Abstract
The term "the artificial" can only be given a precise meaning in the context of the evolution of computational technology, and this in turn can only be fully understood within a cultural setting that includes an epistemological perspective. The argument is illustrated first in two case studies from the history of computational machinery: the first calculating machines and the first programmable computers. In the early years of electronic computers, the dominant form of computing was data processing which was a reflection of the dominant philosophy of logical positivism. By contrast, artificial intelligence (AI) adopted an anti-positivist position which left it marginalised until the 1980s when two camps emerged: technical AI which reverted to positivism, and strong AI which reified intelligence. Strong AI's commitment to the computer as a symbol processing machine and its use of models links it to late-modernism. The more directly experiential Virtual Reality (VR) more closely reflects the contemporary cultural climate of postmodernism. It is VR, rather than AI, that is more likely to form the basis of a culture of "the artificial".
Introduction
The phrase "the artificial" is being used increasingly to denote a new aspect or even a new form of society (e.g. Negrotti, 1991; Berleur, 1993) and warrants further examination. It suggests there is a connection between knowledge and truth as constituents of deep culture and computers as processors of knowledge, but the link between the two is not always obvious. For example, our use of computers may greatly increase the number of facts at our disposal but purely quantitative changes do not necessarily require a significant cultural change in order to be accommodated. However, when the use of computers affects the criteria for what passes as knowledge, as Lyotard suggests when he says,
"Along with the hegemony of computers comes a certain logic, and therefore a certain set of prescriptions determining which statements are accepted as `knowledge' statements." (Lyotard, 1984)then we must contemplate a much stronger relationship.
One of the key issues that any description of "the artificial" must address is its relationship to the distinction between modernism and postmodernism. Definitions are never easy and there are dangers in being over-strict, but some introduction to what will become a key distinction in the argument is required.
Modernism has been variously described as:
Postmodernism, by contrast, has been described as:
One may be tempted at this point to abandon the distinction as unworkable but fortunately Hassan has provided us with two sets of "family resemblances" that sketch out some schematic differences between the two complex concepts (Hassan, 1984). It is not necessary to reproduce the entire list here, but a representative sample will hopefully capture the essence of the distinction.
Modernism Postmodernism
closed open
purpose play
design chance
hierarchy anarchy
mastery exhaustion
finished work process
distance participation
centering dispersal
semantics rhetoric
selection combination
depth surface
signified signifier
readerly writerly
cause trace
determinacy indeterminacy
(Hassan, 1984, pp.123-24)
The aspects of modernism we wish to draw on here assume an external reality
that is represented epistemologically in terms of closed formal structures
discovered by subject experts and who make such knowledge available to others.
The acquisition of such knowledge is believed to lead to mastery over the
physical world and this is seen as inherently progressive. By contrast,
postmodernism describes a world that is unable to clearly distinguish between
appearance and reality, and uses fragmented signifiers to communicate within
open and ambiguous structures which the reader must close in order to derive
meaning. There is no attempt to master the world and no sense of progress.
Within this dichotomy between two pure forms, there are finer divisions which
can be seen as transitional. For example, the avant garde was a particular
form of late modernism which sought to be in advance of whatever form modernism
currently exhibited.
One key to understanding the relationship between "the artificial" and the distinction between modernism and postmodernism is the position of artificial intelligence (AI). It is linked by name to "the artificial", yet it seems to flit uneasily between the two sets of concepts outlined above. It has aspects of both purpose and play, of finished work and process, of the reader and the writer. In attempting to address the question of the status of AI we will first examine some historical precedents for seeing connections between the development of computational machinery and major shifts of culture. Secondly, we will identify the dominant form in which computing emerged in the 1950s as "data processing". Thirdly, we will show that the emerging discipline of AI contained the promise to reflect a more contemporary view of culture, but was limited to a form of avant garde modernism. Finally, we will argue that Virtual Reality (VR) is more closely allied with postmodernism and is, if anything, closest to reflecting an emerging "culture of the artificial".
Problems in the early history of computing
The early history of computational machinery helps throw some light on the problems involved in describing the relationship of computers to cultural processes. William Schickardt built the first known calculating machine in 1623 and by the end of the seventeenth century many others had built similar devices. The philosopher Blaise Pascal was the first to attempt to sell calculators in any quantity, producing about fifty for sale around 1700, but he found no demand for them. It was not until 1820 that Charles Thomas in Paris started the first successful commercial production of calculating machines (Harmon, 1975). This delay of nearly 200 years between the technological development of the calculator and its commercial development was to some degree a reflection of the incompatibility of the idea of such a machine with the dominant ideas of the age, an age we have come to call the Enlightenment.
