This brings us to the third community with which Walter engaged, the nascent community of cyberneticians in Britain. The 1948 publication of Wiener's Cyberneticsboth put the word itself into circulation in Britain and helped crystallize the formation of a self-consciously cybernetic community there. On 27 July 1949 John Bates of the Neurological Research Institute of the National Hospital in London wrote to Walter as follows:
Dear Grey,
I have been having a lot of "Cybernetic" discussions during the past few weeks here and in Cambridge during a Symposium on Animal Behaviour Mechanisms, and it is quite clear that there is a need for the creation of an environment in which these subjects can be discussed freely. It seems that the essentials are a closed and limited membership and a post-prandial situation, in fact a dining-club in which conventional scientific criteria are eschewed. I know personally about 15 people who had Wiener's ideas before Wiener's book appeared and who are more or less concerned with them in their present work and who I think would come. . . .
Besides yourself, Ashby [see the next chapter] and Shipton [Walter's colleague and collaborator at the Burden], and Dawson and Morton from here, I suggest the following:-
Mackay - computing machines, Kings Coll. Strand.
Barlow - sensory physiologist—Adrian's lab.
Hick - Psychological lab. Camb.
Scholl - Statistical Neurohistologist—U.C. Anat. lab.
Uttley - ex Psychologist, Radar etc. T.R.E.
Gold - ex radar zoologists at Cambridge
Pringle
I could suggest others but this makes 13. I would suggest a few more non neurophysiologists communications or servo folk of the right sort to complete the party but those I know well are a little too senior and serious for the sort of gathering I have in mind.
We might meet say once a quarter and limit the inclusive cost to 5/- less drinks. Have you any reactions? I have approached all the above list save Uttley so far, and they support the general idea.
Walter replied the next day to this "exciting letter"—"We also have been having some pretty free CYBERNETIC discussions and your notion of a sort of Dining Club attracts me very much. I agree that it will be nice to keep the gathering rather small, about the size of a witches coven owing to the shortage of broomsticks." Walter also mentioned that Warren McCulloch was visiting Britain in September 1949 and suggested that this would provide a good occasion for the first meeting of the group.26 And thus it came to pass. McCulloch addressed the first meeting of the Ratio Club on 14 September 1949 on the topic of "Finality and Form in Nervous Activity." Sixteen member were present, including Ashby but not Walter, "owing to the delivery of a male homeostat which I was anxious to get into commission as soon as possible." Expenditure on food was £1-4-0; on beer and wine, £7. Thereafter, the club met at least thirty-four times up to 1955 (with decreasing frequency after 1952) before being wound up at a reunion meeting on 27 November 1958.27
There is much that might be said on the Ratio Club, its membership, and their doings, but this would easily carry us too far afield, and I will confine myself to a few observations.28 We should note first Ratio's interdisciplinar ity. Bates described its proposed membership as "half primarily physiologists though with 'electrical leanings' and half communication theory and ex-radar folk with biological leanings" and, later, to Turing, as "half biologists—(mostly neurophysiologists) and half engineers and mathematicians," while remarking to himself that the club was "incomplete—no sociologists, northeners, professors" (Clark 2002, 78–80).29 But beyond that, Ratio was interinstitutional, as one might say. It did not simply elide disciplinary boundaries within the university; it brought together representatives from different sorts of institutions: people from the universities, but also medical men and physiologists based in hospitals and research institutes, including Walter and Ashby, and workers in government laboratories (Albert Uttley at the Telecommunications Research Establishment, the TRE).30 The Ratio Club was the center of gravity for work in cybernetics in Britain from 1949 to the mid-1950s, and it existed transversely,or orthogonally, to the usual institutions for the production of knowledge, cutting across not just academic disciplinary boundaries, but also across the usual institutional classifications, too. And this transversality continued to be a conspicuous feature of British and European cybernetics after the demise of Ratio, when the series of Namur conferences became the key institutional venue from 1956 onward.31
Two observations follow. First, ontology and sociology were entangled here. This transverse crystallization had the character of a purification that was at once social and ontological. From the side of traditional fields of practice, it would be a mistake to think that an interest in the adaptive brain was actively excluded. But the formation of first the Ratio Club and then the Namur conference series attests to a perceived marginality of the cyberneticians in their own fields, and a perceived closeness to workers in other fields with similar interests. From the other side, the shared interest in the adaptive brain came to center precisely on transverse institutions like the Ratio Club. Ratio— rather than their home disciplines and institutions—was where people like Walter found an active and engaged audience for their cybernetics. And, as we will see later, much of the propagation of cybernetics up the present has continued to be located in such strange antidisciplinary and interinstitutional spaces, even as the range of cybernetics has gone far beyond the brain.
