The Cybernetic Brain

by Andrew Pickering

  Beyond the mere possibility of this cybernetic stance, the proof of the pudding is obviously in the eating, though Pask does find a way of recommending it, which has more to do with the "consumption" of art than its production: "The chief merit of externalization . . . seems to be that external discourse correlates with an ambiguity of role. If I look at a picture, I am biased to be a viewer, though in a sense I can and do repaint my internal representation. If I play with a reactive and adaptive environment, I can alternate the roles of painter and viewer at will. Whether there is virtue in this, I do not know. But there might be." So, the cybernetic stance invites both a change in the nature of art objects and, once more, a shift in the power relation between artist and audience, somehow entraining the audience in their production and evolution, as we also saw in the previous chapter in the case of Brian Eno. In the Musicolour performances at Churchill's Club, for example, "we also used the system when people were dancing and discovered that in these circumstances the audience can participate in the performer-machine feedback loop just because they are doing something to music and the band is responding to them" (88), though this turned out not to be the case in the larger setting of the Streatham Locarno.

  The Social Basis of Pask's Cybernetics

  It is clear that the social dynamics of Pask's formative venture into cybernetics bears much the same marks as the others discussed in earlier chapters. There is, first of all, the undisciplined mode of transmission of cybernetics. Pask did not train to be a cybernetician by enrolling in any disciplinary program; instead, a chance meeting with Norbert Wiener served, as we saw, to crystallize Pask's agenda, an agenda that already existed, though in a relatively formless state. Second, as we have also seen, Pask's first project as a cybernetician was undertaken in an undisciplined space outside any conventional institutional structure—he built the first Musicolour machine in his rooms at Cambridge, out of the detritus of war and a technological society. I mentioned bits of Calliope organs and bomb sight computers earlier; Elizabeth Pask (n.d.) recalled that Gordon and Harry Moore built Musicolour from "old relays and uniselectors junked from post office telephone exchanges"—the same components that Walter and Ashby used in their model brains. One could speak here of a lack of material discipline as well as social discipline. Like our other cyberneticians, then, Pask's cybernetics bubbled up outside the normal channels of society. And along with this undisciplined aspect went the protean quality of Pask's cybernetics: Pask was entirely free to follow his own inclinations in developing his cybernetics in a theatrical direction, a more or less unprecedented development.13 At the same time, this lack of disciplinary control helps to account for another aspect of the novel form of Pask's cybernetics—his abandonment, already in the early 1950s, of the idea that cybernetic systems seek by definition to pursue fixed goals.

  One can think along much the same lines about the fate of Musicolour itself. Pask's recollection, quoted above, that "we . . . tried to sell it in any possible way: at one extreme as a pure art form, at the other as an attachment for juke boxes," goes to the heart of the matter. It was not clear what Musicolour was. It did not fit well into the usual classification of material objects. It had something to do with music, but it wasn't a musical instrument. It drove a light show, but it wasn't just lighting. It was an object, but a pretty ugly one, not an art object in itself. One could say that Musicolour was itself an undisciplined machine, incommensurable with conventional forms of entertainment, and the different modes of presentation and venue that Pask and his friends explored in the 1950s have to be seen as a form of experimentation, trying to find Musicolour a niche in the world. In the end, as we have seen, the world proved recalcitrant, and, like that other odd art object, Gysin's Dreamachine, Musicolour was a commercial failure. Mention of the Dreamachine perhaps reminds us that in the later sixties light shows of all sorts—not all using strobes, and some very reminiscent of Musicolour displays—were de rigueur. But by that time Musicolour had been forgotten and Pask had moved on to other projects. One can only wonder what the Grateful Dead might have got out of one of Pask's devices.

  And finally, Pask himself. One should probably understand the boy who built bombs and said that school taught him to be a gangster as someone who enjoyed a lack of discipline—not as someone forced into the margins of society, but who sought them out. No doubt for Pask much of the attraction of Musicolour and cybernetics in general lay in their undisciplined marginality. And this, in turn, helps us to understand his post-Cambridge career, based in a private company, System Research, free from any demands, except that of somehow improvising a living. Now we can pick up the historical thread again.