Enlightenment thinking was dominated by objective universality in science and individual responsibility in human affairs. It has been described as the effort to "develop objective science, universal morality and law, and autonomous art according to their inner logic" (Habermas, 1985). Writers such as Descartes, Newton and Kant held that universal laws operated everywhere, not only in the natural sciences but also in ethics and law, and they further believed that such laws could be discovered by each individual through clear perception or introspection. It was the very universality of such laws, and the inherent ability of any person to observe them directly, that encouraged the philosophy of individual responsibility. If individuals all started with an equal ability to see the world and the right thing to do, then the market was guaranteed to be free and equal and everyone could legitimately be held individually responsible for the outcome of their actions. The human agent of an action was thus seen as the basis of its legitimacy and it was natural that much store was placed at this time upon a person's name, reputation and signature.
According to Laufer (1991), the stress laid upon the human individual as the agent of action within the Enlightenment view of the world meant that it would not have been considered legitimate to pass responsibility for an action to a piece of machinery. If this is true then it goes some way to explaining why calculators were initially seen only as curiosities and were not accepted into commerce or science until the middle of the nineteenth century when Enlightenment thinking began to give way to positivism.
Under positivism knowledge was no longer seen as equally accessible by each person, but as the product of specialised work within autonomous disciplines. As Habermas expresses it, in the break with the Enlightenment "a distance grows between the culture of the experts and that of the larger public" (Habermas, 1985). The language of science also changed. The personal, narrative account favoured by authors such as Newton reflected their belief that subjective experience was also universal and it was therefore natural to write about experiments using the first person singular. By the late nineteenth century scientists had changed to using the third person passive form to indicate an impersonal viewpoint (Halliday 1988).
The abandonment of the belief that everyone was capable of knowing everything coincided with the gradual erosion of the idea of a free market. The public sector was seen to have a legitimate role alongside the private sector and legislation was introduced to control the operation of the market. At a personal level, actions were no longer judged in terms of the agent who performed them, but in terms of the objective outcome. If Kant's ethics are typical of Enlightenment thinking then the utilitarianism of Mill and Bentham is typical of positivism. In utilitarianism an action is good simply insofar as its effect on the world increases the sum total of human happiness. As Kolakowski expresses it,
"[utilitarian] ethics does away with evaluation based on intention in favour of evaluation based on results." (Kolakowski, 1972, p.99)The positivism of Auguste Comte required there to be something outside the system of knowledge that provides unity and purpose. Laufer refers to this as "God/Progress". The concept of progress, as providing an axiomatic direction to the pursuit of knowledge and morally correct action, we take as a central feature of modernism. Raymond Williams has shown the connection between the concept of progress and the concept of modernity. Up to about 1840 the word "modern" was used to refer simply to the present in contrast to the whole of the past. Then the concept came to imply a break with outdated tradition, and by the end of the nineteenth century it was used in the sense of "indicating updating and improvement" (Williams, 1989). There had been a shift in culture during this period to redefine modernity in terms of the new positivist concept of progress. Later, in the twentieth century, the concept of progress was to provide a link between positivism, the specific cultural movement known as modernism, and the social and economic programme of modernisation.
Against this background of the rise of positivist thinking, Charles Babbage produced the first known design of a programmable calculator in 1850. Once again, it was not immediately developed commercially and, in fact, it was over one hundred years before the first commercial programmable computer was produced in 1951. It is sometimes claimed that the failure to develop Babbage's machine was due to the limitations of engineering at the time. It is now known that in the 1850s Babbage examined a working model of a device similar to his Difference Engine which had been built by George Scheutz in Sweden. It is also known that George Grant built a complete Difference Engine in the US in the 1870s (Williams, 1976). Large scale production of Babbage's Analytical Machine would probably have relied upon a precision in engineering that was not developed for a further forty years but it is reasonable to ask why a programmable computer was not developed in the 1890s.
By 1890 positivism had replaced the Enlightenment as the dominant system of legitimacy within Europe (Laufer, 1991). The concentration upon the effects of actions, rather than upon the individual that performed them, legitimised the use of mechanical calculators at a time when there was also an enormous growth in the administrative capabilities of the emerging modern nation state that required the collection, storage and processing of large amounts of data (Wittrock, 1993). Babbage's ideas for a programmable computing machine arose within a society that was entering upon an obsessive affair with data.
The first use of programmable calculating technology on any significant scale was not a development of Babbage's Analytical Engine, but rather in the field of data processing. During the U.S. Census of 1890 the details of each person were punched into a machine-readable card and these were processed by tabulator. This dramatically reduced the time taken to process the census results from an estimated five and a quarter years to eighteen months, and firmly established the advantage of calculating machines in efficiently handling large amounts of data. The next 60 years saw a spectacular growth in all types of office equipment as companies such as IBM and Remington-Rand emerged as market leaders in the field (Harmon, 1975). In the US, government intervention also played a role with Roosevelt's New Deal programme in the 1930s being one clear example where a government initiative in the public sector led to a greatly increased demand for tabulating machinery (Nikolaieff, 1970). Data could be held on punched cards and tabulating machinery could selectively process large amounts of such data and produce listings and reports. It was from this tradition that the concept of the computer as a "number cruncher" was born.
