The Question of Artificial Intelligence

Philosophical and Sociological Perspectives

Brian P. Bloomfield

• Publisher: Routledge
• Year: 2018
• Print ISBN: 9781138585317
• eBook ISBN: 9780429999581
• Edition: 1
• Topic: Social Sciences
• Subtopic: Sociology
• Index: Social Sciences

Originally published in 1987 when Artificial Intelligence (AI) was one of the most hotly debated subjects of the moment; there was widespread feeling that it was a field whose 'time had come', that intelligent machines lay 'just around the corner'. Moreover, with the onset of the revolution in information technology and the proclamation from all corners that we were moving into an 'information society', developments in AI and advanced computing were seen in many countries as having both strategic and economic importance. Yet, aside from the glare of publicity that tends to surround new scientific ideas or technologies, it must be remembered that AI was a relative newcomer among the sciences; that it had often been the subject of bitter controversy; and that though it had been promising to create intelligent machines for some 40 years prior to publication, many believe that it had actually displayed very little substantive progress.

With this background in mind, the aim of this collection of essays was to take a novel look at AI. Rather than following the path of old well-trodden arguments about definitions of intelligence or the status of computer chess programs, the objective was to bring new perspectives to the subject in order to present it in a different light. Indeed, instead of simply adding to the endless wrangling 'for' and 'against' AI, the source of such divisions is made a topic for analysis in its own right. Drawing on ideas from the philosophy and sociology of scientific knowledge, this collection therefore broke new ground. Moreover, although a great deal had been written about the social and cultural impact of AI, little had been said of the culture of AI scientists themselves – including their discourse and style of thought, as well as the choices, judgements, negotiations and competitive struggles for resources that had shaped the genesis and development of the paradigmatic structure of their discipline at the time. Yet, sociologists of science have demonstrated that the analysis of factors such as these is a necessary part of understanding the development of scientific knowledge. Hence, it was hoped that this collection would help to redress the imbalance and provide a broader and more interesting picture of AI.

1

AI at the Crossroads

S.G. Shanker

Suppose I wished to show how very misleading the expressions of Cantor are. You ask: ‘What do you mean, it is misleading? Where does it lead you to?’

Wittgenstein, Lectures on Aesthetics

1.1 PARADIGM REACTION(ARIE)S

It would be difficult to say which is the more striking: the pace of the internal development of Artificial Intelligence (hereafter AI), or its cultural assimilation. Rarely has a revolutionary technology matured so rapidly; and rarely have so many enjoyed the opportunity and the competence to discuss it. From the seeds of an obscure paper on the Decision Problem which appeared in 1936 and a pair of technical papers inaugurating Information Theory twelve years later, there have erupted thriving industries spanning the full spectrum of post-industrial society, providing gainful employment not just for computer scientists and communications engineers, but on the basis of their achievements, for biologists, neurophysiologists, psychologists, linguists, sociologists, and philosophers. Especially for philosophers.

Many from these professions are distressed by the Promethean implications of AI; only philosophers have been preoccupied with its intelligibility. Unfortunately, their efforts to clarify the coherence of the notion of ‘artificial intelligence’ are in danger of being smothered by the good intentions of their humanist cohorts. For it seems perfectly obvious to everyone concerned with the legitimacy of AI why philosophers should have devoted so much attention to what Ayer dismisses as ‘the sterile question whether machines can be said to think.’¹ The lingering influence of the Great Chain of Being manifests its presence in the pervasive resistance to any developments which might impugn mankind’s privileged cognitive position on the evolutionary scale. In addition, since philosophers have hitherto enjoyed a ‘monopoly’ on questions concerning intelligence and the nature of mind, they are predictably reluctant to abandon one of the last remaining bastions of their authority without a prolonged struggle. From a sociological point of view the widespread philosophical opposition to AI was thus entirely to be expected: ‘Such a reaction is not at all surprising given the sensitivity of such a goal to peoples’ images of themselves’. For

The AI approach is seeking to establish and legitimate a view of intelligence and the nature of mind which challenges the received commonsense view of mind and intelligence as something rather special and certainly well beyond the reach of scientific analysis.²

Professional interests apart, the AI-scientist still finds it puzzling how anyone who was familiar with the inexorable erosion of philosophy’s sphere of influence at the hands of science could nonetheless oppose on philosophical grounds the exciting ‘cognitive space’ which has been opened up by AI. Is it simply because philosophers congenitally/professionally lack the imaginative flexibility to respond to new scientific paradigms? To philosophers clearly goes the sorry distinction of upholding the traditions of armchair science. Throughout their checkered scientific history they have emerged as the standard-bearers of archaic theories:

Those unwilling or unable to accommodate their work to [a paradigm revolution] must proceed in isolation or attach themselves to some other group. Historically, they have often simply stayed in the departments of philosophy from which so many of the special sciences have been spawned.³

Contemporary philosophers are thus condemned to bear in silence the ignominy of the past failures of their forbears; what AI scientist is going to believe that they are not guilty of the same purblindness today? Or perhaps it is simply imperiousness? Like an idealistic Jean Monet, the AI scientist would like nothing better than to see ‘the distinctions among AI, psychology, and even philosophy of mind . . . melt away’.⁴ However, philosophers steadfastly refuse to join in this common market; perhaps it is no surprise that so much of the philosophical opposition to AI today stems from Britain.