  Training Machines

  Pask did not lose interest in adaptive machinery after Musicolour, but he had the idea of putting it to a different and more prosaic use, returning to his formative interest in learning. In the mid-1950s, "there was great demand in the commercial world for keyboard operators, both for punch card machines and typing," and Pask set out to construct an adaptive keyboard trainer. He later recalled that the "first Self Adaptive Keyboard Trainer (SAKI) was constructed in 1956 by myself and prototyped by Robin McKinnon-Wood and me" (Pask 1982, 69; see fig. 7.5). This was displayed at the Inventors and Patentees Exhibition at the Horticultural Hall in London, a meeting also regularly frequented by one Christopher Bailey, technical director of the Solartron Electronic Group, who had "from time to time, made valuable contacts with the less loony exhibitors." Pask got into conversation with Bailey about Grey Walter's robot tortoises, and Bailey in turn proposed that Solartron, which was already expanding into the area of AI, should support the development of adaptive training machines by Pask and System Research (E. Pask n.d.; McKinnon- Wood 1993). Thereafter (Pask 1982, 69, 72),

  Figure 7.5.SAKI. Source: Pask 1961, pl. II, facing p. 33. (Reproduced by permission of Amanda Heitler.)

  Bailey participated in the design and development of this and other systems; notably: EUCRATES [figs. 7.6, 7.7] a hybrid training machine and trainee simulation; a device for training assembly line tasks; a radar simulation training machine and several devices for interpolating adaptively modulated alerting signals into a system, depending upon lapse of attention. The acronym SAKI stood, after that, for Solartron Adaptive Keyboard Instructor and a number of these were built and marketed. Details can be found in a U.K. Patent granted in 1961, number 15494/56. The machine described is simply representative of the less complex devices which were, in fact, custom-built in small batches for different kinds of key boards (full scale, special and so on). The patent covered, also, more complex devices like EUCRATES. . . . In 1961 the manufacturing rights for machines covered by these patents were obtained by Cybernetic Developments: about 50 keyboard machines were leased and sold.

  A genealogical relationship leading from Musicolour to SAKI and Eucrates is evident.14 Just as Musicolour got bored with musicians and urged them on to novel endeavors, so these later machines responded to the performance of the trainee, speeding up or slowing down in response to the trainee's emergent performance, identifying weaknesses and harping upon them, while progressively moving to harder exercises when the easier ones had been mastered.15 Stafford Beer tried Eucrates out in 1958 and recorded, "I began in total ignorance of the punch. Forty-five minutes later I was punching at the rate of eight keys a second: as fast as an experienced punching girl" (Beer 1959, 125).16 SAKI was an analog machine; like Musicolour, its key adaptive components were the usual uniselectors, relays, and capacitors. Later versions used microprocessors and were marketed by System Research Developments (Sales). SAKI itself formed the basis for the Mavis Beacon typing trainer, widely available as PC software today.

  The link to Solartron in the development of adaptive training machines was very consequential for Pask and System Research. Much of Pask's paid work from the late 1950s onward centered on the development of adaptive teaching and training machines and information systems, as discussed further below. And again this episode illustrates so
me of the social oddity of cybernetics—a chance meeting at an exhibition, rather than any more formalized encounter, and, again, the protean quality of cybernetics, as a peculiar artwork metamorphosed easily into a device for teaching people to type. From an ontological point of view, we can see machines like SAKI as ontological theater much like Musicolour, still featuring a dance of agency between trainee and machine, though now the dance had been domesticated to fit into a market niche—the machine did now have a predetermined goal: to help the human to learn to type efficiently—though the path to the goal remained open ended (like coupled homeostats searching for equilibrium). Interestingly from an ontological angle, Beer recorded witnessing a version of the Turing test carried out with Eucrates. As articulated by Turing, this test relates to machine intelligence: the textual responses of an intelligent machine would be indistinguishable from those of a human being. Pask's demonstrations with Eucrates were a performative rather than representational version of this: "So my very first exposure to Gordon's science was when he sat me in a room with a monitor, in the capacity of metaobserver, and invited me to determine which screen was being driven by the human and which by Eucrates. It was impossible. The behaviour of the two elements was converging, and each was moving towards the other" (S. Beer 2001, 552).

