Affect and Artificial Intelligence

Elizabeth A. Wilson

• Publisher: University of Washington Press
• Year: 2011
• Print ISBN: 9780295990477
• eBook ISBN: 9780295800004
• Topic: Computer Science
• Subtopic: Computer Science General
• Index: Computer Science

In 1950, Alan Turing, the British mathematician, cryptographer, and computer pioneer, looked to the future: now that the conceptual and technical parameters for electronic brains had been established, what kind of intelligence could be built? Should machine intelligence mimic the abstract thinking of a chess player or should it be more like the developing mind of a child? Should an intelligent agent only think, or should it also learn, feel, and grow? Affect and Artificial Intelligence is the first in-depth analysis of affect and intersubjectivity in the computational sciences. Elizabeth Wilson makes use of archival and unpublished material from the early years of AI (1945–70) until the present to show that early researchers were more engaged with questions of emotion than many commentators have assumed. She documents how affectivity was managed in the canonical works of Walter Pitts in the 1940s and Turing in the 1950s, in projects from the 1960s that injected artificial agents into psychotherapeutic encounters, in chess-playing machines from the 1940s to the present, and in the Kismet (sociable robotics) project at MIT in the 1990s.

1 / THE POSITIVE AFFECTS OF ALAN TURING

The interrelationships between the affect of interest and the functions of thought and memory are so extensive that absence of the affective support of interest would jeopardize intellectual development no less than destruction of brain tissue. To think, as to engage in any other human activity, one must care, one must be excited, must be continually rewarded. There is no human competence which can be achieved in the absence of a sustaining interest.

—Silvan Tomkins

IN NOVEMBER 1946, ALAN TURING WAS IN CORRESPONDENCE WITH THE British neurologist and cyberneticist W. Ross Ashby about the possibility of making mechanical models of the brain. ¹ Turing had recently taken a position at the National Physics Laboratory (NPL), where he was part of a team working on the construction of the Automatic Calculating Engine (ACE), one of the first digital computing machines to be built in Britain. ² Turing didn't last long in this job. A comment to Ashby about the orientation of his interests at the NPL holds a clue to the brevity of his tenure: “In working on the ACE I am more interested in the possibility of producing models of the action of the brain than in the practical applications to computing.” ³ It soon became evident that Turing's vision for the ACE, emerging from an idiosyncratic set of interests and enthusiasms, was incompatible with the institutional constraints at the NPL. While Turing understood himself to be engaged in modeling the brain, the NPL was insisting on a more pragmatic, less fanciful orientation to computational machines. In truth, the NPL wanted only one sliver of Turing's expertise: mathematical logic. His ambition to be involved in the actual building of the ACE, soldering iron in hand, was frustrated when the engineering work was contracted out to a separate organization, and his growing interest in biology and neurology was considered peripheral to the core tasks of the ACE project.

By mid-1947 Turing was “thoroughly disheartened” (Copeland 2004a, 400). He left the NPL for a one-year sabbatical at King's College, Cambridge, to pursue his ideas about biology and computation. Sir Charles Darwin, the NPL director, explained the rationale for Turing's leave:

He wants to extend his work on the machine still further towards the biological side. I can best describe it by saying that hitherto the machine has been planned for work equivalent to that of the lower parts of the brain, and he wants to see how much a machine can do for the higher ones; for example, could a machine be made that could learn by experience? This will be theoretical work, and better done away from here. ⁴

As I have argued in the Introduction, an environment that safeguards conventional intellectual divisions (engineering not theory, logic not fancy, head not hands) was ill-suited for Turing's talents and temperament. An environment that can plainly articulate what work ought to be done here and what work ought to be done elsewhere would be unable to sustain Turing's diverse interests. One of the characteristics of Turing's work is that it was frequently at odds with conventional thinking about digital computers, even as it was setting the standards for programming and design. Maurice Wilkes, a pioneer of computing science at the University of Cambridge, gave the following assessment of Turing's work on the ACE:

