What Turing Himself Said About the Imitation Game
The mathematician and cryptanalyst explained his famous test of computer intelligence during two BBC radio broadcasts in the early 1950s
Photo-Illustration: Chad Hagen; Photo: Archivio GBB/Contrasto/Redux
The Imitation Game, the recent biopic about Alan Turing’s efforts to decipher Nazi naval codes, was showered with award nominations. It even won the 2015 Academy Award for Best Adapted Screenplay. One thing it won’t win any awards for, though, is its portrayal of the “imitation game” itself—Turing’s proposed test of machine thinking, which hinges on whether a computer can convincingly imitate a person. The Turing test, as it is now called, doesn’t really feature in the film. (Given that the movie gets so much of the history wrong, perhaps that’s a good thing.)
Turing described his now-famous test in a scholarly article, “Computing Machinery and Intelligence,” which was published in the quarterly journal Mindin 1950. But his last recorded statements on the topic were, in fact, aired on the radio in 1952, in a broadcast that very nearly didn’t happen. What he had to say to the British public, though little known today, reveals quite a bit about his thoughts on the topic.
The story goes like this: In 1950, a BBC producer named Archibald Clow, on the lookout for new talent, met Turing in Manchester. Clow was unimpressed. Writing to Christopher Holme, a top executive at the BBC, he stated, “He undoubtedly has a lively mind, but I am very doubtful about him as a speaker.” Clow reported that Turing seemed to have “a definite hesitation in his speech” and thus would not recommend him.
What’s more, Clow grumbled, Turing “doesn’t seem keen to talk along the lines of his paper in Mind…. He would rather stick to the mechanico-factual side and I have a suspicion that the reaction to the Mind article may have been unfavourable and this at present is making him excessively cautious.”
Fortunately, Holme knew that Clow was not the best judge of broadcasters. The previous year, Clow had advised against using the astronomer Fred Hoyle on account of his strong Yorkshire accent. Hoyle turned out to be one of the BBC’s most successful broadcasters—it was actually in a BBC talk on the origin of the universe that Hoyle coined the term “big bang.”
Holme took just as much notice of Clow’s reservations about Turing: “Turing, who is by training a mathematical logician rather than a scientist, should be given the fullest opportunity to develop his argument in the Programme even if it requires more than one talk,” he wrote back to Clow. While Holme took a more limited view of Turing’s aims in the Mind article than people do today—“Basically Turing does no more than declare his faith in what is in reach of practical achievement in the immediate future”—he nevertheless thought that the ideas in the article “could be made intelligible to a fairly wide audience.”
Holme was right, and in the end Turing recorded two talks, both broadcast on a BBC radio network called the Third Programme, one in May 1951, the other in January 1952.
The Third was a network that couldn’t exist in today’s ratings-conscious climate. Born on the heels of World War II of what lofty BBC types called a “virtually insatiable demand for serious literature and drama, for good music and intelligent discussion,” it was (they said) a chance to “experiment boldly…without regard for…mass-appeal.” According to its first head, George Barnes, the Third’s goal was to provide the intellectual stimulation that had been so lacking during Britain’s many years of war.
The Third was demanding. It expected a listener to “meet the performer half-way by giving his whole attention to what is being broadcast.” This was exactly the kind of outlet Turing needed for his radio debut. And we should be grateful to the BBC executives who engineered the opportunity, because Turing’s radio talks—archived only in the form of transcripts, alas—provide a wonderful window into his thinking.
Photo-Illustration: Chad Hagen; Photo: Heritage Images/Getty Images
Turing’s first broadcast, titled “Can Digital Computers Think?,” was one of a series of five radio talks by early pioneers of computing. The other speakers were Douglas Hartree and Max Newman, both mathematicians, and Maurice Wilkes and Freddie Williams, who were electronics experts.
Like Turing, these men were scarcely household names. All had spent the recent war doing ultrasecret work that, if only they could have spoken about it, would have awed their listeners. The average person would have seen these scientists’ names only in newspaper articles or letters to London’s Times about the new “electronic brains.” Yet it’s no exaggeration to say that Turing and Newman (also one of the leaders in decrypting Nazi communications) contributed at least as much to the defeat of Hitler as any of Churchill’s celebrated generals.