When the programmable computer was rediscovered in the 1940s, out of an effort to solve several problems faced by the military, its wider social potential was still not clearly perceived. IBM refused an offer to commercially develop one of the first electronic computers, the ENIAC, because they believed at the time that it could only be used for scientific applications and had a very limited potential market. Such an attitude was also to be found in the UK: Douglas Hartree, a pioneer of computing in Britain, said in 1951,
"We have a computer here in Cambridge; there is one in Manchester and one at the NPL. I suppose there ought to be one in Scotland, but that's about all." (quoted in Lavington, 1980, p.104)Despite the range of metaphors that existed to enable people to come to terms with the computer in the late 1940s and early 1950s, for practical purposes it was seen only as a large scientific calculator. In 1950 Alan Turing, the British mathematician who first provided an abstract definition of a computing machine, published a more speculative paper entitled "Computing Machinery and Intelligence" (Turing, 1950). In this paper he described a potential for computers, and thereby a future research and development programme, that would place their operation on a par with human intelligence. Turing's views were idiosyncratic, he did not reflect the generally perceived priorities of the immediate post-War period and his research agenda was not well supported. A similar story can be told about the US at this time, where the rigid model of control espoused by John von Neumann was allowed to dominate to the exclusion of the more holistic, cybernetic view of computers proposed by Norbert Wiener (Heims, 1980).
Sales of the modern electronic computer took off spectacularly only when it became adopted by the successful business equipment industry within the cultural tradition of positivist data processing. Far from replacing the established technology, the early computing industry coexisted with and even complemented it. During 1956/57, about the time the first electronic computers were being installed in many large companies, the British Tabulating Machine Company reported that sales of punched card equipment nearly doubled (Lavington, 1980, p.81). It was the office equipment suppliers who successfully adapted the general-purpose computer to the advanced needs of the tabulator market and led to the programmable computer's first acceptance by society.
It can been said that modern computing arrived just in time to give a new lease of life to positivism/modernism. It provided extra power to perform calculations that were rapidly becoming too large to be practical by any other means. Hence the initial effect of computers in the 1950s and 1960s was conservative: instead of introducing a revolution in thinking they served to prop up systems that would otherwise have collapsed under their own inefficiency (Weizenbaum, 1976).
The 1950s in Britain saw the consolidation of the welfare state, the development of mass consumerism and the growth of multinational corporations, all of which gave rise to an increased demand for data processing. They added to the trend for more and more statistics in pursuit of the positivist goal of control through the measurement of the effects of actions and the modern general purpose computer became accepted into society primarily as a data processing machine to assist with this task. Unready to accept Turing's radical view of the future role of the computer there was a need for a distinctive description of computer technology that was convincing and was compatible with the dominant positivist viewpoint.
Defining "Data Processing"
By 1950 positivism had become concerned with language which it recognised as containing imprecision. This made the use of natural language problematic for the expression of objective certainty. This issue had concerned the Vienna Circle of Philosophers during the 1920s and led to the philosophy of logical positivism, which owed much to the early work of Wittgenstein (Wittgenstein, 1961). Logical positivism concerned itself with the way that all languages must behave if they are to refer to the world. At some fundamental level, it argued, sentences are composed of atomic symbols that simply name objects in the real world. Facts are systematic configurations of these atomic names and the meaning of these facts can be derived from two things: the objects that are named and the logical configuration of the sentence. Put very simply, logical positivism separated the meaning of the components of a sentence from the meaning added by their particular arrangement. One could say that it separated content from form.
Toulmin argues that logical positivism was both the continuation of a project first entered into by Leibniz during the Enlightenment, and the philosophy behind much of present day computing (Toulmin, 1991). The computer's emergence as a data processing machine in the 1950s and 1960s coincides with an adaptation of the language of computing to satisfy the dominant epistemological form expressed through logical positivism. This was achieved in the meta-level dialogue in which a set of new terms was introduced to legitimise the activity of data processing. From the late 1960s through to the 1980s many textbooks were published that sought to explain data processing in strictly positivist terms.
These books argued that data processing is about the transformation of data into information and that the key difference between the two is that information has a meaning or purpose. The links between data processing and logical positivism can be shown by illustrating three major concerns.
1. Data are like logical atoms, they directly represent things in the real
world
For positivists the world is represented by a set of facts: "The world we know
is a collection of individual observable facts." (Kolakowski, 1972, p.15).
Data processing also tries to capture the world as "data". It is said that
data can "represent an idea, object, condition, or situation." (Arnold, 1978)
and it can "describe events and entities." (Verzello, 1982). Data are taken
to be atomic and indisputable within the system that processes them. The
question of how data represents real things is seen as a pragmatic matter. It
concerns how various human tasks such as accounting or stock control can be
represented as data and this forms the descriptive content of these books.
2. Data must be represented uniformly in terms of symbols
Logical positivists are concerned that there should be a uniform use of
language to represent facts. Data processing exhibits a similar concern that
data, "refers to any set of characters" (Davis, 1969) and "are communicated by
various types of symbols" (Verzello, 1982). Such authors insist that data be
expressed in terms of the character set of the computer's current character set
but the implications of this are not discussed.
