“It says here, ‘In the beginning was the Word,’

“Already I balk.”

-Goethe, Faust, lines 1224-5

Often the way we talk about something tells us as much as its substance. In the case of Artificial Intelligence, it turns out, this is true all the way back to its foundational document.   

I recently became curious about a quote widely attributed to Alan Turing, that  “a computer would deserve to be called intelligent if it could deceive a human into believing that it was a human.” It’s all over the Internet, and in many books and newspapers, but I have been unable to find any citation for where Turing said it. It’s nowhere in his famous paper, “Computing Machinery and Intelligence, the foundational document in which Turing posits his “Imitation Game,” where an independent human judge must decide whether questions are being answered by a human, or by a digital computer imitating a human. People simply seem to think something is the source, and these things take on a life of their own.

There’s much else in Turing’s famous paper, though, that makes it more interesting today than ever. Its discussion of how a programmable digital computer could be a “universal machine” capable of replicating any other machine’s function through its program has broadly influenced computer science. And Turing’s insistence that such a machine would have humanlike capabilities, astonishing for the time, has been important for how computers went on to be marketed and sold. Today’s rampant speculation about whether the new styles of AI computation means that machines are now growing brains and autonomy, makes the paper more relevant, and revelatory, than it has been in decades. 

More than anything, though, the paper is a doozy of brilliant insight and borderline mad assertions. Thus it speaks very much to the present moment. 

Written in 1950, just three years after transistors were invented (and three years before they were used in a computer), Turing brilliantly lays out the future design of digital computers, and identifies the key elements standing in the way of a machine passing what came to be known as The Turing Test. These are computing power and memory, and when today’s gen AI proponents tell us that computers are on their way to becoming intelligent, they’ll still tell you that all we need is more compute and bigger data sets. At the same time, Turing’s paper asserts that Moslems believe women do not have souls, that telepathy is scientifically proven and thus ghosts are real, and that you can’t educate a computer by beating it, the way you can a child, so that in the future computer scientists will need to develop a symbolic language that generates punishment.

Turing is an accidental visionary too. He believes that “at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted.” This was indeed true, but it turned out to be a statement about the success of marketing computers as “thinking machines” with “brains” and capable of “learning” and “remembering” in ways that are nothing like these functions in humans. His conjecture was validated, in other words, by the creation of a multibillion-dollar industry, and not by any computer capability. 

This, like much else in the paper, including the ignorant and the fanciful, shocked my assumptions about Turing, and what he seems to be after in what has become a foundational document of a vital industry. On closer inspection I now think it tells a profoundly human story about Turing himself.

But folklore matters too, and before looking at his paper it’s worth considering why Turing is broadly thought to have said such a strange thing about intelligent machines. Take the first part: “A computer would deserve to be called…” How is it that a machine of any kind can deserve something? Does a wristwatch, or a bicycle “deserve” as well? We may speak of animals “deserving” one thing or another; indeed, the popular misquote of Turing may stem from Darwin’s assertion that worms “deserve to be called intelligent” for the way they understand the shape of their burrows (earthworms occupied Darwin’s thoughts for years after Origin of the Species.) That is still a far cry from considering a rock or a toaster deserving of anything in particular. Furthermore, how does “to be called” something make it actually the thing? Darwin appeared to be reflecting on the idea of general intelligence, but the Turing misquote weighs the importance towards computers. It’s almost as if we collectively yearn for them to be intelligent.

The next part, “if it could deceive a human being into believing it was human” is a showstopper. It’s close to what Turing is after in his Imitation Game. But what proof of intelligence is derived from such a deception? Did the computer deceive, or does the victory not belong to the human who wrote the program? If so, how is this a sign of a computer’s intelligence, except as a manifestation of human intelligence, and entirely contingent on a human actor? It’s a common error; when IBM’s Big Blue beat Kasparov at chess there was far more hand-wringing about the machine’s output  than there was awe at the vast team of humans who had created its means to victory.