My second observation is this. The Ratio Club and its successor institutions were undoubtedly successful in maintaining the postwar cybernetic ferment, but they were conspicuously lacking in the means of social reproduction. The Ratio Club had no mechanism for training students: a dining club does not grant PhD's. Among our cyberneticians, only Stafford Beer in the second generation seems to have taken this problem seriously, but we can note now that this ad hoc organization contributed importantly to the way cybernetics evolved. Academic disciplines are very good at holding neophytes to specific disciplinary agendas, and it was both a strength and a weakness of cybernetics that it could not do this—a strength, inasmuch as cybernetics retained an undisciplined and open-ended vitality, an ability to sprout off in all sorts of new directions, that the established disciplines often lack; a weakness, as an inability both to impose standards on research and to establish career paths for new cyberneticians left enthusiasts to improvise careers much as did the founders.
These remarks return us to a topic broached above. Popular writing and, in Walter's case especially, public performances assumed an importance in the propagation of cybernetics that one does not find in established fields. In doing the research for this book I have been surprised to discover just how many first and consequential contacts with cybernetics have been with popular books, articles and performances. We just saw that Wiener's Cybernetics was central to the crystallization of the British cybernetics community, and Beer fell into cybernetics after reading the same book. Walter's cybernetics traveled and mutated along the same lines. In chapter 7 we can discuss the adaptive architecture of John Frazer, who tried to build his own robots after seeing a display of the tortoises as a schoolboy, before falling in with Pask (who declared himself a cybernetician after meeting Wiener in person as an undergraduate). Later in this chapter, we can see how William Burroughs laundered elements of cybernetics into the counterculture after reading The Living Brain. And in the following section I want to bring the discussion of robotics up to the present by focusing on another Living Brain reader, Rodney Brooks.32 The general point to note here, however, is that the propagation of cybernetics was indeed both unsystematic and undisciplined. Walter's cybernetics was addressed to the brain, but Brooks understood it as robotics, Frazer took it into architecture, and Burroughs transplanted it into the domain of altered states and that classic sixties project, the exploration of consciousness. Hence the protean quality of cybernetics, with individuals free to adapt it to their own interests and obsessions, unconstrained by disciplinary policing.33
Ro
dney Brooks and Robotics
Rodney Brooks is currently director of the MIT Computer Science and Artificial Intelligence Laboratory, Panasonic Professor of Robotics at MIT, and past chairman and now chief technical officer of iRobot Corporation.34 Brooks began his career in robotics as a schoolboy in Australia when "I came across a Pelican edition of Grey Walter's book, and tried to build my own version of Machina Speculatrix, using transistor technology rather than vacuum tubes. . . . The subtleties of the original electronics were a little beyond me, but I did manage to get my first robot, Norman, to the point where it could wander around the floor, respond to light, and bumble its way around obstacles" (Brooks 2002, 27). From Australia he moved to the United States, completed a PhD in computer science at Stanford University in 1981, and held postdoctoral positions at Carnegie Mellon University and MIT and a faculty position at Stanford, before rejoining MIT as an assistant professor in 1984. The first machine that Brooks and a few collaborators then constructed was a robot called Allen, which made Brooks's reputation, in certain quarters at least, and began his rise to his current position as leader of one of the most important computer science and AI laboratories in the world. And the point to grasp is that Allen was very much an updated version of the tortoise. Using a ring of twelve sonar range detectors in place of the tortoise's photocell and contact switch, and solid-state logic elements instead of electronic valves, Allen would explore its environment, pursuing goals (such as finding and trying to get to the most distant part of the room) and avoiding obstacles along the way. Even Brooks's construction strategy, which he called a "subsumption architecture," getting different layers of the control system working one after the other, mirrored Walter's transit from the tortoise itself to CORA (see below).35
So, if one is looking for a "weighty" answer to the question, what happened to cybernetics? one answer would be: it is alive and well in Brooks's lab at MIT. But then another question arises. How on earth could one make a reputation in computer science by building an updated tortoise thirty-six years after Walter? Of course, Brooks displayed his own originality, but some important history is needed here, which I want just to mention without going into detail. In the opening chapter I contrasted the performative brain of cybernetics with the representational one of AI, and I need to say a little about the development of the latter field.
The canonical history of AI dates its self-conscious inception to the sixweek workshop "Artificial Intelligence" at Dartmouth College organized by John McCarthy in 1956 (Brooks 2002, 21–31). Many of the principles of the nascent field were present, and what followed was a rapid purification, as I called it above, but going in the opposite direction. From World War II to the mid-1950s speculation about the mind in terms of machine models was an exceptionally rich, diverse, and fascinating field, in which cybernetics in many ways took the lead. From the mid-1950s onward a representationalist strand of AI came to the fore, and it achieved institutional dominance within the space of about ten years. In GOFAI—good, old-fashioned AI—the aim was to mimic mental performances. Alan Newell and Herbert Simon's Logic Theorist program was an early landmark, and it was a program that mimicked the proofs to be found in Bertrand Russell and Alfred North Whitehead's Principia Mathematica. In robotics this translated into the problematic of generating computer representations (maps, models) of environments and operating on them to execute plans, such as moving from A to B while avoiding obstacles. This style of AI and robotics, then, can stand as a piece of ontological theater for the other ontology from that of cybernetics, the modern ontology of knowability. AI robots sought to know their worlds substantively, and to accomplish their goals through that knowledge. AI robotics was the other to Walter-style robotics.