3. Information is data that are systematically organised
Positivists claim that scientific knowledge is gained by ordering facts
(Kolakowski, 1972, p. 15) and the logical positivists argued that facts are
atoms in a particular configuration. DP has a similar view, arguing that
information is data that are systematically organised. Gore says that data,
while inherently unorganised, must also be "able to be organized" (Gore, 1979,
p.5). Frates & Moldrup describe data processing in these terms:
"unorganized facts, or raw data, are transformed into an arranged, ordered, usable form called information." (Frates & Moldrup, 1983, p.10).Data are the atomic components of information but in themselves can only refer directly. By placing these in suitable arrangements we can generate new meanings. As Arnold expresses it,
"No compilation of data ... can be called information unless it has has been organized in a meaningful way..." (Arnold, 1978, p.2)Data processing was not just a link between an emerging technology and a powerful market sector, or the reduction of a general purpose technology to a "number cruncher" to meet the need for more statistics. It was also a process of epistemological realignment whereby the language that determined how we thought about what we were doing with computers was moulded into the form of logical positivism. This system of thought held out the hope that systems of arbitrary symbols might really reflect the totality of the world and therefore that processing such symbols might yield legitimate knowledge and enable us to exercise legitimate control over the real world either with or without human intervention.
Artificial Intelligence
Writing in 1980, R.W.Hamming, wrote,
"An examination of the history of computing shows that around the years 1952-1954 many of us came to the same conclusion that the computer was more than a number cruncher." (Hamming, 1980, p.8)For Hamming and many others the true significance of the computer was not as a data processor but as a "symbol manipulator". The philosophical implications of this change were spelled out by Minsky when he contrasted the emerging field of artificial intelligence to "the positivistic, behavioristic tradition in psychology whose rigid grip is just beginning to crumble" (Minsky, 1968, p.2). Positivism, and more particularly its manifestation within psychology in the form of Skinner's behaviourism, was under attack from the development of AI which was innovative in its methodology and its use of mentalistic concepts.
The emerging AI openly expressed its rejection of positivism but we can find the origins of this position in Turing's paper, "Computing Machinery and Intelligence" (Turing, 1950). In his description of the Imitation Game, Turing has often been seen as reflecting a classical behaviourist standpoint: if two objects provide the same responses to the same stimuli then they are of the same type. If we test two objects to see whether they display behaviour we call "intelligent" and both results are positive, then it is irrelevant that one of them happens to be human and the other a computer. The Imitation Game is therefore presented as a test to see whether computers can think, even though Turing explicitly warns against such an interpretation. In arriving at such a conclusion we had to assume that Turing's paper is primarily about a challenge to our status as privileged thinking beings, but this is questionable. Turing was, after all, more interested in writing about machines than writing about human beings. He was primarily writing about a new type of machine, one that displays both the properties of human beings and those of electromechanical devices. The mere possibility of such a machine does not sit comfortably within behaviourism for it undermines the positivist context within which behaviourism sits. If we accept that there could be an entity that obeys both the natural laws that govern the physical world and the cultural conventions that govern our everyday actions then the epistemological basis, not just of positivism, but of all Western thought since the Enlightenment is brought into question. Our traditional and most fundamental dualisms, such as those between mind and matter, nature and culture, science and tradition, are no longer reliable. It is in this sense, we argue, that Turing's paper and the subsequent development of AI can be linked to the undermining of positivism/modernism and thereby to the establishment of the conditions for the emerging culture of postmodernism.
Following Turing's lead, AI launched itself in an anti-positivist direction. It ignored traditional boundaries and became instantly interdisciplinary, embracing disciplines as previously diverse as electrical engineering, psychology, linguistics, mathematics, logic, computer science and philosophy. While positivism saw data as something given, as passive "raw" input that was processed into something useful, AI saw data as something to be sought out in the attempt to solve a problem. Minsky says that AI programs "do not have to start from an unstructured basis to evolve everything they will need", but must "understand" the information "as opposed to merely being able to exploit it" (Minsky 1968, p.17). Knowledge representation, which was a pragmatic issue within DP, became a central area of research in AI during the 1970s and 1980s.
A further break with positivism occurred over the status of theories and the empirical verifiability of AI. Minsky described AI programs as "experiments" and as "first trials of previously untested ideas". The gradual growth of the discipline marked a new type of scientific endeavour with a new role for theories and models (Hayes, 1984; Narayanan, 1986). AI became a kind of "practical philosophy" where theories could be embodied in models and the behaviour of the model was a new reality that could be interpreted in a way that was inconceivable before. AI broke with the DP tradition of generating new information and developed as a new medium within which thinking took place.
In distinguishing between modernism and postmodernism we adopted the technique of schematic lists which suggest "family resemblances" generally shared by a concept. We borrow that technique and use it again to indicate the schematic differences between DP and AI.