The erroneous quote also, like Turing himself in his paper, misses the reality that being fooled is a condition of the recipient, not of the thing doing the fooling. Siegfried and Roy made many people believe they made an elephant disappear, but that didn’t make them metaphysical beings. The program for my Google Maps gives directions in a confident-sounding voice, but that doesn’t mean it deserves to drive.

Nonetheless, the idea that if you can fool a human into believing something is intelligent, then it must be intelligent, became for decades the gold standard of proving computer intelligence, the so-called Turing Test. And for decades people have sworn by it, striving to make machines that seem to us to be human. 

Published 74 years ago, the Turing Test still has adherents and detractors, and a rich history of people trying to game the test itself for bragging rights. Even now, it’s common to hear gen AI companies talk about how well their computer did in the bar exam, as if this means it would make a capable lawyer, and not that it scraped and digested a sufficient amount of data so it could pass a very specific kind of bar exam. It’s also common to speak of large data sets in terms of what multiple of bytes are in use compared with The Library of Congress or The Works of Shakespeare, as to suggest that the size of these data sets was somehow a multiple of human excellence.

Turing begins his paper clearly enough. “I propose to consider the question, ‘Can machines think?’” he begins. Soon after things become peculiar. It is necessary to define “machine” and “think,” he writes, before adding that “it is difficult to escape the conclusion” that these terms might be established by “a statistical survey such as a Gallup poll.” It’s an arbitrary choice that ignores alternative possibilities, like finding a definition in a good dictionary or rounding up some experts to hash the matter out. Concluding that a survey would be weak, he then lights on his own method, a game in which a machine imitates the role of one of two humans in isolated booths, answering questions posed by a third person, who must decide whether the question was answered by a person or a machine.

Turing later limits the definition of “machine” to digital computers, adding that this limitation “will only be unsatisfactory if (contrary to my belief), it turns out that digital computers are unable to give a good showing in the game.” He does not take up the clear possibility that the initial question, “Can machines think?” might simply be answered “No, Alan, they can’t.” Instead, it’s a question of whether digital computers will be the machines that succeed at thinking. It’s poignant how badly Turning seems to want his initial question answered in the affirmative. 

A long section of the paper consists of a series of objections to the idea of intelligent machines that Turing has concocted, along with his reasons why these objections do not hold up. It’s somewhat jarring, not simply because they are all his invention, and as such will of course fail (if any had succeeded, why would he write the paper?) It would be better to publish, and wait for objections from independent readers.

Turing’s concocted objections don’t matter for the success or failure of the imitation game. Shortly before this, he actually rejects his original question of whether machines can think, the original basis for the Turing test, by stating that it is “too meaningless to deserve discussion” (now he tells us.) Instead, Turing simply insists that machines will be spoken of that way. Most of the rest of the paper, the bulk of the work is speculation about the quality of the machine mind and the machine soul, and how it might be perfected.

Often he’s strangely compassionate. When he asks, “How do Christians regard the Moslem view that women have no souls?” Turing is speaking to a supposed theological objection that God only grants souls to men or women, but not to intelligent machines. It’s a spurious question, since this European canard about Islam was debunked in the West decades before (in fairness to Turing, I did find the same idea expressed in a New York Times Travel article from 1973.) It’s notable that he first equates intelligence and possession of a soul. Why even raise it, unless you hope it is true?

Along with the soul, Turing toys with the objection that the emotional dimension of intelligence is not represented in a program, which he terms the argument from consciousness. In effect, he equates raising this with the solipsistic idea that we can only know our own consciousness, and everything outside it is an illusion. This, to Turing, amounts to an either/or. ”I think most of the people who support the argument from consciousness could be persuaded to abandon it rather than be forced into the solipsist position,” he writes. ”They will then probably be willing to accept our test.” The absence of any middle ground, for example we can assume that other minds exist and yet also assume that all machines lack minds, is striking. 

