine what ‘success’ people may achieve in later life” (idem) especially professional on rational criteria rather than strictly school criteria (out academic careers, if that). The success reminded by Weizenbaum is limited to grandes écoles [literally ‘High Schools’ but rather specialized high education establishment], grounded on a very selective principle opposed to the French University general system [NOTE 54].

Even in the artificial intelligence debate, the human intelligence monopoly seems to be associated to the refusal of its evaluation, in the lineage of the egalitarianist ideal, or the classic relativist or Third-Worldist leftist principles, applied on the alleged ethnocentrism of IQ tests (Weizenbaum, pp. 203-206). This principle is itself intrinsically contradictory, because it denies any right to western culture to decree any rules defining “the very stuff of human worth” for itself (idem). While it is pretty obvious that, everywhere, everyone is judged more or less intelligent (or possessing any other human quality) in function of a local criterion. Thus, Weizenbaum judges the AI partisans to be absurd, pretentious, and certainly not so intelligent, or anyway not demonstrating his own values.

Anyway, this AI rejection fantasy is denied by the epistemological or cognitive value of the practice of plain data processing. Its adversaries even incidentally admit it. Possibly, AI can be considered to a minimum as a technical extension of intelligence (Dreyfus, p. 301). And nothing can justify the idea that it can be otherwise in the future, neither that it is possible to deprive voluntarily oneself of its assistance. But AI can precisely allow, by its own characteristics, self-improvement.

Talking about computer intelligence raises the problem of knowing whether AI is only a metaphor or not. This term often comes in critics of AI (Weizenbaum, p. 176), who does not seem to admit either computer scientists’ intelligence:

“It follows that the claim that programs like SHRDLU have a little bit of understanding is at best metaphorical and at most outright misleading.” (Dreyfus, p. 8).

This kind of negative remark could characterize as well animals’ intelligence. Nevertheless, La Fontaine’s beavers one (see above) was inferred from observation of their behavior – precisely consisting in displacing trunks, like the SHRDLU program, moving blocks too!

It seems rather than the term metaphor represents, in the scientific environment, a derogatory version of the notion of paradigm or model. Some AI community members thus believe as more prudent or more correct to avoid using the term artificial intelligence, they sometimes consider as controversial, advertising, or clumsy. Others believe this AI term rather be replaced by a paraphrase in jargon: “the execution of computer systems inferential capacities” (Mario Borillo, in Dreyfus’ French edition, p. 423). But the AI idea is precisely an analogy/paradigm allowing identifying human and machine information processing by designating it, analogically therefore, by the human origin of the operation. Indeed, the metaphor is an abstraction process like any other, then allowing considering other common features than scientific and technical ones. It would have indeed been possible to find a new word. But the negation of what would have no longer been an analogy would have been nevertheless possible to be produced by Dreyfus (by the utilization of devaluing analogies).

In the same way, it is not coherent with philosophical referencialism to be surprise by the strengthening by AI of the mechanistic hypothesis (cf. Weizenbaum, pp. 8-9), because in the past, automaton had raised the same debates, from the greatest names in philosophy. It is therefore easy to think that philosophers in question would have used this AI metaphor too.

In front of the AI partisans metaphors attacked by the (one-sided) purism control, those of the practical imaginative human beings certainly aren’t perfect too, as one of our authors notices himself:

“It’s hard, when one sees a particularly offensive television commercial, to imagine that adult human beings sometime and somewhere sat around a table and decided to construct exactly that commercial and to have it broadcast hundreds of time. But that is what happens. These things are not products of anonymous forces. They are the products of groups of men who have agreed among themselves that this pollution of the consciousness of the people serves their purpose.” (Weizenbaum, p. 273).

No squeeze out! Well, there isn’t any anonymous action of the Overall Mentalist Holy Spirit. Things like this are man-made as machines are. As we can see, human beings (and not machines) do what they are able to do. Quality is not always as expected, but it is relative, by definition. More, ad-people are in general literary persons, or artists, obsessed by imaginary work, myths, and especially metaphors. They are even explicitly against logical demonstrations, judged heavily didactic, and elsewhere not very favorable to purchase motivation!

Dreyfus objects to the human behavior modeling the paradigm/example principle. He justifies it by the Kuhn epistemological authority, whom he corrects – rightfully – the Minsky (and usual) interpretation of his main notion, not as “an abstract explicit descriptive scheme utilizing formal features, but rather [as] a shared concrete case” (Dreyfus, p. 39). It is indeed important to note too, in the same passage, the fact that the living scientific thought isn’t characterized by the formal stage of completed science. And contrary to a deductive, logicist principle, frequent in history of sciences, the very paradigm term does mean example (according to its etymology). While Kuhn explicitly refers to grammatical examples as those of verbs chosen to illustrate conjugations (‘aimer’ [to like/to love] is the paradigm of the first group verbs conjugation in French). More, an example has actually both typical and pedagogical virtues more “cognitively economical” (Dreyfus, p. 24). But why computer and human being should be opposed on this point? AI is practicing typicality too, precisely answering to this definition of a representative specimen of a class. And it could be possible to generate it automatically by frequency analysis.