Historically, representational, or symbolic, AI quickly became the dominant paradigm in the universities, largely displacing cybernetics from its already tenuous foothold, not only from computer science departments and their ilk, but from social science departments, too, in the so-called cognitive revolution, in which human mental powers were conceived by analogy to digital computers as information processors (Gardner 1987). Of course, the rise of AI and the associated "cognitive sciences" is an immense historical story in itself, but let me just comment briefly. How did AI come to exert such a fascination over the academic and popular imagination? Part of the answer must lie in its very familiar ontology. It is easy to think of the brain and mind as the organ of knowledge, and AI thus conceived presents a straightforward problem of mimicking very familiar (especially to academics) mental performances. At the same time, AI was uniquely associated with digital computers and their programming and thus fitted very naturally into the agenda of novel postwar departments of computer science (unlike the odd machines of Walter et al.). And third, the military bought it. Almost all the funding for AI research was provided by the U.S. military, and almost all of that went to research in symbolic AI (Edwards 1996).36
Cybernetics thus lost much of its social basis in the universities from the mid-1950s onward; the cyberneticians became even more marginal there than they had been before—which is another kind of answer to the question, what happened to cybernetics? But this gets us back to the story of Rodney Brooks. In robotics, symbolic AI promised much but never quite delivered. Machines were never quite fast enough to accomplish real-time control.37In his first years in the United States, Brooks worked within this tradition, focusing on computer models of environments, but became increasingly frustrated with it. In the late 1970s at Stanford, he helped Hans Moravec, a future leader in AI-style robotics, on a robot which moved so slowly (due to the time taken for computation) that, outdoors, the movement of sun and shadows would confuse its internal repesentations (Brooks 2002, 30):
Despite the serious intent of the project, I could not but help feeling disappointed. Grey Walter had been able to get his tortoises to operate autonomously for hours on end, moving about and interacting with a dynamically changing world and with each other. His robots were constructed from parts costing a few tens of dollars. Here at the center of high technology, a robot relying on millions of dollars of equipment did not appear to operate nearly as well. Internally it was doing much more than Grey Walter's tortoises had ever done—it was building accurate three-dimensional models of the world and formulating detailed plans within those models. But to an external observer all that internal cogitation was hardly worth it.
It was against this background that Brooks's 1985 robot, Allen, stood out as a revolutionary alternative. Allen dispensed with the "cognitive box" (Brooks 2002, 36) that was the hallmark and center of attention in contemporary robotics in favor of the performative and adaptive engagement with the environment that was the hallmark of the tortoises.38 This, of course, put him on the wrong side of the law as far as the academic establishment was concerned, and he has repeatedly told the story of how, during his first scholarly presentation of his new approach, one senior computer scientist whispered to another, "Why is this young man throwing away his career?" Three referees unanimously recommended rejection of his first paper on this approach—though it was published anyway (Brooks 1999 [1986]) and went on to become "one of the most highly cited papers in all of robotics and computer science" (Brooks 2002, 43). In the event, though the "arguments . . . continue to today" (Brooks 2002, 43), Brooks's approach did succeed in redirecting the work of a substantial fraction of the robotic community back into Walterian, cybernetic channels. One token of this success came in 2002, with the organization of a major international conference, "Biologically-Inspired Robotics," held at the Hewlett-Packard Laboratory near Bristol and close to the Burden Institute. Marking the twenty-fifth anniversary of Walter's death, the subtitle of the conference was simply "The Legacy of W. Grey Walter." Many of the principals of this "new" field" gave invited addresses, and graduate students presented an impressive array of talks.39
After a decades-long hiatus, then, at Brooks's lab at MIT, and many other academic centers, too, the robotic wing of cybernetics finally gained what
it conspicuously lacked in its formative years, a solid institutional base not only for research but also for social reproduction, the training of graduate students as future researchers with a prospect of recognizable career paths in the field.40 And one concluding remark is worth making for future reference. In the following chapters we will encounter many imaginative initiatives in cybernetics which eventually fizzled out, and one inevitably wonders whether this points to some essential flaw in cybernetics itself. I think we should remember that Walter's robotics once fizzled out, but that in retrospect it is clear that the fizzling had more to do with the lack of an institutional base and support than any inherent flaws.41
CORA and Machina docilis
AFTER FOUR YEARS [IN CAMBRIDGE] SPENT LITERALLY IN A CAGE AND CHAINED BY THE ANKLE—NOT FOR PUNISHMENT BUT FOR ELECTRICAL SCREENING—. . . IMAGINE, THEN, HOW REFRESHING AND TANTALIZING WERE THE RESULTS FROM PAVLOV'S LABORATORY IN LENINGRAD TO THOSE ENGAGED ON THE METICULOUS DISSECTION OF INVISIBLE NERVE TENDRILS AND THE ANALYSIS OF THE IMPULSES WHICH WE INDUCED THEM TO TRANSMIT.
The Cybernetic Brain Page 8