Could a machine think?—Could it be in pain?—Well, is the human body to be called such a machine? It surely comes as close as possible to being such a machine. But a machine surely cannot think!—Is that an empirical statement? No. We only say of a human being and what is like one that it thinks. We also say it of dolls and no doubt of spirits too.
—Ludwig Wittgenstein
I’m a machine and you’re a machine, and we both think, don’t we?
—Claude Shannon
If Shannon had peculiar work habits before the publication of his information theory, his growing reputation granted him the license to indulge those peculiarities without reservation. After 1948, the Bell Labs bureaucracy could not touch him—which was precisely as Shannon preferred it. Henry Pollak, director of Bell Labs’ Mathematics Division, spoke for a generation of Bell leaders when he declared that Shannon “had earned the right to be non-productive.” Shannon arrived at the Murray Hill office late, if at all, and often spent the day absorbed in games of chess and hex in the common areas. When not besting his colleagues in board games, he could be found piloting a unicycle through Bell Labs’ narrow passageways, occasionally while juggling; sometimes he would pogo-stick his way around the Bell Labs campus, much to the consternation, we imagine, of the people who signed his paychecks.
Colleagues may have bristled at all this, but Shannon was, by this point, a legend masquerading as an ordinary employee. He had engineered as close to an emeritus role for himself as he could under the contractual obligations of full-time employment. That meant being able to work with his door closed, practically a sin in Bell Labs circles. It also meant pursuing personal projects to whatever conclusion suited him. One receipt from this time records a series of hardware store purchases that he billed to the Labs, presumably part of a machine Shannon was building, the result of which could hardly have mattered to the practical work of the phone company.
But none of this was cause for alarm within Bell Labs. There was no doubting the quality of Shannon’s mind, and thus no one thought to ask rigorous questions about how he was keeping it occupied. After all, the “founder of information theory” had, essentially, dropped the theory into everyone’s laps after finishing it in private. Who was to question what else he might be up to behind closed doors?
One curious side effect of all this freedom: Shannon became, during this period, an inconsistent correspondent, even as the volume of his correspondence began to grow with his reputation. Letters went unanswered for long stretches, so many and for so long that Shannon collected them all in a folder labeled “Letters I’ve procrastinated in answering for too long.” In the words of Jon Gertner, “it seemed lost on Shannon that the scientist who had declared that any message could be sent through any noisy channel with almost perfect fidelity was now himself a proven exception. Transmissions could reach Claude Shannon. But then they would fail to go any further.”
These weren’t all anonymous fans and unknown admirers, either. Shannon received correspondence from eminent scientists, high-level government officials—and even L. Ron Hubbard.
It’s only with the benefit of several decades that we can judge this to be one of the stranger correspondences in Shannon’s life. It’s worth being emphatic on one point: yes, the founder of dianetics and the Church of Scientology sought Shannon out. No, Shannon was not a Scientologist himself. Hubbard, it appears, was more interested in Shannon than Shannon was in Hubbard. Shannon, for his part, wrote a letter to Warren McCulloch, a leading cybernetics researcher at MIT, asking if McCulloch would meet with his “friend” Hubbard.
Hubbard, it seems, was better known to Shannon as an author of space operas than as a budding religious crackpot. “If you read Science Fiction as avidly as I do you’ll recognize him as one of the best writers in that field,” Shannon writes. “Hubbard is also an expert hypnotist and has been doing some very interesting work lately in using a modified hypnotic technique for therapeutic purposes. . . . I am sure you’ll find Ron a very interesting person, with a career about as varied as your own, whether or not his treatment contains anything of value.”
Hubbard would later write to Shannon thanking him for help on his research and promising a copy of Dianetics when it was released. No further correspondence between the founder of information theory and the pope of Scientology has been recorded. Yet, as William Poundstone notes, “to this day Hubbard’s Scientology faith cites Shannon and information theoretic jargon in its literature and web sites.”
The correspondence with Hubbard was positively staid in comparison with many of the other letters that piled up on Shannon’s desk. Alongside the usual traffic of scientific colleagues reaching out for a paper or a book review, there was also a steady stream of cranks who sought out Shannon’s approval for their private researches, or whose paranoia about the phone company led them to contact one of its prominent faces. One handwritten letter began, “Dear Dr. Shannon, I am enclosing a ‘Theory of Space.’ I have sent it to several other eminent scientists but thus far have not received a reply. . . . ” A letter from the self-identified “IDEA MAN” requested Shannon’s assistance “to complete and verify a 15 year search to locate and exactness OF life, mind, and energy.”