In detective stories older than Dreyfus’ theses, the Agatha Christie’s character, Miss Marple, precisely uses her village life experience to solve criminal enigmas. Which embodies, in this case, both an acknowledgement of the common sense expertise, and an example of its universalization. We can call this the Marple’s principle. But obviously, a practical case used as sifting, acquires an abstract value, as when a name becomes a common word (a Shylock, to Xerox ®, etc.). Reciprocally, an abstract model can remain a particular case, when a parrot, or a pedant, or the vulgar scientist, doesn’t know how to apply it outside the canonical origin example. This usually allows criticizing, often with reason, the textbook cases, technocrats, or unreal abstraction i.e. precisely when it isn’t one!

The metaphorical or scientific abstraction whole problem lies in the justification of models used to describe reality. Conversely, culturalism (the Sapir-Whorf’s principle) raises particularism to the status of incommensurability, and this to dogma, like current psychoanalysis, with its refusal of any supra-individual generalization. Solipsism is at the end.

The Dreyfus’ paradigms apology should warn him against his suspicion concerning computer scientists’ micro-worlds. The contradiction lies, as the case may be, sometimes in denying their validity (Dreyfus, p. 119, cf. here in Micro-Worlds), sometimes in claiming their representativeness – guess what, according to the fact they contradict or the strengthen the AI adversaries thesis:

“Chess is an ideal micro-word […]. But while the game’s circumscribed character […] there is a great deal of evidence that human beings play chess quite differently from computers.” (Dreyfus, p. 29).

It can be seen right here, besides extrapolated indices, that a micro-world is used as negative criterion. The champions phenomenological intuition myth prevent from raising the problem, nevertheless obvious, of the relevance of the chess game paradigm, because the true problem of any model is relevance, and chess game is not the best intelligence criterion, as judiciously noticed by partisans of AI:

“These young scientists were explicit in their faith that if you could penetrate to the essence of great chess playing, you would have penetrated to the core of human intellectual behavior. No use to say from here that somebody should have paid attention to all the brilliant chess players who are otherwise not exceptional, or all the brilliant people who play mediocre chess. […] Intellectually speaking, a good chess player is nothing more and nothing less than a good chess player.” (Edward Feigenbaum, Pamela McCorduck, The Fifth Generation, pp. 58 and 44).

The fact that computers play, and win today, by any means, actually hasn’t any importance. It is notorious that the chess grand masters do not play as other human beings do, without that this human being quality is removed to amateurs. Elsewhere, before the computer arrival in this area, the general impression was that the grand masters were not normal human being, and this was obviously not in their favor. More, all specialty pushed to its limit is usually producing results not only valued for its performances, but run down for its monstrosity too. Weizenbaum emphasizes the idiotic specialist (fach idiot) blinkers. In France, child prodigies are often called performing monkey too. And the Weizenbaum’s rejection of IQ tests also belongs to rejection of the exceptional, or egg heads one.

All the difference between good and bad metaphors seems to be grounded on the confusion between these philosophers’ infinite gloss ability, and the modeling ability as explanation. The traditional gloss tries all the combinatory of possible variants, whose result is the symbolist, i.e. pre-metaphysics, explanatory system. A systematization would be available in Gilbert Durant’s classifications (in Les structures anthropologiques de l’imaginaire [The anthropological imaginary structures]), which allow understanding how symbolism can be reduced to an anthropomorphic modeling too. But the cultural coherence of symbolism built by Durant isn’t inevitably equivalent to the amateur symbolists’ real discourses, which grope at best, and are short of this systematicity. Thus, it can be observed that, in contrast with “The mental component is an intention” (Searle, p. 63), the term memory system is explanatory in the theory of action framework. The term intentionality is anthropomorphic all right, and contrary to their pretensions, the philosophers’ ‘parables’ are contented by more than vague reasoning:

“Suppose a group of researchers said, ‘We will understand how clocks work if we design a machine that is functionally the equivalent of a clock, that keeps time just as well as a clock.’ So they designed an hourglass and claimed: ‘Now we understand how clocks work’.” (Searle, p. 56).

It seems here that a mockery of model to ridicule the adversary is considered as a reductio ad absurdum. But let’s analyze generously this banter: i) First, if a problem of this kind is raised, it would have rather been met in the historical reverse order. This reduced this model exactly to a humorous contrivance! ii) More, the example is clumsy for an intentionalist, due to archetypal role of mechanism played by the clock. iii) Finally, if an hourglass was as precise as a clock, it would be actually the model of one of its functions: the passing of time. We would be therefore logically grounded to conclude that the analogy correspondence is demonstrated, contrary to the self-satisfied tirades (Searle, p. 56).

The philosophical option can again consist in using its so much absurd examples than the intellectual putsch allows making forget that they are precisely self-contradictory!