So it was only fitting that these men had a chance to speak to the British public. And fortunately, Holme’s confidence in Turing as a broadcaster was justified. The producer of his first talk said that Turing spoke “quite naturally and unaffectedly.” The BBC then asked him to make a second broadcast, this time as a participant in a symposium titled “Can Automatic Calculating Machines Be Said to Think?” There were three other participants: Newman, philosopher Richard Braithwaite, and neurosurgeon Geoffrey Jefferson, whose views Turing had attacked in his Mind article.
The BBC recorded this symposium in London on 10 January 1952. The company paid Turing’s train fare and hotel expenses, plus a fee of 20 guineas, which is about US $800 in today’s money—probably quite a bit less than the movie’s lead actor, Benedict Cumberbatch, would receive for appearing on a British talk show.
The symposium first aired at 9:35 p.m. on Monday, 14 January, a prime evening slot. In line with the Third’s goal of “presenting the great classical repertoire,” the talk was preceded by French chamber music and followed by a Schubert piano sonata. The actual symposium lasted nearly 45 minutes—four scholarly voices against a background of crackle and hiss. (The Third wasn’t always easy to tune in to: Even the BBC acknowledged that there were “too many places where to hear it is physically impossible or where reception is so bad that listening to it is a penance.”)
Although Turing’s audience was not huge by today’s standards, it was nevertheless large—approximately 100,000 people, according to BBC estimates. The folks who tuned in were likely a more diverse group than you might expect. About a third of them were “working class,” according to a report on the Third’s market. Some probably lived in continental Europe, where the Third was popular. Indeed, Turing would have been audible as far away as Switzerland, even though 30 percent of Britain wouldn’t have been able to hear him clearly or even at all.
How did all these people react to Turing’s radio talks? Mathematicians were as prone to caricature in the 1950s as they are today. “Mathematicians are queer folk,” said the president of the London Mathematical Society on BBC radio a year before Turing’s first broadcast. “Mathematicians live in intellectual isolation,” he told his listeners. “They are usually shy and reserved about their work, and in the company of other learned men they tend to feel isolated and to look upon their studies as an incommunicable and delightful secret.” In some respects, he could have been talking about Turing, who was undoubtedly shy.
The late Sarah Baring, Lord Mountbatten’s goddaughter and one of the female “slaves,” as Turing called the women who assisted the almost entirely male code breakers at the Government Code and Cypher School at Bletchley Park, said that “on the rare occasions when he was spotted like a protected species, he would be shambling down to the canteen in a curious sideways step, his eyes fixed to the ground.” Nevertheless, Turing seemed self-assured in his radio broadcasts, and he addressed his audience directly and in an accessible way.
The public still didn’t have a clue about Bletchley Park’s computing machines. But the postwar electronic brains being built by Turing, Wilkes, Williams, and others were well known, thanks to a blaze of publicity about these superfast machines in the national dailies and even the local papers.
In the growing debate about the possibility of computer intelligence, Turing was almost a lone voice on the side of the machines. His BBC broadcasts gave him a chance to get his viewpoint across to “the man in the street” (Turing’s words), and his scripts reveal him to be a good communicator—a man dedicated to explaining difficult ideas to the public and who did this rather well. The Imitation Game portrays Turing as unable to understand jokes, but his broadcasts display a nice, if dry, sense of humor.
Turing’s 1952 broadcast shreds three modern myths about the Turing test, as my colleague Jack Copeland at the University of Canterbury, in Christchurch, New Zealand, has argued. First, according to many accounts—including Andrew Hodges’s biography of Turing, which inspired the screenplay of The Imitation Game —Turing’s aim was to provide an operational definition of thinking. In fact, he expressed the opposite intention, saying on the radio, “I don’t want to give a definition of thinking,” and adding, “I don’t really see that we need to agree on a definition at all.” Indeed, several listeners polled by the BBC complained that Turing and his fellow speakers had avoided giving a definition.
The second myth is that Turing predicted a machine would pass his test around the beginning of this century. What he actually said on the radio in 1952 was that it would be “at least 100 years” before a machine would stand any chance with (as Newman put it) “no questions barred.”
The third myth is that Turing’s test is flawed, because it can be passed by machines that obviously don’t think—for example, a computer that merely searches through a huge database of ready-made conversations, looking for matching responses to the judge’s questions. But Newman and Turing made it clear in the broadcast that brute-force searches like this might take the computer “thousands of millions of years.”[shortcode ieee-pullquote quote="The computer is "permitted all sorts of tricks so as to appear more man-like"" float="right" expand=1]
During the broadcast, Braithwaite, Jefferson, and Newman obligingly drew Turing out on key points of his proposed test of a machine’s ability to think, including such questions as Is the computer allowed to lie? Turing said: The computer is “permitted all sorts of tricks so as to appear more man-like.” This even gives the machine the option of trying to evade snarly questions by pretending to be a foreigner, with a limited grasp of local culture and vocabulary. (The winner of a recent computer-imitates-human game, which feigned being “Eugene Goostman,” supposedly a 13-year-old Ukrainian boy, did just this—but even then the program did not pass the Turing test, despite much media hype to the contrary.)