Data Processing Artificial Intelligence
autonomous disciplines interdisciplinary
preexisting world created world
presents facts seeks facts
information thought
purposeful experimental
proof analogy
reference metaphor
determinate indeterminate
context free contextual
Artificial intelligence challenged a world in which the computer was seen as a
rather mundane number crunching machine. The emerging discipline caused
considerable disquiet in academic and research establishments, particularly
during the 1970s. In Britain, the Lighthill Report commissioned by the (then)
Science Research Council divided AI into three areas of basic research:
advanced automation, computer based central nervous system research, and robot
building (Fleck, 1982). The Report declared that the first two of these areas
presented no a priori grounds to support AI as an autonomous disciplines
and any successes in these areas would simply be absorbed by existing
disciplines. Robot building, it argued, had the potential to ground AI's claim
to be an independent discipline but it found that no significant progress had
been made in this regard. Any appearance of success, it argued, was due to the
importation of knowledge that had been discovered separately in other
disciplines. AI, it concluded, could not demonstrate any success that could
not be accounted for by traditional disciplines.
AI was seen by the Lighthill Report as a rogue discipline that had no claim for autonomy. Partly as a response to such attacks, there began a movement within AI to establish a body of core knowledge around which a new discipline may coalesce. One early example of this was the concentration upon search as a major technique of AI. (A search problem is defined by specifying an initial state, a goal condition, and the rules whereby one state may change into another.) It served as a useful model for representing certain elementary problems and enabled early researchers in AI to produce superficially impressive demonstration programs. However, it soon became studied as a technique in isolation from any context and can now be seen as an attempt to map the territory initially explored in the name of AI in a positivist manner. Such reliance upon pure technique in AI was criticised by Minsky who has always held that techniques need to be at the service of some real intelligence that knows when and how to use them. He argued that the blind use of technique fails to represent complex skills, such as those of an advanced chess player when generating a plan (Minsky, 1968, p.11).
Amongst those who have resisted this incorporation back into positivism, the strongest defence has often been to define a new science of AI around the study of "intelligence", not necessarily as a property of human beings but as an entity that is abstracted from the experience of the human subject. Minsky distinguished between the type of AI that attempts to build models of human behaviour, from the type of AI that is
"an attempt to build intelligent machines without any prejudice towards making the system simple, biological, or humanoid" (Minsky 1968, p.7).Minsky's famous definition of AI as the attempt to build machines that can do things that would require intelligence if done by humans is a major step in this process. The origin of intelligence as a purely human property is kept intact, but a new object is reified from it. A class of behaviours is defined with respect to human behaviour, and anything that can exhibit this behaviour can be said to possess this new form of "intelligence". The underlying question of whether such a class of behaviours can be properly defined is not addressed. Ryle, for example, argued strongly that that "intelligence" is a disposition and cannot be attributed to single acts (Ryle, 1949).
The reification of "intelligence" can be seen as similar to the development of the concept of competence in linguistics. Chomsky found that everyday speech and writing was too irregular to provide a sound basis for the scientific study of language, but he did find that most mature speakers of a language agreed when asked whether a particular group of words formed a grammatical sentence. He developed the notion of "linguistic competence" to describe this latter phenomenon and defined the science of linguistics as the study of such regularities. Minsky's brand of AI sought to define a new object of study that might be called "cognitive competence" (Beardon, 1988). We can see this as rational symbolic activity that most adults would accept as correct if they considered it with sufficient time and attention.
Whilst the weaker versions of AI have adapted their methods to be more in keeping with traditional positivist disciplines, the stronger versions have persisted in anti-positivism. As such they can be seen as preparing the grounds for a culture of "the artificial" but we would question whether the break is yet complete. We see a parallel with the avant garde in art, which has been defined as,
"art of the current moment that is more advanced, more up to date, more radical, or extreme, than all other art present times." (Walker, 1973)Artificial intelligence can be seen as an avant garde form of computing corresponding to a late-positivist stage of cultural development. It disagreed with significant aspects of data processing: data was no longer given but had to be actively sought; the point was to build experimental models, not describe a real world. But despite this significant break, AI was still trapped within the concept of the computer as a machine that uses symbols and builds models. It is bound to the idea of progress and advancement and in these regards even strong AI is trapped within modernism, though desperately trying to outreach it.