“Now, we can describe the water as if it were doing information‑processing. […] But it doesn’t follow from that that there is anything of psychological relevance about water running downhill.” (Searle, p. 50).

In order to make this example valid, it would be precisely necessary here to be about a memory system, on the computers principle. And to fulfill the conditions, water would have to be an organism and an autonomous system: i) Memory System: because if the water could memorize information, it could actually allow to build a representation of the world it go across, as a weather report recording allows to make forecast. ii) Organism, because it is obvious that the water is not the river – animistic idea – each molecule is potentially independent. iii) Autonomous System, because the river or water molecules, are obeying physical laws without any choice, while living beings have to control their movements in the environment, what implies elsewhere a memory system, and an organism.

The rejection of the mere modeling possibility is linked, for AI adversaries to the behaviors simulation one. It can be wondered elsewhere if the very refusal principle doesn’t consist in challenging in advance any criterion of success of practical realizations. This is shown in wordings (Dreyfus, pp. 72-73) questioning the Turing’s test grounded on the capability to differentiate program and human answers. This refusal is especially argued by the doubt against the analogue system representation by a digital system one (Dreyfus, p. 327, note 11). A simpler mentalism is even limited to the petitio principii (Searle, p. 37). It is then necessary to note that the body simulation is formally established in a simple mechanism consisting in testing armchairs by making sit a form simulating the body only concerned part. It is the same for any isolated intellectual operation. Even if, obviously, this kind of modeling can always be trivialized:

“If simulation is taken in its weakest possible sense, a device is simulated by any program which realizes the same input/output function. […] This clearly lacks what is necessary for a psychological theory […].” (Dreyfus, p. 168).

As it can be assessed that it concerns the very definition of simulation, it is especially obvious that if it hasn’t anything to see with psychology, the negation of its own qualities has a great deal to see with insincerity, or lack of understanding. Especially when asserting the importance of the body, valorizing then only psychology represents a regression to philosophers’ intellectualist biases. If the problem is about psychological processes definition, it is not so obvious that modeling, or features reproduction, would not teach us anything psychologically speaking:

i) It can be considered that brain is defined by a engraved structure, therefore simulated by connectionism today. This option seems to fulfill Dreyfus’ duplicative conditions, while denying its predictions about perceptrons of the AI first time, whose connectionism is the continuation. But it can be estimated too that this duplication is useless.

ii) Or it can be considered that computerized formalization is about reasoning. Usually, intelligence is obviously defined by evaluation of such performances. It is therefore currently confused with coherent reasoning, and human or natural sciences data.

iii) Finally, it would be also necessary to choose what has to be simulated in human behavior. The question is raised especially for connectionism, which claims to get a simulation without formal control. The consequence could be to get a so perfect imitation of the human behavior that intelligence could be the result as well as stupidity. And who needs an AS instead of an AI. As it seems essential to remind that sometimes human beings do not simulate, but pretend to be reasoning. It seems that we have some manifestation of it in phenomenological philosophy.

The generous critique can understand however that resistance can be more vivid concerning intelligence simulation, looking as transgressing a taboo, than in another simulation case. But this resistance reaches some reasoning limitations:

“The mere fact that the brain might be a digital computer is in no way ground for optimism as to the success of artificial intelligence as defined by Simon or Minsky.” (Dreyfus, p. 159).

In this case, especially in the critique of biological assumptions framework, it yet would be a simulation true to the original brain! But in these assumed negations of hypothetical conditionals, a classic simulation without identity of the device is anyway refused, because the Dreyfus’ model, which can be also reduced finally to a petitio principii, is grounded on the biological reductionism and on the absurd model/object identity only reaffirming the impossibility of any simulation. When he thinks that “equivalence in the psychological respect demands machine processes of the psychological type” (Dreyfus, p. 168), he agrees with Searle who demands computer simulation of a mental process in order to having a mental state. It is the homunculus problem (or fiction of the mind) encased in the brain like a nest of Russian dolls. It is about forging a dream mental state to think the thought. While, it is precisely simulation itself that is the mental state, without causal consequences, as contradictorily acknowledged by our author:

“No one supposes that a computer simulation of a storm will leave us all wet, or a computer simulation of a fire is likely to burn the house down.” (Searle, pp. 37-38).

The meteorological hypothesis quoted above is precisely a good example of the case in which a simulation claims a real situation duplication, in order to forecast weather, and not to wet the audience. If a human being talks about or thinks to these phenomena, they won’t occur either, contrary to the parapsychological mental causation consequence. It looks as if mentalism doesn’t distinguish, as any parapsychological standpoint, the phenomena under the thought control and those independent of it.