Other issues emerged during the discussion, such as What sort of thing is the judge permitted to ask? “Anything,” Turing said and added, “And the questions don’t really have to be questions, any more than questions in a law court are really questions.”
Won’t a computer be easy to detect, since it is infallible? No, because, as Turing said, the computer can use tricks to avoid being spotted. In any case, his view was that “computing machines aren’t really infallible at all.” Turing knew very well that computers make mistakes.
What kind of computer program could do well at this test? Turing’s idea, brilliantly innovative at this time, was that a computer set up in the right way could learn from its teacher and from its experiences, as a human child does. He called these “child-machines” and mentioned his own experiments along these lines.
Won’t a computer stand out as particularly unemotional? Turing suggested that child-machines might even develop humanlike emotions, which he said are “likely to occur as a sort of by-product of genuine teaching.” If he was right about this, computers need not pretend in order to appear emotional.
Isn’t an electronic brain very different from a human brain?The speakers agreed that there are indeed large differences, such as in the size of the memory store. But Turing challenged the idea that the electronic brain would have to be similar to the human brain in all respects to count as something that could think: “To take an extreme case, we are not interested in the fact that the brain has the consistency of cold porridge.” Indeed, his proposed test, where the machine is interviewed by teleprinter, filters out the many differences that are irrelevant to the issue of whether computers can think.
Did listeners buy Turing’s arguments? Audience reaction was mixed, with comments ranging from “This was a pure gem” to “I think it sad to hear men of such learning wasting 45 minutes in such a fashion.” Some listeners raised exactly the same objections Turing anticipated and countered in his 1950 article in Mind. Thinking, some said, is “the God-given prerogative of man” and computers “will never be anything other than a cleverly constructed mass of metal.” Yet overall 54 percent of listeners polled by the BBC gave the broadcast an A or A+.
It seems that these people didn’t detect (or object) that these experts were reading previously prepared comments, as was standard practice in those days. Prior to airtime, all four departed from the original script, with Turing making the most changes. The final version—the one actually used—is, fortunately, preserved at the BBC Written Archives Centre, which holds a vast and only partly cataloged collection of old scripts and other historical papers.
Turing’s biggest single change to the script is to his discussion of learning by analogy, such as when cosmologists explain the idea of the expanding universe using the analogy of an inflating balloon. Turing was interested in how an analogy can, as he said, “ring the bell” in the brain. Today’s skeptics of what is now called strong artificial intelligence say that a computer is incapable of these lightbulb moments. Turing, by contrast, thought that if only scientists could discover a mechanical explanation of how analogical thinking works in the human brain, they could program a computer to do the same.
Turing’s fundamental message about thinking in his 1952 radio broadcast was that we shouldn’t set the bar any higher, or any differently, for computers than we do for people. We don’t decide that our fellow human beings think by putting their brains under a microscope—ordinary everyday interaction is enough. That was Turing’s astute observation. If you want to tell whether a machine thinks, try communicating with it.
A.M. Turing’s and M.H.A. Newman’s words reproduced courtesy of, respectively, the copyright holder to Turing’s works and Damien Newman. BBC copyright material reproduced courtesy of the British Broadcasting Corporation. All rights reserved.
About the Author
Diane Proudfoot heads the department of philosophy at the University of Canterbury, in New Zealand. Her interest in artificial intelligence led her naturally to Alan Turing. For a talk several years ago, she wanted to replay excerpts of Turing’s BBC radio broadcasts, but no recordings existed, so she recorded actors reading from an original script. “This was my short-lived career as a theatrical producer,” says Proudfoot. Her study of Turing’s radio scripts became the basis for this article.
To Probe Further
S. Barry Cooper and Jan van Leeuwen, eds., Alan Turing: His Work and Impact. Amsterdam: Elsevier, 2013.
Jack B. Copeland, “Why Eugene Goostman Did Not Pass the Turing Test,” Huffpost Tech, 10 June 2014.
Jack B. Copeland, Turing, Pioneer of the Information Age. Oxford: Oxford University Press, 2012.