Virtual Reality
Virtual reality (VR) is one of the latest trends within computing and its material base has been the development of hardware such as high resolution colour monitors, digitised sound recorders, digitised video boards, force-feedback devices, etc. The combination and synchronisation of these, in conjunction with software that enables them to run interactively in "real time", creates a new potentiality for computing that to many is just as exciting as artificial intelligence in its heyday. Along with a similar revolutionary zeal goes a similar non-conventional method of work. Benjamin Woolley says that VR,
"... has been subject to none of the usual academic controls: there has been no coordinated experimental work to test its hypotheses" (Woolley, 1992, p.18)and, in a passage reminiscent of the Lighthill Report into AI, says of VR
"most of the research done under its auspices could as easily have been done, and in some cases is being done, under the auspices of more mundane disciplines" (Woolley, 1992, p.37)
These similarities between AI and VR are superficial and do no more than place both of them outside mainstream positivist science. Their differences are more marked and more significant. Artificial intelligence is a form of computing that represents a terminal point of modernism, while virtual reality is a form of computing that has more to do with postmodernism. Virtual reality marks a significant break from AI in that its hardware enables us to interact with the computer without the use of arbitrary symbols. The ability to interact directly as if we are dealing with sense data means that VR "takes visualisation and interpretation of data to a new dimension" (Eisner, 1991). Thus, whereas at some level of description we can still argue that the computer is a symbol-processing machine, at the level of most user experience symbolic representations are peripheral to the reading of the computer. From the standpoint of such users we must see the computer semiotically as a system of signs changing in real time.
Virtual reality is more than the passive reception of artificial experiences, it is also about acting in an artificial world. Within the AI paradigm the user and the computer combine to symbolically understand a future action. AI assumes that before we act we make a plan and this will involve solving problems. It aims to help us solve such problems and make such plans, thereby assisting us by telling us what we should do. In the virtual reality paradigm the user's actions take place directly within the virtual world modelled by the computer. The experience is immediate, VR "allows a person to explore a computer-generated world by actually being in it." (Sherman, 1992), or as Meredith Bricken puts it, "What you do is what you get."
The concept of virtuality is the important concept that needs to be abstracted from VR because it finds expression in other areas of computing. Virtuality is concerned with illusion: the appearance of something that is not real. Initially we see that the concept of virtuality retains AI's commitment to a real world: we must believe that what we experience is real, but we must also secretly know that it is not. Ted Nelson, the founder of hypertext and a leading figure in thinking about new forms of computing, makes this clear.
"The virtuality of a thing is what it seems to be, rather than its reality, the technical or physical underpinnings on which it rests." (Nelson, 1990, p.239)Many multimedia systems strive for what has been described as a "believable magic", or what Alan Kay and Larry Tesler of Xerox PARC have called "user illusion" (Kay, 1990, p.199). The idea of virtuality means that the designers create a believable world inside the machine and that users will enter this world and come to believe, not that it is the real world, but that it does exist. Sherman, for example, in describing VR in the classroom, says that "... disbelief is completely suspended", and that "The pupils are literally in another world" (Sherman, 1992).
The VR world must not only be believable in itself but it must exclude other possibilities. In virtual reality, the power of photorealistic images and a seamless means of interaction combine to overpower the senses and force the subject to believe in the world that has been created. VR demands our exclusive attention. In virtual reality a person "is engulfed by the model that he or she has created" (Eisner, 1991), an experience sometimes described as "total immersion virtual reality" (Brown & Slater, 1991). In VR we are in a world that is a total experience, is highly credible, yet we know has no underlying reality. Woolley has described the computer as,
"... a `virtual' machine .. an abstract entity or process that has found physical expression, that has been `realized'. It is a simulation, only not necessarily a simulation of anything actual." (Woolley, 1992, p.68/9)We seem very close to the world described by Baudrillard as "the generation by models of a real without origin or reality." (Baudrillard, 1983, p.2).
Many pioneers of contemporary computing, people such as Ted Nelson and Alan Kay, argue that we should see the computer as a virtual world in which to act. The act of communication takes place between the user and the world inside the machine. Brenda Laurel invites us to see the computer as Aristotelian theatre through which the audience/users empathise with the characters and are purged via the process of catharsis (Laurel, 1990). This use of the theatre is one in which the audience willingly suspends belief in order to be healed and strengthened. It can be contrasted with the didactic theatre of Bertholt Brecht who was opposed to the idea of cathartic drama and wanted the theatre to be an experience that enabled users to better understand and act in the real world.
In The Postmodern Condition, Lyotard attempts to address the question of what has happened to knowledge in the information age (Lyotard, 1984). He first outlines the limitations of a positivist approach to knowledge. A positivist science that demands that experiments be capable of being reproduced is limited because there may be many events that happen but are not reproducible and therefore must be excluded from this kind of science. What is more serious is that the criterion of reproducibility implies a determinism in the world which recent science has itself shown not to exist.
Lyotard argues that the computer, as a symbol processing machine, is closely allied to positivist thinking and promotes only the formal, scientific aspects of knowledge. Scientific knowledge is not only limited by the exclusion of some factual knowledge about the world, it also excludes other types of knowledge. It excludes all practical knowledge concerning how to do things, for example, and also "the narrative", which is "the quintessential form of customary knowledge". The narrative is important for science because it explains such things as the validity of proofs and arguments. The association of computers with positivist views of science and the limitations of those views raise serious questions about the status of computers as knowledge processors in our society.
Positivism has, according to Lyotard, led to "the exteriorization of knowledge with respect to the `knower'", one result of which is that knowledge is separated from the "training of minds". That is to say, skill and intelligence are no longer the inextricable products of a lifetime of learning and experience by a human being but can be posited as things existing as objects outside of human minds. Hence Lyotard dismisses both data processing, where knowledge is represented as data in impersonal data banks, and Minsky's brand of AI, which separates the processes of knowledge from any human agent, as brands of positivism.