  Figure 7.6.Eucrates. Source: Pask 1961, pl. I, facing p. 32. (Reproduced by permission of Amanda Heitler.)

  Figure 7.7.Eucrates: detail showing settings. (© 2002 by Paul Pangaro.)

  Teaching Machines

  The machines we have discussed so far—Musicolour, SAKI, Eucrates—were all directly concerned with performance and nonverbal skills. In the 1970s, however, Pask turned his attention to education more generally and to machines that could support and foster the transmission of representational knowledge. CASTE (for Course Assembly System and Tutorial Environment) was the first example of such a machine, constructed in the early 1970s by Pask and Bernard Scott (who completed a PhD with Pask at Brunel University in 1976; see fig. 7.8). Figure 7.9 shows a more sophisticated version, Thoughtsticker, from around 1977.17

  There is no doubt that Pask was extremely interested in these machines and the overall project in which they served as markers, or of their worldly importance in sustaining him and System Research, but I am going to give them relatively short shrift here. The reason for this is that much of the technical development in this area circled around the problematic of representing articulated knowledge within a machine, and that is not a key problematic of this book. Pask quickly arrived at the position that bodies of knowledge consist of sets of concepts related to one another in what he called an "entailment mesh" (Pask and Scott 1973) an idea which was itself central to what he called "Conversation Theory" and later to his "Interaction of Actors" theory (Pask 1992). I leave it to others to explore the historical evolution of this aspect of Pask's work, set out in three books (Pask 1975a, 1975b, 1976a) and many papers, all of which are, in Bernard Scott's words (2001a, 2), "notoriously difficult" to read.18

  Figure 7.8.Intuition, a portable version of CASTE. Source: G. Pask, "Conversational Techniques in the Study and Practice of Education," British Journal of Educational Psychology, 46 (1976), 24, fig. 3. (Reproduced with permission from the British Journal of Educational Psychology. © The British Psychologcial Society.)

  Figure I. An epistemological laboratory. The figure in placed first to give an idea of how empirical results are obtained but many of the labels will not be intelligible until, at slater stage in the paper, the functioning of the equipment and the types of man machine transactions are spelled out. The key is as follows:

  A. Random access slide projector with control keyboard, for displaying slide mounted graphics.

  B. Entailment mesh display with overlay multi sheets, containing 60 mode positions and 4 × 60 independently addressed coloured signal lamps and touch sensors.

  C. Tutorial mode keyboard with special function keys.

  D. Course assembly mode keyboard with special function keys.

  E. ARDS graphic display tubes with control unit and keyboard used for displaying "pruned" meshes.

  F. Video display units with control keyboards used for topic text input-output.

  G. Pigeon holes filling system with slots for 60 files and containing 3 × 60 independently addressed signal lamps and 60 sensors.

  H. Dual drive floppy disk unit.

  I. CAI 32k computer.

  J. Digital Cassette unit used as mass storage device.

  K. ASR 33 teletype used for "hard copy" output.

  L. Electronics rack, containing special electronics and system interface.

  Figure 7.9.Thoughtsticker. Source: G. Pask, "Organizational Closure of Potentially Conscious Systems," in M. Zeleny (ed.), Autopoiesis: A Theory of the Living (New York: North Holland, 1981), 269, fig. 1.