The limitations of simulation in general do not forbid either to be exclusively interested in the way human beings deal with information, or even to only one aspects of this processing, without consideration of some perturbations. If some real, but irrelevant parameters had not been voluntarily, intentionally, neglected modern physics would have never existed, as reminded by Weizenbaum himself. And without any other reductionism that this normal abstraction process, it is legitimate to think that it can be the same in human or social sciences. In the same way, as a specialist, Weizenbaum, although an AI opponent for moral reasons, cannot obviously back the demand of model identity (Weizenbaum, pp. 164-165). In the experiments of aviation beginnings, he distinguishes three modes (simulation, performance, and theory), without privileging a priori any one. It is necessary to note that at the same time, some scientists could claim: ‘the heavier-than-air will never fly!’ what indicates that it is difficult, for a contemporary, to know on what criteria rely on, since none of the modes, performance, simulation, or theory was currently satisfactory. It is certainly the only rational meaning of the Feyerabend’s epistemological slogan “Everything goes.”

More, the paradigm principle has some limitations, since there is bad examples too, as recorded by Dreyfus himself, without drawing any theoretical consequences (what is therefore normal): “Bar-Hillel’s point is well taken; his example, however based on a particular physical fact is unfortunate” (Dreyfus, p. 215). And this bad example can be a good one. But doubt crept into the mind about this author’s other productions, like precisely his report against automatic translation, grounded on this kind of arguments.

More specifically, it could have been simply considered that, in lack of modeling, the necessity to understand the world requires to generate more or less exact analogical explanations. The passage from metaphor to model could be characterized by taking seriously this progression of representation, contrary to both ironies below:

“The brain, always understood in terms of the latest technological inventions, was understood as a large telephone switchboard or, more recently, as an electronic computer. […] This model is still uncritically accepted by practically everyone not directly involve with work in neurophysiology […].” (Dreyfus, p. 159).

“In my childhood we were always assured that the brain was a telephone switchboard. (‘What else could it be?’) [NOTE 55] I was amused to see that Sherrington, the great British neuroscientist, thought that the brain worked like a telegraph system. Freud often compared the brain to hydraulic and electro‑magnetic systems. Leibniz compared it to a mill, and I am told that some of the ancient Greeks thought the brain functions like a catapult. At present, obviously, the metaphor is the digital computer.” (Searle, p. 44).

The a posteriori irony is obviously pathetic. If all these great minds have had recourse to these metaphors, it is most possibly because: i) They were useful and heuristic in their times; ii) They allowed to differentiate processes from other ones; iii) The idea of mechanism is rationality, modularity.

Nevertheless, each of these analogies is gradually improved, i.e. precisely less analogical (in the derogatory meaning), or more analogical (in the modeling meaning). This evolution is equivalent to the taking into account of cumulative results of science. Even the analogy of the “brain as a general-purpose symbol-manipulation device [operating] like a digital computer” (Dreyfus, p. 162) can turn out to be fertile. It is here only a distribution and specialization problem, because it is necessary to choose between specialization benefits (the rapid sublation of former models), and those of interdisciplinarity (provoking the protracted usage of these former models). We recognize here the trivialization, easy for ungrateful heirs:

“No matter what dunce laughs with Pascal at ideas of the Father Mersenne, no matter who mocks his millions of ancestors who did not know that the earth was round.” (Jean Fourastié, Faillite de l’université [University Collapse], p. 115)

Let’s notice too that the intentionalist idea, grounding its phenomenological analogies on ‘mind’s intent’ (more clearly ‘visée’ [aiming], in French), therefore indicates the primitive brain analogy with a catapult! What would seem to limit phenomenology to a paraphrase of antique analogies, which is easily explained by the traditional philosophy academic practice.

The cause of all these confusions, or of this dependence on some metaphysical clichés, is certainly the rejection of empiricism. The epistemological fashion or vulgate, in this 20th century end, seems to have ratified the anti-empiricism, anti-positivism or anti-Observation victory. The terms ‘fashion’ or ‘vulgate’ could make think of a rhetorical device to depreciate the adversary’s standpoint. On the contrary it is a euphemism, what is a rhetorical effect too but, in this circumstance, it decreases the blunder. It can be wondered, elsewhere, in the critique of assumptions framework, what is an “empirical hypothesis which has had its day” Dreyfus, p. 162), if asserted in the same time that empiricist discourses are grounded themselves on theoretical assumptions. The term ‘empirical’ is simply equivalent to an insult (or a paraphrase of non-mathematization), which abandons here, contradictorily, the concrete aspect of paradigm.

The philosophical tradition stemmed from Kantianism, although it likes to repeat that he represents the end of metaphysics, has finally regressed in ontology. Kant begins actually, as soon as the first page of the Critique of Pure Reason (1781) first edition preface, by a spiritualism refusal:

“Human reason […] therefore finds itself compelled to resort to principles which overstep all possible empirical employment, and which yet seem so unobjectionable that even ordinary consciousness readily accepts them. But by this procedure human reason precipitates itself into darkness and contradictions; and while it may indeed conjecture that these must be in some way due to concealed errors, it is not in a position to be able to detect them. For since the principles of which it is making use transcend the limits of experience, they are no longer subject to any empirical test. The battlefield of these endless controversies is called metaphysics. Time was when metaphysics was entitled the Queen of all the sciences; and if the will be taken for the deed, the preeminent importance of her accepted tasks gives her every right to this title of honor. Now, however, the changed fashion of the time brings” (Kant, Critique of Pure Reason, p. 29).