Turing began his lectures in London on 12 December [1946] and I was there… . I found Turing very opinionated and considered that many of his ideas were widely at variance with what the main stream of computer development was going to be. I may have gone to his second lecture but I certainly went to no more. [Douglas] Hartree continued to go and insisted on giving me his notes, but I found them of little interest. ⁵

Notwithstanding his variance from the mainstream, Turing's name was later attached to the award for career achievement in computing from the Association for Computing Machinery (ACM). Lauded by the ACM as the Nobel Prize of computing, the A. M. Turing Award is now sponsored by the computational behemoths Intel and Google and is worth $250,000. Despite his lack of interest in Turing's London lectures, Wilkes was the second recipient of the A. M. Turing Award, in 1967. In his award lecture, Wilkes notes:

The programming system that [Turing] devised for the pioneering computer at Manchester University was bizarre in the extreme. He had a very nimble brain himself and saw no need to make concessions to those less well-endowed. I remember that he decided … that the proper way to write binary numbers was backwards, with the least significant digit on the left. He would, on occasion, carry this over into decimal notation. I well remember that once, during a lecture, when he was multiplying some decimal numbers together on the blackboard to illustrate a point about checking a program, we were all unable to follow his working until we realized that he had written the numbers backwards. (Wilkes 1967, 201)

The difference between the mainstream and the irregular, between here and elsewhere, between forwards and backwards is often opaque in Turing's work—and usefully so. An eccentric set of intellectual and affective concerns shapes his canonical 1950 paper on machine intelligence. In this chapter, I will argue that it is his errant curiosity, his capacity for enjoyment and surprise, and his childish engagement with computational machinery that underwrite the importance of the 1950 paper for a contemporary audience. It is my hope that Turing's orientation toward work that is “better done elsewhere” can give us a way of reading traditional AI for its waywardness and peculiarity. Turing can help us locate the twists in AI, from the inside out. This is not an argument for his status as an outsider: the ACM, we must presume, doesn't name awards after those who have no influence. But neither is Turing's eccentricity simply embroidery on otherwise conventional concerns (he is not, in this sense, the archetypal hero inventor). To the extent that eccentricity and conventionality coexist in Turing, it is because each enables and sustains the other. Each prospers through an engagement with its sibling: eccentricity is cultivated by its proximity to conventionalism, and conventionalism is nourished by eccentricity. It is this coexistence that makes Turing such a compelling figure for AI, and an examination of these tendencies in his work might help us understand how conventional and orthodox ambitions circulate in AI in the present day.

Fortunately, Turing's peculiar, heterodox interests survived his lack of institutional support at the NPL. His formal report to the NPL about his Cambridge sabbatical contained the following:

One way of setting about our task of building a “thinking machine” would be to take a man as a whole and to try to replace all the parts of him by machinery. He would include television cameras, microphones, loudspeakers, wheels and “handling servo-mechanisms” as well as some sort of “electronic brain.” This would of course be a tremendous undertaking. The object if produced by present techniques would be of immense size, even if the “brain” part were stationary and controlled the body from a distance. In order that the machine should have a chance of finding things out for itself it should be allowed to roam the countryside, and the danger to the ordinary citizen would be serious.Moreover even when the facilities mentioned above were provided, the creature would still have no contact with food, sex, sport and many other things of interest to the human being. Thus although this method is probably the “sure” way of producing a thinking machine it seems to be altogether too slow and impracticable. (Turing 1948, 420)

Meltzer and Michie (1969) report that Turing's document caused a “furore” at the NPL; it was claimed by some that “Turing is going to infest the countryside … with a robot which will live on twigs and scrap iron" (n.p.). Yet, as Copeland (2004a) notes, this document also “introduced many of the concepts that were later to become central in the field” (401). Disillusioned by lack of progress in the building of the ACE, Turing did not return to the NPL after his year at Cambridge. In 1948 he accepted a new post at the University of Manchester to work on their automatic digital machine, a position he held until his untimely death in 1954.