Another was more menacing:
Dear Sir, Your mechanical robot Bel, the idol in the Bible, is a mechanical monstrosity. Your robot is breaking five amendments of the Constitution (1, 3, 4, 5, + 13th). God admits that I am laughing at you. You are making a traitor out of the President of the U.S. And the F.B.I. By letting your robot deceive you. I have threatened to sue the NY Telephone Co. Of NY City, and I will, if you don’t wake up.
Shannon was good-natured about it all—able to use his charm to defuse tough inquiries, or more often, to ignore them without a second thought. Unlike many scientists who parlayed successful research careers into lives as public intellectuals, he did not seem to consider using his growing standing within the world of science as an opportunity to expand his network outside of it. Nor did he take it as a responsibility to opine on policy or act as a public educator. If anything, he closed himself off further, ignoring letters, colleagues, and projects, and spending his time and attention absorbed by the puzzles that interested him most. Shannon had earned this right—information theory was painstaking work—and he found himself drawn now to new problems and fresh horizons, including some that seemed, to colleagues, borderline ridiculous for someone of Shannon’s stature.
“I think the history of science has shown that valuable consequences often proliferate from simple curiosity,” Shannon once remarked. Curiosity in extremis runs the risk of becoming dilettantism, a tendency to sample everything and finish nothing. But Shannon’s curiosity was different. His kind meant asking a question and then constructing—usually, with his hands—a plausible answer. Could a robotic mouse navigate its way through a maze? Build one and find out. Could a machine turn itself off? Make one that is trained to commit technological hara-kiri. What other people called hobbies, he thought of as experiments: exercises in the practice of simplification, models that filed a problem down to its barest interesting form. He was so convinced of a machine-enabled future, and so eager to explore its boundaries, that he was willing to tolerate a degree of ridicule to bring it to pass. He was preoccupied, as he wrote to a correspondent, “with the possible capabilities and applications of large scale electronic computers.” Considered in the light of that future, our present, his machines weren’t hobbies—they were proofs.
One starting point for all of this tinkering with mechanics was benign: a wife’s Christmas gift. “I went out and got him the biggest Erector set you could buy in America. It was fifty bucks and everyone thought I was insane!” Betty Shannon later told an interviewer. Claude Shannon added: “Giving it to a grown man! But the fact of the matter is that it was extremely useful and I used it to try out different things.” And like a child with a new present, Shannon became obsessed: the basement became a mess of loose erector parts, and he stayed up late into the night, building away.
The first idea was a dry run: a mechanical turtle that stalked the Shannon house, bumping into the walls and turning around only to bump into a different wall. But the hapless turtle anticipated the next invention, the one that would, unexpectedly, attract national attention: Theseus, the maze-solving mouse. As one report had it, the idea for a mechanical mouse that could navigate a maze grew out of Shannon’s attempt to escape the famous garden maze at London’s Hampton Court Palace. It took him twenty minutes; he figured it could be done faster. Later, the most famous photo of Claude Shannon would depict him with a finished Theseus and the maze, Shannon’s hand setting the mouse down within the walls. It was named, somewhat optimistically, for the Greek hero who slew the Minotaur and escaped the fearsome Labyrinth; but for now it was a three-inch piece of wood with copper whiskers and three wheels.
It was Shannon’s research on switching, and his work for the telephone company, that inspired the guts of the contraption. Seventy-five electromechanical relays, the sort used as switches in the phone system to connect one call to another, toggled like railroad tracks shifting trains to allow the mouse to navigate. Betty completed the wiring for the earliest prototype. “We did all this at home at night after work,” she said.
Theseus was propelled by a pair of magnets, one embedded in its hollow core, and one moving freely beneath the maze. The mouse would begin its course, bump into a wall, sense that it had hit an obstacle with its “whiskers,” activate the right relay to attempt a new path, and then repeat the process until it hit its goal, a metallic piece of cheese. The relays stored the directions of the right path in “memory”: once the mouse had successfully navigated the maze by trial and error, it could find the cheese a second time with ease. Appearances to the contrary, Theseus the mouse was mainly the passive part of the endeavor: the underlying maze itself held the information and propelled Theseus with its magnet. Technically, as Shannon would point out, the mouse wasn’t solving the maze; the maze was solving the mouse. Yet, one way or another, the system was able to learn.