One of the origins of this spiritualism explicit rejection is grounded on the critique of Swedenborg, because his logical outcome is spiritism. Unfortunately, this source is not quoted explicitly by Kant, neither this specific critique of latent or obvious spiritism, is truly developed. Anyway, it won’t have been retained, by the Kantian or neo-Kantian tradition. It can even be told to have been inverted in critique of empiricism, from the same Kantian sources. And metaphysics impossibility has been transformed, that beats everything, into empiricism impossibility. It is possible to admit the Kantian principle: ‘This is not because knowledge begins with experience that it derive from experience’. But it is a petitio principii, still waiting its demonstration. And phenomenology cannot bring it, since it valorizes the practical and human experience, as Pamela McCorduck had reported it:

“Robert K. Lindsay, in his review in the prestigious journal Science, was amused to point out that Dreyfus felt compelled to use empirical evidence to shore up an argument against the empirical tradition.” (McCorduck, Machines who think, p. 203).

This particular method of empiricism negation, which it is necessary to conceptualize, is characterized by the moral benefit appropriation of the Kantian innovation, with the negation of its content. Many epistemology works amount to the gloss and illustration of this fragment. This defines these works in more academic than scientific terms [NOTE 56]. Some scholars do not act like this in a cynical way but they believe or finally believe in it. Students are therefore under the teaching in question. And a phenomenon of double thought or double language become established. We will retain therefore this interpretation error of Kantianism that it implies a permanent reference to literal sources (here, the critique of spiritualism, or rather of the spiritualist solution to the problem of knowledge), on pain to reproduce the same mistake. Dreyfus excellently formulates it: “One must not become so fascinated with the formalizable aspects of a subject than one forgets the significant questions which originally gave rise to the research […].” (Dreyfus, p. 233).

Some philosophers think, indeed, that the abstraction is a sign of thought (Pythian’s thought reminiscence?), as to produce many discourses with an obscure generality – at least ¾, usually by removing any contextual information. This can justify, it is true, the phenomenological reaction! On the contrary, let’s notice that Searle has a specialty: the reestablishment of the empirical source of big (and small) philosophical questions. But he make then obvious their speculative nature (as he says himself) from no matter what harmless, even fictional, experience: “If you reflect on the fact that a punch in the eye produces a visual flash (‘seeing stars’) even though it is not an optical stimulus.” (Searle, p. 10).

Mentalism is therefore grounded on the obviousness of mental state, even the most doubtful. The philosophers’ practice consists then in agglomerating their prejudices (here critique of science and positivism, etc.) to no matter what sensitive experience like the subjectivity of pain (Searle, p. 16). This kind of phenomenon usually produces recourse to spiritualist explanations tending to parapsychology. There are many great classics of allegedly unexplained and inexplicable facts but explained jus the same. With “the phantom‑limb pains felt by amputees” (Searle, p. 19), he builds a theory justifying idealism and the negation of the body materiality (in contradiction with Dreyfus), from a trivial fact – and mysterious in the dualist framework.

The idealistic thought is mythological as providing always a coherent answer to any problems or mysteries. Common sense is rather considering them, therefore rightfully, as oddity and entertainment, because common sense is neither coherent, nor – obviously – complete. And precisely, the specificity of science is also to accept incompleteness. Blaming the negligence of what science is not pretending to deal with is an epistemological incomprehension. Even if some behaviorists like Skinner (in Weizenbaum, p. 245) can go so far as a hegemony, or a clumsy, but human, arrogance.

In the referential epistemological framework for common sense, medicine, grounded largely on empiricism, doesn’t cure what it doesn’t know how to cure – and that it doesn’t have to claim knowing how to cure. The patient or his close relative can consult the bonesetter, or parallel medicine. And it is possible that some of these practices are efficient in some cases. But they are epistemologically wrong in claiming overall solutions. And their possible validity belongs to medicine, despite the physicians’ possible prejudice. These prejudices concern corporate monopoly problems and not medical knowledge (from which considerations more specifically sociological are valid).

Finally, empiricism cannot be denied, especially due to fact that it common practice every day. Even when it is about cognitive phenomena, it is necessary not to forget that the origin of our knowledge is observation. The fact that it is empirical or experimental doesn’t change a lot. It can be admitted that experimental method solves the traditional opposition between theory and practice by structuring experience in a theoretical framework, providing not to use this stratagem to regress to the only productivity of theory. Actually, the confusion rests most possibly on the fact that a logical status of empiricism is not acknowledged. Yet, the solution is quite simple, while this status simply lies in its nature of existential proposition producer, obviously conditioning experimental method.