  Here, then, I will confine myself to a few comments. First, CASTE and its descendants remained clearly within Musicolour's lineage. CASTE was another aesthetically potent environment with which the student could interact, exploring different routes around the entailment mesh appropriate to this subject matter or that, being invited to carry out the relevant performative exercises, with apparatus at the laboratory bench if relevant, responding to queries from the machine, and so on. Again we find the performative epistemology that I have associated with cybernetics in the earlier chapters—of articulated knowledge and understanding as part and parcel of a field of performances. Thoughtsticker went so far as to interact with the subject-experts who fed in the contents of entailment meshes—generalizing aspects of the mesh in various ways, for example, and checking back about the acceptability of these generalizations. In this version, the very content of the mesh was, to a degree, a joint product of the human and the machine, as I said earlier about a Musicolour performance.

  Second, we can note that, again like Musicolour, Pask's teaching machines also functioned as experimental setups for scientific research, now in the field of educational psychology. Pask and his collaborators toured schools and colleges, including Henley Grammar School and the Architectural Association, with the portable version of CASTE and made observations on how different learners came to grips with them, echoing Pask's observations in the fifties on Musicolour. These observations led him to distinguish two distinct styles of learning—labeled "serial" and "holist" in relation to how people traveled around the mesh—and to argue that different forms of pedagogy were appropriate to each.

  Third, we can return to the social basis of Pask's cybernetics. The very ease with which one can skip from a general sense of "adaptation" to a specific sense of "learning" as manifested in schools and universities suggests that education, if anywhere, is a site at which cybernetics stood a good chance of latching onto more mainstream institutions, and so it proved for Pask. In 1982, Pask and Susan Curran (1982, 164) recorded that "over a period of 11 years (five of overlapping projects and a six-year research project) the Social Science Research Council in Britain supported a study by System Research . . . on learning and knowledge." Around the same time, Bernard Scott (1982, 480) noted that "despite the painstaking way in which Pask prepared the ground for the theory's presentation [conversation theory], it is fair to say that it has not won general acceptance in psychology. The dominant attitudes were too strong." But nevertheless, "the theory (through its applications) has had far greater impact in educational circles and is recognised, internationally, as a major contribution to educational praxis" (on which see B. Scott 2001b).

  Pask was integrally involved, for example, in pedagogical innovation at Britain's new Open University (OU). David Hawkridge joined the OU as head of their Institute of Educational Technology in 1969. He knew of Pask's work on SAKI and hired Brian Lewis who had worked at System Research as his deputy (Hawkridge 2001, 688–90):

  [Lewis] lost no time in introducing me and other Institute staff to Gordon, his great friend and former boss. In fact, quite a few of us went on expeditions down to System Research Ltd in Richmond in the early 197
0s, and came back bemused and sometimes confused. What was Pask up to and could it be turned to advantage in the OU? I suggested to Brian that we should ask Pask to be our Visiting Professor (part-time). That would regularise the Richmond visits, and Brian said he thought Gordon would be delighted to give a few seminars at the Institute. I had little idea what these might involve, but the Institute had just won a large grant from the Ford Foundation, and Gordon's appointment (and, indeed, Bernard Scott's as a consultant) seemed entirely appropriate. He was the pre-eminent British scholar in our field. . . . It was probably Brian who suggested to Gordon that there was a DSc to be had from the OU if only he would submit all his publications. Finding sufficiently knowledgeable referees for his case was not easy, but I had the pleasure of seeing him receive the award to great applause. I think he was delighted with it, and with the robes.19

  There is much more that could be said on Pask's teaching machines, but I want to close this section by noting that his work on these devices led him to a distinctive general perspective on mind.20 In earlier chapters we saw how, in different ways, cybernetics shaded into Eastern philosophy and spirituality. None of this figures prominently in Pask's work, nor does religion in general (though he did convert to Catholicism shortly before his death: Amanda Heitler, personal communication). But a 1977 essay, "Minds and Media in Education and Entertainment," is worth examining from this perspective. Here the initial referent of "mind" was the human mind, and the "media" were the usual means of communication between minds: speech, texts, information systems like CASTE. But (Pask 1977, 40)