The exteriorisation of knowledge has turned it into a commodity whose value is decided in terms of the market, rather than in terms of truth with respect to a real world. This is not just an economic point, though it is true that we confront today the concept of "value-added" with respect to information. With the breakdown of positivism has gone a breakdown in the authority of the institutions of knowledge and their role as arbiters of truth. In an interdisciplinary world there is no authority and the pragmatic solution is to create an intellectual market-place, an environment in which the history of a statement on the network is the principle means of determining its value. Baudrillard takes this to its conclusion, arguing that the exchange-value of information has taken over from its use-value and this means that we are permanently dealing with simulations. The essence of a simulation is that there is no connection between information and some underlying reality it is taken to refer to.
Nowhere is the play between appearance and being more marked than in the quest for greater realism within VR. Frederick Brooks, a leading software engineer who now works developing virtual reality systems, makes the distinction between realism and truth. It is possible to model mountain using a number of connected irregular geometric shapes and images generated from such a model can be quite accurate, though the appearance of flat surfaces means that the effect is not very realistic. It is also possible to use fractal generated models to simulate the irregularities in mountains and the resulting images may be far more realistic than the geometric shapes. Brookes comments,
"The fractal mountains are a good clear visual image of an important distinction between realism and truthfulness. The danger of more and more realism is that if you don't have corresponding truthfulness, you teach people things that are not so." (Rheingold, 1992, p.45)
When Lyotard closely tied the computer to scientific forms of knowledge he was accurately reflecting the form that computing had taken at that time and we have seen that in his criticisms of positivism he has included one argument that indicates that Minsky's form of AI is included in this. Since then we have seen the emergence of a type of computing that embodies virtuality, in which appeal is made directly to our senses and we are invited to believe in illusions presented to us. Virtual reality is a different type of computing and there seems to be complete agreement between postmodernists and virtual realists in questioning the very status of reality.
Conclusions
We have attempted to show that there is a relationship between various models or types of computer use and aspects of our culture which we describe as epistemological. We first tried to establish this through the examination of two examples from the early history of computing. In the case of the first calculating machines, we argued that the delay of approximately two hundred years between the first known prototype machine and the first recorded commercial exploitation was related to the perceived illegitimacy of using machines to replace human action within the framework of Enlightenment thinking. In the case of the first programmable computer, we argued that Babbage's Analytical Engine was inappropriate to a society that was engaged in the collection and assessment of greater and greater amounts of data. We then argued that when the modern electronic computer emerged around 1950 it became part of a positivist/modernist revival that formulated it as a "number cruncher". This is explained not only by reference to economic and social factors, but also to the development of the meta-language of "data processing" which aligned computing with logical positivism.
Having established this evolutionary model relating the development of computers to cultural and particularly epistemological factors, we turn to a definition of the term "culture of the artificial". Traditionally, this has been closely associated with the development of artificial intelligence and symbol processing machines. We argue that even the development of strong AI is no more than an avant garde development of late positivism/modernism and does not represent a clean break from it. It is still committed to symbol processing and to models which attempt to refer to an underlying reality and it is still committed to the idea of progress and advancement. The "culture of the artificial" must embrace the artificial in the sense that Baudrillard describes. It must deal in illusions that are, at the same time, real. Because of its commitment to virtuality and illusion, the form of computing most closely associated with postmodernism is virtual reality (though we would not want this term to be restricted to those systems that describe themselves as such). We use the term to refer to any system that we interact with in non-symbolic ways and which attempts to create an illusion.
The purpose of this paper has been analytical and no attempt has been made to pass value judgments on which form of computing may be better or more desirable than any other. The central point has been to illuminate the close connections that exist between different models of computing and different cultural, and particularly epistemological, movements in society. If there is a contemporary point to be taken away from such an analysis it is concerned with the implications of the apparent oxymoron of artificial reality, which
"reveals that the things we assume to be independent of us are actually constructed by us. It reveals that being `natural', is not simply a value-free, unproblematic, apolitical objective state - though part of its mythology is to make itself appear to be so." (Woolley, 1992, p.8/9)
References
Arnold, R.R., Hill, H.C. & Nichols, A.V. (1978) Modern Data Processing. 3rd edition. John Wiley, Santa Barbara.
Baudrillard, J. (1983) Simulations. Semiotext(e), New York.
Beardon, C. (1988) Explanations in Cognitive Science. Artificial Intelligence Review, 2, 2, pp.181-193.
Berleur,J. (1993) Risk and vulnerability in an information and artificial society. In: J.Berleur, C.Beardon & R.Laufer (eds.) Facing the challenge of risk and vulnerability in an information society. Elsevier Science Publishers B.V. (North-Holland), Amsterdam, pp.3-23.