An explanation of the phenomenological standpoint could also come from the method that demands that science be defined by the break with previous scientific theories. In the Galileo’s case, when the debate has actually taken place, we know today, that there were good reasons to doubt its system, whose arguments were actually false on several points; and because it anticipated dubiously on demonstrations on others. It is necessary to note too that its opposition at influence of the moon on tides contradicted his contemporary sailor empirical knowledge! Wrongly obviously, also because a theoretical knowledge has to account for experience (‘saving phenomena’), and not to be blandly satisfy in contradicting idées reçues, as the vulgate of epistemological breaks would demand it – what is an idée reçue as well. Contrary to the appearances rejection (old) fashion also met in philosophy, it is often correct to claim that reality can be used as sifting to theoretical hypotheses. An explanation doesn’t consist in denying phenomena, besides a possibility of reinterpretation, it aims to reduce them to their determination (and not to say as Searle should: ‘It’s the way it is’).

Holistic philosophy of course refuses technoscience. While all human skills, to which it refers, are nevertheless technical. Each is about a limited area, with routine behavior what is more. But the critique of technique being classically a philosophical leitmotiv, AI adversaries do not want to lose its benefits. It is enough to call any realization ‘simple technique’ (Weizenbaum, pp. 196-197, 269), to disqualify it epistemologically speaking. This knowledge/technique splitting has nevertheless shown its insufficiencies, emphasized by Weizenbaum and Searle themselves. Manufacturers of astrolabes, or cartographers of former centuries, have expressed too a comprehension of their universe (as shown by Boorstin in The Discoverers). This did represent what Weizenbaum claimed, one of the “other ways of understanding” (p. 16) that science would have made illegitimate. What else is left except parapsychology?

It can be noticed too that this same philosophy, when it wants to valorize its own discourse by diverting a term from its common use, claims using it ‘in its technical meaning in philosophy’. That simply means it is about a local jargon, generally badly translated from German or Greek, or interpreted according to a fantasy etymology (a French sociologist Michel Maffesoli’s specialty), whose examples are plentiful: ‘to comprehend, is to take with’ (but what means ‘to take with’?); ‘Religion, in the religare meaning’ (while religions are obviously what unites (religare) a community, but also what separates it from others, especially by food taboos and other practices).

The applied science itself also demands an epistemological acknowledgement into scientific practice. This question is not absolutely solved by epistemologists. Weizenbaum is only a flagrant illustration among others, while he began by easily criticizing Marxism, so discredited today, which advocating planned science, through the opposition of Michel Polanyi to Nicolai Bukharin (Weizenbaum, p. 1). But he criticizes in extremely severe terms this idea of pure science when his adversaries choose it: “Scientists who continue to prattle on about ‘knowledge for its own sake’ in order to exploit that slogan for their self-serving ends have detached science and knowledge from any contact with the real world.” (Weizenbaum, p. 173).

Besides these contradictions, this author denies in the same way the possibility of an instrumental reason. But the debate on the opposition between pure and applied science is actually, obviously, settled by technique for long. Its efficiency is its best guarantee (operative criterion). One more contradiction won’t make much difference for the phenomenological critique of technoscience, by the lack of taking into account of material world, that is quite equivalent to technique, kind of applied and coherent phenomenology.

With quite a lot indulgence in front of its (natural or human) sciences refusal, the phenomenological theory could be understood as limited to a thought on fiction. This would explain this strange and silly apology of art reigning in current philosophy: it would have forgotten that art is its starting point. And philosophy would made his own the presumptuous artistic cowardice consisting, when it suits, in claiming to be the expression of the deepest truth about world or human being, until explicit naturalism; but fleeing, when being fault, in the facility of the creation arbitrary, possibly guaranteed by Kantian subjectivism. To think one is a God’s gift for mankind does induce rationalizations.

An important principle of social game consists in mastering archaisms, at least to decipher old texts. To handle artificially these archaisms in view of an overcoding for selecting happy-fews can characterize a part of this literary game – more or less exclusively, because scientific texts allow too some liberties, or also are under the weight of habits. Culture is usually characterized by the decoding possibility of superpositions of interpretation layers, of discoveries, rumors, and errors transmitted over the centuries. A given civilization is precisely defined by this transmission space of learned answers, true or false, to questions that its members have raised.

This taking into account of the new hermeneutic criteria has however to be limited to doesn’t sink, as emphasized by Umberto Eco or Raymond Boudon: in what the first considers as an over-interpretation delirium (and that gives him an explanation of parasciences); and what the second analyses by bringing an assumption to light, “Everything is meaning something,” motivating these over-interpretations.

The usage of fiction by AI adversaries allows them to use metaphors themselves, by sheltering behind a literary valorization, admitted by intellectuals. Dreyfus considers as a good prophecy “the film 2001, A Space Odyssey, [where] a robot named HAL who is cool, conversational, and very nearly omniscient and omnipotent” (Dreyfus, p. 80). This standpoint can also be a lack of understanding of the metaphor, because as it consists in illustrating a thesis, to say the thesis is true because its illustration is identical, becomes strongly specious. More, we will notice the self-satisfactory illusions, about the same metaphors, argued by the principle of authority:

“A Space Odyssey was made with scrupulous documentation. The director, Stanley Kubrick, consulted the foremost computer specialists so as not to be misled as to what was as least remotely possible.” (Dreyfus, p. 80).