When it arrived at Murray Hill, Theseus became a minor Bell Labs celebrity. The mouse earned Shannon and the Labs a patent. The Labs also commissioned a short film starring Shannon and Theseus. The seven-minute short was produced with the general public in mind.Shannon, nattily dressed in a dark suit with a light red tie, explains the maze-solving mouse and its mechanisms in the deliberate, step-by-step manner of a college professor. “Hello,” he begins, “I’m Claude Shannon, a mathematician here at the Bell Laboratories.” He dives in, explaining both what the viewer is watching—the mouse making its way through a maze—and what underlies the system. When it comes to extending the analogy of the mouse and maze further than what’s in front of him, Shannon only hints at it, only gestures in the direction of what Theseus means for the possibilities of a robot brain:
Of course, solving a problem and remembering the solution involves a certain level of mental activity, something akin, perhaps, to a brain. A small computing machine serves Theseus for a brain. . . . We have placed the brain cells of Theseus, if you like, behind a small mirror here.
Shannon explains that Theseus’s brain is something more basic and familiar, something akin to the system that powers the telephone’s elaborate network of switches and wires. “Here at the Bell Telephone Laboratories, we’re concerned with improving your telephone system,” Shannon says, coming the closest he ever will come to shilling for his employer. That moment, along with the images of telephones dialing and switches activating, and the cheery music in the background, was a necessary piece of PR: concerned as they were about regulatory interest in their work, the higher-ups at Bell Labs and AT&T couldn’t allow Claude Shannon to go into theaters, schools, or universities with a robotic mouse and risk giving the appearance that the enormous leeway and profits they had been granted by the U.S. government was being devoted to frivolities.
Shannon closes the video by changing the contours of the maze and putting Theseus inside a square with no exit. “Like the rest of us, he occasionally finds himself in a situation like this,” Shannon says, as the mouse moves, hits a wall, moves, hits a wall, and ends up trapped. The camera cuts to Shannon, who smirks, and the music cues the end of the demonstration.
The outside world took an unexpected interest in Theseus, and the celebrity it earned Shannon and Bell Labs impressed Shannon’s bosses. One story lodged itself in Bell Labs lore. Henry Pollak recounted what happened when Shannon gave a demonstration of Theseus for the AT&T board of directors:
I was told that one of the board members at the end of the presentation said, “Now that’s the kind of original thinking we need at AT&T! I propose that Claude Shannon be made a member of the board!” And they had a hell of a time dissuading this guy from the idea of making Claude Shannon a board member, and they finally got around it by the fact that Claude Shannon did not own enough shares of stock to be made a board member.
Time magazine featured Theseus in a short article: “Mouse with a Memory.” Life published a photo of Theseus finding the cheese. Popular Science ran a three-page spread under the headline “This Mouse Is Smarter Than You Are.” Theseus found its way into more serious quarters, as well. The mechanical mouse was a featured subject of discussion at the famed 1951 Macy Conference, an interdisciplinary meeting of scientists and scholars in New York. Shannon was in attendance along with many of the leading authorities on artificial intelligence and computing, as well as the anthropologist Margaret Mead. The incongruity of such leading minds discussing a mechanical mouse was mitigated by the fact that Theseus (or, to be exact, the mouse-maze system as a whole) was a working example of the “artificial intelligence” that many of the esteemed attendees had spent their careers pondering only in theory. Theseus was artificially intelligent. When an attendee pointed out the obvious—that if the metallic cheese were removed, the mouse would simply sputter along, searching in vain for a piece of cheese that was no longer there—conference attendee and social scientist Larry Frank responded, “It is all too human.”
In the end, the editors of the conference proceedings offered a skeptical assessment of Theseus (in the process demoting him, perhaps unconsciously, from “mouse” to “rat”):
The fascination of watching Shannon’s innocent rat negotiate its maze does not derive from any obvious similarity between the machine and a real rat; they are, in fact, rather dissimilar. The mechanics, however, is strikingly similar to the notions held by certain learning theorists about rats and about organisms in general.
In other words, Theseus was not a real intelligence, but he did model one aspect of how a rat or any other another creature might learn. If Shannon indulged in a good-natured eye roll, it is not recorded.
Shannon would later tell a former teacher of his that Theseus had been “a demonstration device to make vivid the ability of a machine to solve, by trial and error, a problem, and remember the solution.” To the question of whether a certain rough kind of intelligence could be “created,” Shannon had offered an answer: yes, it could. Machines could learn. They could, in the circumscribed way Shannon had demonstrated, make mistakes, discover alternatives, and avoid the same missteps again. Learning and memory could be programmed and plotted, the script written into a device that looked, from a certain perspective, like an extremely simple precursor of a brain. The idea that machines could imitate humans was nothing new. But Theseus had made that idea—and the promise that a machine could memorize and deduce—seem vividly real.