This is not simply a case of the analytic sons inheriting the sins of their natural philosophical fathers; for the picture developed by one of the most influential of their own peers seems to guarantee that the more they struggle the more they undermine their own efforts, however subtle these might become. The philosophical controversy which the advent of AI has unleashed is itself the crowning proof for the advocates of the Mechanist Thesis that theirs is a bona fide scientific revolution as Kuhn has described it.⁵ For if philosophers are unable — or unwilling — to concede the significance as opposed to the feasibility of the Mechanist Thesis, this can only be because they are committed to an outmoded set of ‘preconceptions’ about the nature of man and machine. Viewed from an Enlightenment perspective of man’s inalienable rational superiority, the new AI-paradigm will indeed seem unintelligible. Thus, the argument that ‘“AI” is not an analogy, it is a confusion’ is exactly how it would appear to someone operating within the parameters of an outmoded paradigm; and the more carefully philosophers demonstrate the ‘category mistakes’ involved in the Mechanist Metaphor the more they confirm the naive sociological thesis that they are paradigm reactionaries. On this picture the philosopher’s greatest weapon — the demonstration that the Mechanist Thesis violates the rules of logical grammar — is rendered totally innocuous, and philosophical critique is silenced before it can even begin.

The AI scientist impatiently protests that the philosopher’s objections to the intelligibility of the Mechanist Metaphor should apply to all scientific analogies; for, as many have pointed out, the very essence of scientific metaphor is its illogicality.⁶ And yet philosophers are understandably loathe to condemn scientific analogies tout court. But then, why should ‘artificial intelligence’ be any different from (e.g.) the hydro-dynamic analogy in the theory of electricity? At least, how can the philosopher object to the former but not the latter on a priori grounds? To the Mechanist what philosophers fail to appreciate is simply that it is always pointless to attack the logical consistency of a metaphor; all that matters is whether the field in question displays its own internal logic which — as in the case of AI — has proven to be demonstrably fruitful. Even if current theories are wrong, that in no way entails that they are unintelligible. Hence ‘Artificial Intelligence is neither preposterous nor inevitable: rather, it is based on a powerful idea, which very well might be right (or right in some respects) and just as well might not.’⁷ The ultimate answer to what is seen as philosophical casuistry is thus: how do you reconcile your pedantry with the growing success of expert systems? That is, how could a confusion be so useful? The very success of the Mechanist Metaphor serves at one and the same time as compelling testimony to the fact that ‘artificial intelligence’ marks the emergence of a new paradigm, and per consequens, one which is only criticisable from a position within that paradigm !

If philosophers genuinely wish to participate in the revolution would they not be wiser to accept their officially sanctioned role as scientific underlabourers whose minor responsibility may be to clarify what these inferential machines can and will be able to do and how they accomplish this, but whose major challenge is to accept the light which this sheds on such perennial philosophical issues as the nature of the obscure processes which occur in either the mind and/or the brain? Thus Ayer indicates that the real problem here is not so much whether computers think as whether thinkers — i.e. their brains — compute.⁸ This is very much the central theme of J. David Bolter’s Turing’s Man: ‘By promising (or threatening) to replace man, the computer is giving us a new definition of man, as an “information processor,” and of nature, as “information to be processed.’” ⁹ The source of Anthony Kenny’s qualms over the so-called ‘homunculus fallacy’¹⁰ might forseeably be removed by an arresting conceptual change brought about by scientific advance: ‘Men and women of the electronic age, with their desire to sweep along in the direction of technical change, are more sanguine than ever about becoming one with their electronic homunculus.’¹¹ In the face of such dramatic socio-economic issues, philosophical squabbles over the cogency of ‘artificial intelligence’ may well seem a sterile controversy sparked off by technological impotence. However, what if the former only seem pressing because the latter have not yet been resolved? Or more problematically still, what if Ayer’s two questions are merely different sides of the same problem — mutually reinforcing or pernicious, as the case may be — whilst humanist anxieties and mechanist fantasies alike are really the consequence of deep-seated philosophical confusions?