The director in question has certainly not consulted the foremost specialists for the philosophical extrapolations concluding the movie. And due to respect of refutation principle, he certainly hasn’t consulted the other literary authorities, like Isaac Asimov, concerning robots.

Finally, the literary discourse can be considered, from an expressive point of view, with the same sifting used by Wolf Lepenies, in his book The three cultures (about the science, literature, and sociology opposition). This author characterizes literature as a freedom of expression substitute in dictatorial countries, as illustrated, according to him, by the human sciences nonexistence, in South America. It is a fact that fiction is always a symbolic or descriptive genre about the current society. The most traditionally, it concerns more or less tout observation of human behavior, generally with a moralistic or critical involvement. From this point of view, fiction simply represents a form of (political) sociology, or psychology, on topics neglected by scientific works.

The principle of symbolization, subjectively speaking, would characterize thus the lack of direct and explicit grip on political or material world. The unfree subject, or incapable to express himself with appropriate concepts, would therefore use literature to compensate his lack of personal fulfillment. The best example would be the traditional feminine culture. The literary discourse has generalized the model of high culture for it own sake, feminine specialization of the 19th century, due to women exclusion of the (upper/middle) civil life [NOTE 57].

But also, in what can be called academism or classicism, art can be explained as frequently by a particular thesis illustration. Paradoxically, the modern discourse on art, inspired by Kantianism, perceives precisely the artist’s work as (cultural or ideological) assumption carrying out. This could rather be the definition of programming. On this mode, since some decades, French novel is devoted: on the one hand almost entirely to the Freudian catechism; and on the other hand, for the elitist minority, to the writing self-reference; and since a few years, the new obliged topic is Holocaust [NOTE 58]. It can easily be observed that good reviews, and literary or cinema prizes, nearly always award the conformity to these neoclassicisms. This literature conformism thus contradicts this appreciation (except if we consider it as its definition):

“An artist, for example, does not have a criterion of what counts as a solution to his artistic problem. He invents the problem and the solution as he goes along. His work may later determine standards of success, but his success is prior to the canons later introduced by the critics.” (Dreyfus, pp. 341, note 2).

With the classicism principle, this artist elitist image is therefore questioned. More, like always, some exceptional productions can be used as guarantee to the usual or mediocre productions heaps. On the other hand, it is quite certain that any research worker is in the same situation, whatever his field is, with however the same reservations than for artists.

Fiction wrecks too the negation of any rule by phenomenology, because, on the contrary, in literature, there is application of construction or encoding rules. A good example of finite rules game is the classic detective novel, what Weizenbaum notices too:

“A good detective story, perhaps we should say a ‘fair’ one, is one that gives the reader all the information necessary to discover the truth, e.g., who did it, before explaining how the detective made his deductions.” (Weizenbaum, p. 70).

A limitation to this observation lies in the new (detective) novels, or other contemporary artistic works, neglecting these epistemological closure principles of narration – and which are supposed to be creative (what is not so sure [NOTE 59]). It is rather the partisans of parapsychological solution (like the new TV serials, Twin Peaks, Millenium, X Files, etc.) beloved of AI adversaries, which regress to the narrative solutions of a deus ex machina). But conversely, the rules existence is sometimes claimed to characterize (true) literature. Another AI opponent proposes by this way the exclusion of science fiction (SF) from the legitimate literary area for the psychological validity sake:

“How to characterize this literature [SF] and why does it have such a place in the DP men or computer scientists’ culture? The ordinary literature is a game inside rules, more restricting than they look like, of the social life we know currently. […] By counterpoint, authors of SF literature are especially bound to the creation of news social rule.” (Philippe Breton, La tribu informatique [The Computer Tribe], p. 76).

For this author, SF seems only good enough to satisfy the DP men or computer scientists obsessions – even if micro was not yet in fashion in the beginning of the SF revival. And of course if data processing did not even exist in Jules Verne or Herbert G. Wells’ time.

Nevertheless, in SF, the creation possibility of new social frameworks to define plots or characters seems to represent the most concrete fulfillment of a modern writing theory. The fact that these scientific fictions are more or less mythical is simply a continuation of the normal essence of literature. Without inevitably demanding literature to become SF or futuristic one, it would be normal of not assuming to be obliged any longer, as today, having an anti-scientist discourse (with Heideggerian guarantees). Conversely, the scientific aspect of SF allows the integration of current knowledge, and the rejection of pre-scientific clichés. A literature not being science fiction could, most often, be called ignorance-fiction!

