In mid May 1951, Alan Turing gave one of his few talks on BBC’s Third Programme. The recorded lecture was entitled, “Can Digital Computers Think?”. By the time of the broadcast, a year had passed since the publication of Turing’s (now) famous Mind-article, “Computing Machinery and Intelligence”, with its thought provoking imitation game (Turing 1950). The BBC program—stored on acetate phonograph discs prior to transmission—was approximately 20 minutes long, and basically followed the arguments Turing had proposed in his earlier article. Computers of his day, in short, could not really think and therefore not be called brains, he argued. But, digital computers had the potential to think and hence in the future be regarded as brains. “I think it is probable for instance that at the end of the century it will be possible to programme a machine to answer questions in such a way that it will be extremely difficult to guess whether the answers are being given by a man or by the machine”, Turing said. He was imagining something like “a viva-voce examination, but with the questions and answers all typewritten in order that we need not consider such irrelevant matters as the faithfulness with which the human voice can be imitated” (Turing 1951).
The irony is that Alan Turing’s own voice is lost to history; there are no known preserved recordings of him. The acetate phonograph discs from 1951 are all gone. The written manuscript of his BBC lecture, however, can be found at the collection of Turing papers held at King’s College in Cambridge—partly available online (Turing Digital Archive 2016). The BBC also made a broadcast transcript, taken from the recording shortly after the programme was aired. As Alan Jones has made clear, Turing’s radio lecture was part of a series the BBC had commissioned under the title “Automatic Calculating Machines”. In five broadcasts, an equal number of British pioneers of computing spoke about their work. The fact that these talks were given by engineers themselves, rather than by journalists or commentators, was “typical of the approach used on the Third Programme”. Naturally, it is also “what makes them particularly interesting as historical sources” (Jones 2004). Then again, Jones was only able to examine surviving texts of these broadcasts. Consequently, there is no way to scrutinze or explore Turing’s oral way of presenting his arguments. His intonation, pitch, modulation etcetera are all lost, and we cannot conceive the way Turing actually spoke. Perhaps, he was simply presenting and talking about his ideas in a normal way. Yet, according to the renowned Turing biographer Andrew Hodges, the producer at BBC had his doubts about Turing’s “talents as a media star”—and particularly so regarding his “hesitant voice” (Alan Turing Internet Scrapbook 2016). The point to be made is that audiovisual sources from the past often tend to be regarded as textual accounts. By and large audiovisual sources have also been used by historians to a way lesser degree than classical (textual) documents. Sometimes—as the case with Turing—archival neglect is the reason, but more often humanistic research traditions stipulate what kind of source material to use. In many ways, however, the same goes within the digital humanities.
In theoretical physics, the concept of fine-tuning refers to circumstances when parameters of a theory (or model) needs to be adjusted in order to agree with observations. In essence, a substantial part of our ongoing Spotify project— were we are repeatedly working with bots—has been about fine tuning the both highly influential and widely criticized classical Turing test. By focusing on the deceptive qualities of technology—particularly regarding the difference between man and machine—a number of the notions proposed in Turing’s essay “Computing machinery and intelligence” have never really lost their relevance. The imitation game, Turing stated in his 1950 essay, “is played with three people, a man (A), a woman (B), and an interrogator©”. The object of the game was for the interrogator to determine “which of the other two is the man and which is the woman”. Already at the beginning of his essay, Turing however, asked what would happen if “a machine takes the part of A in this game?” As N. Kathryn Hayles famously put it, gender hence appeared at the “primal scene” of humans meeting with their potential evolutionary successors, the machines (Hayles 1999). Still, following her interpretation of Turing, the ‘gender’, ‘human’ and ‘machine’ examples were basically meant to prove the same thing. Aware that one of the two participants (separated from one another) was a machine, the human evaluator would simply judge natural language conversation (limited to a text-only channel) between a human and a machine—designed to generate human-like responses. If the evaluator could not reliably tell the machine from human—the machine was said to have passed the test. It might then be termed artificially intelligent.
Towards the end of Turing’s article, the initial question, “Can machines think?” was consequently replaced by another: “Are there imaginable digital computers which would do well in the imitation game?” (Turing 1950). Naturally, Turing thought so—and only 15 years later, the computer scientist Joseph Weizenbaum programmed what is often regarded as the first bot, ELIZA. She (the bot) had two distinguishing features that usually characterize bots: intended functions that the programmer built, and a partial function of algorithms and machine learning abilities responding to input. In his article, “ELIZA—A Computer Program For the Study of Natural Language Communication Between Man And Machine”, Weizenbaum stated that “the program emulated a psychotherapist when responding to written statements and questions posed by a user. It appeared capable of understanding what was said to it and responding intelligently, but in truth it simply followed a pattern matching routine that relied on only understanding a few keywords in each sentence.” (Weizenbaum 1966]). ELIZA was hence a mock psychotherapist—online today it is still possible to interact with her. In 2005, Norbert Landsteiner reconstructed ELIZA through the implemention elizabot.js: “Is something troubeling you?”, the bot always starts by asking. Later, Landsteiner added graphics, real-time text and even speech integration: “E.L.I.Z.A. talking”—complete with both American and British intonation (Landsteiner 2013).