How can philosophers hope to break out of this impasse and escape the presumption that all we are concerned with here is a conflict between opposing Weltanschauungen? How can they attack the AI paradigm without thereby endangering the role of paradigms per se, or seeing their arguments rebuffed as yet another example of philosophy’s ongoing countermarch from progress? And most significant of all, perhaps, how can they attack the Mechanist Metaphor without resembling Victorian parents forbidding their children to read novels because they are not ‘true’? The most obvious starting-point is to reconsider what we mean by describing AI as a ‘paradigm’: an exercise greatly complicated, not just by the multiplicity of meanings which the term enjoys, but even more importantly, by the categorial distinctions which these exhibit.¹² For it is perfectly conceivable to speak of the value of the ‘AI paradigm’ from a sociological point of view — e.g. in terms of identifying the goals and/or techniques guiding a community of practitioners — and even to acknowledge the many creative possibilities suggested by the Mechanist Metaphor, while maintaining that the ideological cornerstone of AI — the Mechanist Thesis — is strictly unintelligible. The net result of referring to ‘artificial intelligence’ as either a ‘paradigm’ or a ‘metaphor’ is that this shelters the argument from philosophical scrutiny; for philosophy can have no conceptual quarrel with the evolution of new perceptual tools.¹³ Even Douglas Berggren’s argument that ‘myth’ results when metaphors are taken literally seems to disarm philosophical scrutiny ab initio.¹⁴ However, no reputable science can hope to shelter forever behind the protection afforded by the analogical, and for this reason AI was forced — in order to acquire scientific credibility — to venture into the realm of theory, where the cries of metaphysics could no longer be dismissed as the symptoms of presuppositional confusion. For when philosophers oppose the Mechanist Thesis what they are objecting to is just that: a thesis, not a normative or metaphorical approach.

To revert to Kuhn, we can best appreciate what is problematic with accepting AI as a paradigm in the latter context by considering the type of problems to which this exposes us. Kuhn explains that ‘one of the things a scientific community acquires with a paradigm is a criterion for choosing problems that, while the paradigm is taken for granted, can be assumed to have solutions.’¹⁵ There is a tacit hint here — which several have picked up — that scientific hypotheses share with mathematical conjectures the fundamental trait that they can only be answered within an appropriate system. The proper philosophical response to this thesis, however, is that what scientific hypotheses share with mathematical conjectures is that they are only intelligible within an appropriate system, where rules exist for their use.¹⁶ However, what concerns philosophy is neither whether these new problems are warranted — from whatever outlook — nor if they are unintelligible from some particular viewpoint, but simply, whether they are answerable: whether and in what sense they are problems. Kuhn, however, implicitly conflates this quintessentially philosophical interest with the former — sociological — concern when he continues: ‘Other problems, including many that had previously been standard, are rejected as metaphysical, as the concern of another discipline, or sometimes as just too problematic to be worth the time.’¹⁷ However, these latter remonstrations are by no means equivalent: to object that a paradigm is metaphysical is not at all the same as rejecting its utility. Indeed, in the sociological sense there are no grounds to say that a paradigm would be unintelligible when approached from another paradigm. On the contrary, in this respect the clash between paradigms would be perfectly meaningful to opposing practitioners; any dissension would be purely empirical: e.g. on pragmatic or aesthetic grounds.¹⁸ To introduce the notion of unintelligibility the argument must shift to a grammatical plane where conflicting paradigms are comparable to, e.g. rival geometrical systems. (For example, it is not false, it is meaningless to assert in the context of Euclidean — as opposed to Bolyai-Lobatchevskian — geometry that the path of light rays might constitute a triangle the sum of whose angles is less than 180°.) But is philosophy committed to any — let alone a rival — conceptual scheme in the ‘cognitive science of mind’?

The problem with Kuhn’s picture is that it seems to force philosophy nolens volens into either a conventionalist or empiricist framework. One way or another, to challenge a new paradigm can only be to question its ability to account for diverse phenomena; and more importantly, to demonstrate one’s prior commitment to an alternative — and in this case, supplanted — paradigm. The AI scientist would hardly be so foolhardy as to claim that computers will provide the ultimate paradigm for understanding the ‘mysteries’ of the mind or the brain: that such an analogy will adequately account for all m...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Table of Contents
• Acknowledgement
• List of Figures and Table
• Preface
• 1 AI at the Crossroads S. G. Shanker
• 2 The Culture of Artificial Intelligence B. P. Bloomfield
• 3 Development and Establishment in Artificial Intelligence J. Fleck
• 4 Frames of Artificial Intelligence J. Schopman
• 5 Involvement, Detachment and Programming: The Belief in PROLOG P. Leith
• 6 Expert Systems, Artificial Intelligence and the Behavioural Co-ordinates of Skill H. M. Collins
• Appendix
• Index