Besides his work on computers, Turing's death by suicide at the age of forty-one is one of the things for which he is commonly remembered. Turing's body was found in his bed by his housekeeper on June 8, 1954. He had died from asphyxia due to cyanide poisoning. Half an apple, with several bites taken from it, was found by the bed. The coroner hypothesized that the apple had been used to take away the taste of cyanide. ⁶ Turing had appeared to be in good health and in a stable frame of mind, yet the three years prior to the suicide had been difficult. In particular, he had been convicted of gross indecency (homosexuality) in unfortunate circumstances arising from a burglary of his house and he had undergone twelve months of organotherapy (chemical castration) in lieu of a prison sentence (Hodges 1983). ⁷ Turing had responded to these difficulties with both good humor and distress. In a 1952 letter to his friend Norman Routledge, he wrote:

I am afraid that the following syllogism may be used by some in the future:
Turing believes that machines think
Turing lies with men
Therefore machines do not think
Yours in distress
Alan ⁸

Turing's fear of a hostile conflation of his sexual and intellectual interests was realized at the inquest into his death. The Manchester Guardian reported that the coroner was “forced to the conclusion that it [the death] was a deliberate act for, with a man of that type, one would never know what his mental processes were going to do next.” ⁹ Whether Turing's type was homosexual or intellectual is unclear; and perhaps it is this very confusion (and the insinuation of an identity precariously formed) that makes both tendencies conventionally credible sources of suicidal intent.

I want to pause here for a moment to consider how we might interpret this cluster of events (homosexuality, trial, suicide). Can the affective data of this period provide us with a schema for reflecting on Turing's response to his conviction and punishment? And might this provide a bridge for thinking anew about his work on AI? In a letter to Norman Routledge after the court case, Turing is chirpy:

I have a delightful story to tell you of my adventurous life when next we meet. I've had another round with the gendarmes, and it's positively round II to Turing. Half the police of N. England (by one report) were out searching for a supposed boy friend of mine. It was all a mare's nest: Perfect virtue and chastity had governed all our proceedings. But the poor sweeties never knew this. ¹⁰

Making the best of his enforced celibacy (“Being on probation my shining virtue was terrific, and had to be. If I had so much as parked my bicycle on the wrong side of the road there might have been 12 years for me” ¹¹), Turing seems to enjoy the fact that it is not just he who is pursuing this young man, but half the northern England constabulary as well. Poor sweeties, all. It seems crucial to me that we take into account that Turing countered his legal and medical difficulties with, among other things, enjoyment. Routledge remembers Turing describing “with giggles” the biological effects of the organotherapy (he was growing breasts). Similarly, Shaun Wylie, a mathematical colleague at Bletchley Park, reports that Turing carried round the legal documents relating to his arrest in a folder entitled “Burglary & Buggery.” ¹²

These data suggest that Turing's legal, medical, and governmental problems were caught up with psychological and affective currents that render them more complex than some narratives of persecution suggest. For example, Judith Halbertsam (1991) claims that Turing's “homosexuality made him seem an unfit keeper of state secrets: he was exploitable, fatally flawed, a weak link in the masculinist chain of government and the military. He had a sexual secret that the enemy … could prey upon, and his secret made him incontrovertibly Other" (444). It seems to me, however, that declarations like this need to take into account the ways in which Turing's secret was an open, enjoyable one. His homosexuality allowed him (as it did many other men of his class) to find an intellectual and emotional home in the British intelligentsia, and in places like Bletchley Park, where the wartime code-breaking work was undertaken. Donald Michie (one of Turing's wartime colleagues) reports, for example, that “Bletchley had some flamboyant homosexuals” (Lee and Holtzman 1995, 38). ¹³ Stefan Helmreich's factual slip-"Turing committed suicide in 1954 while in a U.K. prison serving time for being homosexual" (1998, 246)-literalizes this peculiar insistence that Turing was incarcerated and doomed by his sexuality. I am suggesting that more attention be paid to Turing's capacity to divert his legal and medical harassment through delight and giggling, and that we begin to chart how these affective tendencies inflect his work in AI. The capacity of positive affect to make constricted spaces (intellectual, sexual, legal, computational) more expansive is one of Turing's most important legacies to us.