Let’s notice too, that the fiction deciphering competence is elsewhere put to the test today, because usually, the message situation is to be understood contradictorily. Professionals who pride themselves on modernity do not inevitably establish this competence. It can be shown for instance in the fact of taking voluntarily the side of anti-hero (bad guys in fiction, delinquents in reality, and looser in both). This inverted identification to anti-hero, perceived as subtler than the stupidly positive hero one, took especially place during the institutions leftist protesting phase. But this confusion possibility is not especially new, as reminded us the French historian Jacques le Goff:

“The examples (exempla) were written to amuse, entice peoples. They delighted of them while it was said: ‘Beware! Do not act like this guy whom this story is talking about’.” (Jacques le Goff, in Télérama, 22 Oct. 1986, p. 61).

[NOTE 43] Paradigm as a ‘concrete case,’ instead of ‘abstract scheme,’ like Dreyfus (p. 39) wishes it.

[NOTE 44] The attempt to call Newton a mystic is not correct while Descartes does not consider the empirical observation of magnetic phenomenon (like movement without contact), whose taking into account represents the Newton’s explanatory principle.

[NOTE 45] Let’s notice that the French philosopher Alain Finkielkraut wants to benefit from both systems of valorization, by claiming to detest slogans, ads, TV, etc., while developing a phenomenological refusal of automation in general, and of reasoning more especially.

[NOTE 46] Note that the legitimacy of functioning in term of secrets is grounded on the lack of the strictly intellectual property protection. It is especially obvious since the computer generalization: a software is duplicable ad infinitum, without no expense, contrary to the book.

[NOTE 47] Contrary to the anti-scientist cliché, there is nothing like a ‘myth of knowledge,’ or ‘myth of progress,’ like it can also be heard. Myths in question are those of absolute knowledge, and of saving progress (or eudaimonistic progress). But if it happens to be a myth anyway, it is saving ignorance one, excepted maybe under the happy idiot form.

[NOTE 48] Minsky, Semantic Information Processing, p. 16.

[NOTE 49] Feigenbaum et Feldman, Computers and Thought, p. VI.

[NOTE 50] Can we see in the overabundance (twelve time in an hour) of this mannerism in France Culture (French cultural broadcast station), in a former morning program, an expression of the natural social controller role played by intellectuals? Can it be satirized by calling this station ‘France Culture in quotation marks’?

[NOTE 51] One could have conversely remind another rhetorical option, according to whether: “Gravity is only idiots’ happiness.” (Montesquieu).

[NOTE 52] In general, a Heideggerian philosopher would specify the etymological origin of the term enthusiasm: “possessed by the divine,” that it would be reserved to poets. This refusal to admit the DP men or computer scientists enthusiasm, therefore, does not acknowledge poet and dream in any human, in the Nietzschean and anti-Christian tradition. It is not necessary therefore to pretend being surprised of historical consequences of this segregating strategy.

[NOTE 53] What can represent a solution of a France Culture discussion, with French social scientists Jean Cazeneuve and Alfred Sauvy, about humor, in which this question of the opposition between humor and joy had been asked.

[NOTE 54] French universities (i.e. American colleges + universities) are open to everyone having a high school degree, and are both very specialized but not professional, and almost free. Oppositely, the grandes écoles are both more general, directly professionally oriented, and the entry is very competitive (generally the selection happens in the two last years of high school, to be admitted in preparatory ‘prépa’ classes during two years). A few of this grandes écoles are private, while almost all the universities are public, and the proportion is less than 1% for the very grandes écoles among 2,200,000 students (but of course there is other professional studies like medicine, law, engineering, etc.).

[NOTE 55] The ironic commentary is from Searle.

[NOTE 56] An often observed example of empiricism refusal is the fact that Darwin’s theory would not be grounded on observations but on former theories more or less mysteriously left into darkness. As it happens to be the Malthus’ ones, it is enough then to ascend to preceding mortality tables to stem them from observation. Would it be claimed possible to refer to a former theory in a regress to infinity? The dogma according to which it is always necessary to ascend to theoretical sources, is therefore easy to handle, for those respecting taboos. It is enough to stop when it is satisfied.

[NOTE 57] Under its verbalization ‘making dream,’ this model is still fashionable in French high schools and in the media.

[NOTE 58] Out of the necessary historical researches (or testimonies collection), this matter could be considered as an unhealthy exercise of styles. But it is true that some persons seem take consciousness of some realities only on the fictional or narrative form, like it was the case for the Gulag.

[NOTE 59] The best example of this pseudo-liberty is the Godard films in which characters and situations are not clarified for the audience. A contemptuous and explicit claim of the traditional narration refusal is elsewhere his leitmotiv. Contrary to his pretensions, one can see there simply an intellectualist bias (or more specifically critical, since Godard was critic himself). The principle then simply amounts to the accumulation of cultural and political references, or happy-fews codes, simply juxtaposed. The other critics have noticed it … lately … when they no longer share the same prejudices, or when the artist’s work breaks a taboo, like in Télérama, the Christian French television newspaper (and the only one in the world hating television), concerning Je vous salue Marie [Hail Mary], of this same author.