Over the years, Shannon’s thinking and nonthinking machines took on a range of shapes and styles. Some served as an oblique social commentary: the “Ultimate Machine,” when its single switch was flipped, would reach out a mechanical hand and turn itself off. THROBAC (“Thrifty Roman-Numeral Backward-Looking Computer”) was a calculator whose keys, processing, and output all worked in Roman numerals, useless except to those who could decipher the difference between, say, CLXII and CXLII. These gadgets had the character of sly, private jokes. But Shannon also placed a high value on his tinkering. “The design of game playing machines may seem at first an entertaining pastime rather than a serious scientific study,” he allowed, but there was “a serious side and significant purpose to such work, and at least four or five universities and research laboratories have instituted projects along this line.”
His goals were as grand as the means, at least at the time, were simple. “My fondest dream is to someday build a machine that really thinks, learns, communicates with humans and manipulates its environment in a fairly sophisticated way,” Shannon admitted. But he was not bothered by the usual fears of a world run by machines or a human race taking a backseat to robots. If anything, Shannon believed the opposite: “In the long run [the machines] will be a boon to humanity, and the point is to make them so as rapidly as possible. . . . There is much greater empathy between man and machines [today] . . . we’d like to close it up so that we are actually talking back and forth.”
That quote, and several of the anecdotes that followed Shannon until the end of his life, originated in a now largely forgotten profile of Shannon in Vogue magazine, titled “The Man-Machines May Talk First to Dr. Shannon.” As a part of the profile, Shannon spoke at length with writer Brock Brewer about the connection between automata and their creators. (And as it was a feature for Vogue, and not, say, Scientific American, Shannon was expected to endure a photo shoot, with the renowned Henri Cartier-Bresson behind the lens. It put Shannon in illustrious company: Cartier-Bresson’s other shoots included Mahatma Gandhi’s funeral, Queen Elizabeth’s coronation, and the first several months of Mao Zedong’s ascendance.)
The piece opened with what, at the time, must have seemed the musings of a madman: “Dr. Claude E. Shannon . . . who creates, plays with, stays a think ahead of thinking machines, looks forward to man and machine talking back and forth. For him, why not?” For Shannon, the prospect of artificial intelligence was a tangible reality, not a futuristic fantasy. Imagining how “computer-controlled exploratory robots” would handle accidentally falling into a hole on the moon (and anticipating the Roomba in the process), he said,
you have to think of problems like this when machines are running around loose in the real world. A machine on the moon must protect itself—not fall down a hole, without your having to tell it not to. It’s the same problem we’re going to have some day with furniture when there are robot housekeepers running around the house, picking up things.
Shannon was happily oblivious to fears of exponentially expanding artificial intelligence, of robots begetting ever-more-advanced robots and putting the human race at risk. In fact, his was a thoroughly optimistic vision of technological progress—one in which machines ought to be given increasing abilities, responsibilities, and information. In response to the question of what the point of all his robot work might be, Shannon remarked that his goals were threefold: “First, how can we give computers a better sensory knowledge of the real world? Second, how can they better tell us what they know, besides printing out the information? And third, how can we get them to react upon the real world?”
Or, as he told a later interviewer, in an even more optimistic mood:
I believe that today, that we are going to invent something, it’s not going to be the biological process of evolution anymore, it’s going to be the inventive process whereby we invent machines which are smarter than we are and so we’re no longer useful, not only smarter but they last longer and have replaceable parts and they’re so much better. There are so many of these things about the human system, it’s just terrible. The only thing surgeons can do to help you basically is to cut something out of you. They don’t cut it out and put something better in, or a new part in.
In fact, when it came to human superiority over machines, “thinking is sort of the last thing to be putting up a fight.” While Shannon did not expect a computer to pass the famous, and famously open-ended, Turing Test—a machine indistinguishably mimicking a human—within his lifetime, in 1984 he did propose a set of more discrete goals for artificial intelligence. Computer scientists might, by 2001, hope to have created a chess-playing program that was crowned world champion, a poetry program that had a piece accepted by the New Yorker, a mathematical program that proved the elusive Riemann hypothesis, and, “most important,” a stock-picking program that outperformed the prime rate by 50 percent. He was about half right: a computer did defeat the world chess champion in 1997, four years before Shannon’s deadline, and computers do conduct the bulk of the world’s stock trading.
Yet there were moods in which Shannon’s cheeriness over the future of machines curdled into misanthropy. “We artificial intelligence people are insatiable,” he once wrote. Once machines were beating our grandmasters, writing our poetry, completing our mathematical proofs, and managing our money, we would, Shannon observed only half-jokingly, be primed for extinction. “These goals could mark the beginning of a phase-out of the stupid, entropy-increasing, and militant human race in favor of a more logical, energy conserving, and friendly species—the computer.”