The element of artifice programmed into ELIZA again testifies to the deceptive qualities of technology (which the Turing test underlined). In fact, ever since, fraudulence (in one form or the other) seems to be a distinguished part of an evolving bot culture constantly capitalising on advancements in artificial intelligence. When Weizenbaum decided to name his bot ELIZA, he did so with explicit and ingenious reference to the flower girl, Eliza Doolittle in George Bernard Shaw’s play Pygmalion (1913), as well as—one might assume—to the more recent Hollywood musical adaptation, My Fair Lady (1964). The ancient Pygmalion myth—in Ovid’s poem Metamorphoses, Pygmalion was the sculptor who fell in love with his own statue—has often been artistically deployed to examine human(s) ability to ‘breath life’ into, for example, a man made object. In other words, the myth belongs to the domain of artificial humanity; a copy of something natural, with Shaw’s play acting as an ironic comment on class society. Learning impeccable speech and cultivated behaviour without real understanding (like a bot), his play was about a bet where a phonetics professor claimed that he could train a flower girl (Eliza Doolittle) to pass for a duchess at a garden party. Or as Weizenbaum declared: “Like the Eliza of Pygmalion fame, [ELIZA] can be made to appear even more civilized, the relation of appearance to reality, however, remaining in the domain of the playwright” (Weizenbaum 1966).
Bots appear to be human—which is why they are interesting. Bots give an impression of being able to act as a normal user and/or person. If they could (almost) pass for humans half a century ago (like ELIZA) the possibilities of such intelligent machines (or rather software robots) have since naturally increased. Today, the most sophisticated bots react instantly to public information, like the advanced algorithmic bots on the stock option market. They seems almost like disembodied cyborgs, part human and part automaton. Nevertheless, bot culture, artificial intelligence and ultimately the Turing test has naturally also been criticized. The latter has, for example, been deemed not particularly useful to determine if a machine can think (or not)—most famously so by John Searle in his 1980 article, “Minds, Brains, and Computers”. Via his “thought experiment” with himself “locked in a room and given a large batch of Chinese writing”, such text was to Searle just “so many meaningless squiggles”. As is well known, Searle’s experiment also involved him being given “a second batch of Chinese script together with a set of rules for correlating the second batch with the first batch. The rules are in English, and I understand these rules”, he stated. Via these rules, but without understanding a word (or sign) of Chinese, it would theoretically be possible, Searle argued, to appear fluent in Chinese. Searle, in short, thought of himself as a bot. In fact, in the beginning of his article he made an explicit reference to ELIZA, stating that it could pass the Turing test simply by manipulating symbols of which ‘she’ had no understanding. “My desk adding machine has calculating capacities, but no intentionality”, he summed up. Searle hence tried to show that a computational language system, in fact, could have “input and output capabilities that duplicated those of a native Chinese speaker and still not understand Chinese, regardless of how it was programmed.” Famously, he concluded: “The Turing test is typical of the tradition in being unashamedly behavioristic and operationalistic” (Searle 1980).
Hayles 1999. Hayles, K. How We Became Posthuman: Virtual Bodies in Cybernetics, Literature, and Informatics. Chicago, University of Chicago Press.
Jones 2004. Jones, A. Five 1951 BBC Broadcasts on Automatic Calculating Machines. IEEE Annals of the History of Computing 26(2), 3–15.
Landsteiner 2013. Landsteiner, N. E.L.I.Z.A. Talking. Available at: http://www.masswerk.at/eliza/
Searle 1980. Searle J. Minds, Brains, and Computers. Behavioral and Brain Sciences 3, 417-424.
Turing 1950. Turing, A. Computing Machinery and Intelligence. Mind 49, 433–460. Available at: http://www.csee.umbc.edu/courses/471/papers/turing.pdf
Turing 1951. Turing, A. Can digital computers think?. Annotations of a talk broadcast on BBC Third Programme 15 May. Available at: http://www.turingarchive.org/browse.php/B/5
Turing Digital Archive 2016. Available at: http://www.turingarchive.org/
Weizenbaum 1966. Weizenbaum J. ELIZA—a Computer Program for the Study of Natural Language Communication between Man and Machine. Communications of the ACM (9) 6. Available at: http://web.stanford.edu/class/linguist238/p36-weizenabaum.pdf