Brown, M. & Slater, M (1991) A review of interaction technologies as applied to virtual reality. Proc. Computer Graphics 1991. London, pp 309-327.
Davis, G.B. (1969) Computer Data Processing. McGraw-Hill, New York .
Eisner, R. (1991) Researchers see a wealth of applications for Virtual Reality. The Scientist, 18 Mar 1991, pp.14,16.
Fleck, J. (1982) Development and establishment in artificial intelligence. In: N.Elias, H.Martins & R.Whitley (eds) Scientific establishments and hierarchies. Sociology of the sciences, Volume VI. D.Riedel Publishing Company. pp 169-217.
Frates, J. & Moldrup, W. (1983) Computers and Life: an integrative approach. Prentice-Hall, Englewood Cliffs, NJ.
Gore, M.R. & Stubbe, J.W. (1979) Computers and Data Processing. McGraw-Hill, New York.
Habermas, J. (1985) Modernity - an incomplete project. In: H. Foster (ed) Postmodern culture. Pluto Press, London, pp 3-15.
Halliday, M. (1988) On the language of physical science. In: M. Linadessy (ed) Registers of written English: statistical factors and linguistic features. Pinter, London.
Hamming, R.W. (1980) We would know what they thought when they did it. In: N. Metropolis, J. Howlett & G-C. Rota (eds) A history of computing in the Twentieth Century. Academic Press, New York, pp 3-9.
Harman, M. (1975) Stretching man's mind: a history of data processing. Mason/Charter, USA.
Hassan, I. (1984) The culture of postmodernism. Theory, Culture and Society, 2, 3, pp.119-31.
Hayes, P. (1984) On the differences between psychology and AI. In: M.Yazdani & A. Narayanan (eds) Artificial Intelligence: human effects. Ellis Horwood, Chichester, p.157-162.
Heims, S. (1980) John von Neumann and Norbert Weiner: from mathematics to the technologies of life and death. MIT Press, Camb., Mass.
Kay,A. (1990) User Interface: a Personal View. In B. Laurel (ed.) The Art of Human-Computer Interface Design. Addison-Wesley, Reading, Mass. pp. 191-207.
Kolakowski, L. (1972) Positivist philosophy: from Hume to the Vienna Circle. Penguin Books, Harmondsworth, Middx.
Laufer, R. (1991) The history of computers: an epistemological point of view. In: J. Berleur, A. Clements, R. Sizer & D. Whitehouse (eds) The Information Society: evolving landscapes . Captus/Springer Verlag, Toronto/New York.
Laurel,B. (1990) The Computer as Theatre. Academic Press, New York.
Lavington, S. (1980) Early British Computers. Manchester University Press, Manchester.
Lyotard, J-L. (1984) The Postmodern condition. Manchester University Press, Manchester.
Merquior, J.G. (1989) Spider and Bee: towards a critique of the postmodern ideology. In: L. Appignanesi (ed.) Postmodernism: ICA documents. Free Association Books, London.
Minsky, M. (1968) Semantic Information Processing. MIT Press, Cambridge, Mass.
Narayanan, A. (1986) Why AI cannot be wrong. In: K. Gill (ed) Artificial Intelligence for Society. Wiley, Chichester.
Negrotti, M.. (1991) Understanding the artificial: on the future shape of artificial intelligence. Springer-Verlag, London.
Nelson, T.H. (1990) The Right Way To Think About Software Design. In B. Laurel (ed.) The Art of Human-Computer Interface Design. Addison-Wesley, Reading, Mass. pp.235-243.
Nikolaieff, G. (1970) Computers and society. H.W.Wilson, London.
Rheingold, H. (1992) Virtual Reality. Mandarin, London.
Ryle, G. (1949) The concept of mind. Hutchinson, London.
Sherman, B. (1992) Birth of a brave new world. Times Ed Suppl., Mar 92 Update pp.3-4.
Toulmin, S. (1991) The dream of an exact language. In: B. Göranzon & M. Florin (eds) Dialogue and technology: art and knowledge. Springer-Verlag, London.
Turing, A. (1950) Computing Machinery and Intelligence. Mind, LIX, 236.
Verzello, R.J. & Reutter III, J. (1982) Data Processing: systems and concepts. McGraw-Hill, London.
Walker, J. (1973) Glossary of Art, Architecture and Design since 1945. Clive Bingley, London.
Weizenbaum, J. (1976) Computer power and human reason: from judgment to calculation. W H Freeman.
Williams, M. (1976) The Difference Engine. Computer Journal, 19, 1, pp. 82-89.
Williams, R. (1989) When was modernism? In: Williams R, The politics of modernism. Verso, London .
Wittgenstein, L. (1961) Tractatus logico-philosophicus, trans. D.F. Pears & B.F. McGuiness. Routledge & Kegan Paul, London.
Wittrock, B. (1993) Polity, Economy and Knowledge in the Age of Modernity in Europe. AI & Society, 7, 2, pp.127-40.
Woolley, B. (1992) Virtual worlds. Penguin Books, London.