Questions of sexuality (cottaging, heterosexual panic, asexual attachment) will be discussed again in chapter 4. There, the juxtaposition of Walter Pitts's muted affections with Turing's libidinal attachments will allow us to see how different kinds of attachment between men were part of the fabric of early AI innovation. In the meantime, this chapter focuses on the affective currents in one of Turing's most important papers. Concentrating on the period immediately prior to his arrest—a time when the positive affects (interest, enjoyment, surprise) were readily available for intellectual work—I look at the affective character of Turing's canonical meditation on thinking machines.

CURIOSITY

It is often supposed that computers began with heavy arithmetic, and that with this successfully achieved, computer scientists wandered to more ambitious fields. This may be so of others, but is quite untrue of Turing, who had always been concerned with modeling the human mind. (Hodges 1997, 32)

Turing's place in the history of twentieth-century computational science has been cemented by two eponymous contributions: the Turing machine (Turing 1937) and the Turing test (Turing 1950). These mathematical and philosophical milestones bookend a short but diverse career: the publication in 1937 of a precocious and important paper on the Entscheidungsproblem that outlined the logical parameters for a universal calculating machine (the Turing machine); the years spent at Bletchley Park during World War II breaking code and constructing a speech-encryption device; post-war appointments working and consulting with teams building the world's first electronic digital computers; and the publication in 1950 of the massively influential article “Computing Machinery and Intelligence,” which formalized the conditions for evaluating intelligence in a computer (the Turing test). ¹⁴

While Turing is known most simply as an inventor of the computer, a closer examination of his interests indicates a more expansive intellectual life. ¹⁵ As his comment to Ashby and his report to the NPL suggest, he was perhaps less focused on computers than we have presumed and more captivated by the interrelation of certain mathematical, biological, emotional, social, and engineering puzzles. In particular, computational logic, the building of mechanical devices, and fantastic anticipation were always intimately allied for Turing. One aspect of the code-breaking work at Bletchley Park that he had relished was that the institution encouraged the fusion of conceptualization, implementation, and speculation. Turing spent as much time with wires and valves and reverie as he did with mathematical calculation. One of Turing's obituaries notes that he “combined in a rare and remarkable way great powers of abstract reasoning and analysis with a very concrete imagination, and a keen desire to make with his own hands things that would “work'.” ¹⁶ During the post-war years at the NPL and the University of Manchester, when he was employed solely as a computational theoretician, he was consistently frustrated by the institutional distance placed between his theoretical labor and hands-on engineering.

However, even in these emotionally and intellectually restrictive occupations there was a flowering of his curiosity; in particular, his post-war research in computing was amplified by an active interest in both neurology and morphogenesis (the emergence of pattern in developing biological organisms). While some of these ideas found shape in standard academic publication, many others were pursued in less formal ways: he was an occasional member of the London-based Ratio Club that discussed cybernetic research, he had met and become friendly with Claude Shannon while on a research trip to the Bell Laboratories during the war, he attended lectures at the University of Manchester given by Jean Piaget in 1952, he entered into public debate about the future of computers (Turing 1951, 1952), he corresponded with neurologists and biologists, he conducted all manner of chemistry experiments in a makeshift laboratory in his house, and he entered into a Jungian analysis that focused his interest on the interpretation of dreams. It will be the argument of this chapter that any assessment of Turing's computational achievements and historical legacy needs to be located within the broader terrain of these enthusiastic and eccentric concerns.

I will also argue that these concerns bear on, and are fashioned by, psychological questions; particularly questions about affect, and particularly in relation to childhood. Sometime in 1922, whe...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• Contents
• Preface
• Acknowledgments
• Introduction: The Machine Has No Fear
• 1 - The Positive Affects of Alan Turing
• 2 - Shaming AI: Helplessness, Confusion, and Error
• 3 - Artificial Psychotherapy
• 4 - Walter Pitts and the Inhibition of Affect
• Notes
• Appendixes
• References
• Index