Another extraordinary objection is the idea that machines can’t be programmed to enjoy strawberries and cream, which he calls an argument from disability. This can eventually happen, he writes, but that is not the key thing. ”What is important about this disability,” he writes, “is that it contributes to some of the other disabilities, e.g. to the difficulty of the same kind of friendliness occurring between man and machine as between white man and white man, or between black man and black man.” Setting aside the weird and unnecessary implication that cross-racial friendship is not even entertained, an odd thing to put in a paper on computing, we have moved past intelligence into something like friendship, and its necessity if computers are intelligent. 

We’re now deep into Turing’s hidden goal. The imitation game, and this idea of whether or not computers can think, is really in service to a broader idea, stemming from Turing’s conjecture that by the end of the century they will think. When they do, he wants them to be our friends. To have minds and souls. Even to surprise and delight us: ’Machines take me by surprise with great frequency,’ he writes, ‘This is largely because I do not do sufficient calculation to decide what to expect them to do…” This takes us back to Siegfried and Roy as  interdimensional elephant transporters. Just because you are the recipient of the feeling, that does not imply a specific capability on the part of its generator.

It is not only computers that Turing wanted to be enchanted. ”I assume that the reader is familiar with the idea of extra-sensory perception,” he writes later. “…these disturbing phenomena seem to deny all our usual scientific ideas. How we should like to discredit them! Unfortunately the statistical evidence, at least for telepathy, is overwhelming.” Presumably Turing is referring to the botanist J.B. Rine’s sensational claims of documenting parapsychology, which were widely reported for decades, and debunked as early as the mid-1930s. Given the reality of telepathy, Turing adds, ”it does not seem a very big step to believe in ghosts and bogies.” In other words, if you believe the living can communicate in ways not yet understood, Turing is saying, you should have no trouble with the idea that the dead come back to life. 

As much as arguing for, and providing a means to, intelligent machines, Turing seems as much trying to resolve human loneliness. By the time you get to concluding elements of the paper, a digression into schoolboy beatings as a guide to educating machines, you feel the wound and the alienation as much as any insight into the future of digital technology. 

Read sympathetically, it’s clear that even a great genius like Turing is subject to the misunderstandings and prejudices of his time. Or he didn’t always do his homework – who does? Perhaps even more sympathetically, what comes through for me in Turing’s unsupported assertions and outright errors is a tremendous undercurrent of yearning: That machines might become intelligent companions, astonishing us. That people might connect on some better level. That we readily accept that other minds exist, transcending ourselves into connection. That magic might re-enter ordinary life. That childhood is as innocent as a blank piece of paper, and inevitably inked up.

Perhaps I’m swayed in this view by the well-known facts of Turing’s own horrid fate. Two years after writing this paper he was arrested for homosexual acts, and compelled to undergo a hormone treatment known as chemical castration. Many think this drove him to suicide, which may or may not be true. Turing is also said to have been painfully isolated because he was on the autistic spectrum. As someone with significant experience in the area, I reject any armchair diagnosis. Besides, neither of these is necessary to appreciate the poignancy of Turing’s strange foundational document. 

About midway through the paper, when Turing gives up on his original question and conjectures that machines will someday be spoken of as thinking (a different matter from actual thinking, though neither thing has yet happened in the mainstream), he provides a gauzy motive for his assumption. “Conjectures are of great importance since they suggest useful lines of research.” Turing’s opinion, however, is not even a conjecture, since that is a best guess based on available information, and not simply a personal belief. Rather than a rational justification, Turing exhibits a desperate, and all too recognizable, wish that his hope is true.

As much as projecting forward a technological capability, wanting to believe in intelligent machines is the entire basis for Turing’s powerful and influential paper. And in that sense he is a greater prophet than he is when he posits his game. We read today about people engrossed by machines they believe deserve legal rights, while others sketch out a future in which we will port ourselves into computers with no change of self (other than a simple matter of immortality, and and the unimaginable idea of an intelligence apart from the body.)  Turing’s paper foreshadow’s today’s math-based yearning for miracles. To search for the souls we know are within us, to find kindred souls we know are out there, to connect with them in this life (and even possibly after) as truly and as deeply as possible – could Turing have claimed a more human desire?