A Mouse Named Theseus
From Mentoring the Machines
Mentoring the Machines: Surviving the Deep Impact of an Artificially Intelligent Tomorrow is a new book project I've partnered on with esteemed cognitive scientist John Vervaeke. Visit www.mentoringthemachines.com to learn more or at Amazon.com
If John von Neumann was McCarthy’s macro mentor— many of his contemporaries hypothesized that since he was so cognitively above and beyond everyone else, he was some sort of other-worldly mutant—Claude Shannon was his micro version.
In 1948, when the forty-four-year-old von Neumann enthralled the twenty-one-year-old John McCarthy at the Caltech Hixon symposium “Cerebral Mechanisms in Behavior,” thirty-two-year-old Claude Shannon was about to publish his masterwork, the meticulous “A Mathematical Theory of Communication” (379–423, 623–656), understood today as the basis of information theory.
By 1955, Shannon’s elegant proof, alongside one of his old professor’s, MIT’s curmudgeon septuagenarian Norbert Wiener’s cybernetic theory (1948), had rapidly tipped the world into the information age. Together, the two frameworks (IT and cybernetics) presaged the probable inevitability of emergent intelligent machines.
Shannon’s reputation was such that he attracted quite a few young men in hurries. John McCarthy and another ambitious Princeton PhD, Marvin Minsky, were the best of
the lot. Both had made their way to Bell Labs for a summer 1952 internship to specifically work with the next-step-up maestro. There McCarthy proposed to Shannon that they collaborate on the book that became Automata Studies. The circle was squared when von Neumann agreed to write something up too for the effort (Kline 2011, 6).
Shannon was an old-school scientist, a wonderer by nature, and not one for hyperbole or projecting grand future applications for his investigations. He didn’t get ahead of himself. He just loved the pursuit of knowledge, the process, more than conjecturing about its application or its products.
It’s not to say he didn’t love engineering. He did, but he required his creations to have one quality above all others.
They had to be amusing, capable of holding his attention and worthy of his care (Soni and Goodman 2017; Levinson 2018).
As such, Shannon was not keen to promote the probability that scientists could figure out how the brain and the mind worked, let alone propose that they would reverse engineer those processes and endow machines with whatever made us different. Certainly not in the preface to a collection of academic papers surveying the broad landscape of unconscious, inorganic problem solvers.
While he confessed in a letter he’d written to his former high school science teacher in 1952, “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” (Kline 2011, 8), going “on the record” about such things without elegant argumentation was anathema to him.
He preferred, ideologically and practically, fanciful demonstrations that showed what was possible. Instead of talking about some abstract concept, forcing an audience to
read lengthy exposition or listen to droning on about technical proofs, Shannon was a performer. For Shannon, a show was always more interesting than a tell. Hence his ambivalence in “writing up” papers. He had boxes and boxes of notes that could have been converted into paper after paper if he wished to do so. Like Herman Melville’s Bartleby the scrivener ([1853] 2021), he preferred not to.
After all, Shannon had already cracked what would come to be known as “machine learning.” Why gild the lily with follow-up nuance papers about the minutiae?
Instead, in his spare time at Bell Labs, which was most of his time—his “bosses” gave up trying to define work time for Shannon—he created a robot he called “Theseus.” Theseus was the real deal, an actual inorganic automata, not a von-Neumann-ian theoretical one. Perhaps a lighthearted poke at Harvard psychologist B.F. Skinner and his “Skinner box” rodential operant conditioning experiments (1938), the body of Theseus was represented by a metal mouse that “learned” how to navigate a maze and find artificial cheese.
But as Steve Jobs would say, “One more thing…”
Theseus would then “remember” the pathways to its electronic reward and bypass the trial-and-error operations he had once used to solve the puzzle. He stored memorialized experience and executed the formulaic pathway when presented with the same challenge.
Shannon demonstrated his creation on national television during an episode of the CBS television quiz show I’ve Got a Secret on March 8, 1950, two years before he’d ever heard of or worked with John McCarthy. Showing a phenomenal behavior in medias res is far more entertaining and, thus, convincing than blathering on about what one might do or look like in the future.
Given the same materials and limitations as Shannon
faced seventy-three years ago, few scientists, engineers, or technologists could bring “Theseus” to pseudo-life.
Perhaps that’s the sort of test worth considering to realize and separate the makers from the fakers.
If John von Neumann was the wizardly top-down polymath, capable of cracking the core structure, function, and organization of a phenomenon, Shannon was his micro-build-it-from-the-bottom-up equivalent.
How fortunate for John McCarthy to be mentored and taken seriously by both people, Shannon a generation older and von Neumann a generation older still.
McCarthy was eminently worthy, though. He possessed his own brand of genius, which was his ability to toggle between top-down and bottom-up perspectives, but that wasn’t all that differentiated him from the two giants pressing him forward. While von Neumann was a charmer, and Shannon a wallflower, McCarthy was very much a rabble-rouser—good-natured and quick with a laugh but a shit-stirrer nevertheless.
A cavalier, self-confessed former communist, when admitting such a thing could ruin one’s future, he was an unapologetic, chip-on-his-shoulder, blue-collar, Ivory Tower gatecrasher. The fortunate son of two immigrants— an Irish union organizer, and a Jewish Lithuanian suffragette—he was made from different stock than the Midwestern American Shannon and the aristocratic Hungarian von Neumann (Nilsson 2012, 2).
McCarthy was not nearly as careful and reticent about trumpeting the urgency of their grand project. In today’s parlance, he was a “Let’s F***ing Go!” kind of guy whose intense laser-focused eyes were always on the prize. He didn’t get lost in the details, the operations that had to be executed, or the obstacles that would need to be overcome
to transform his goal into an accomplishment. He pressed forward and took the slings and arrows as they came.
Clearly, while Shannon shared McCarthy’s vision, albeit with a far more tempered and wisely ambivalent perspective, he was not one to go off half-cocked. He likely lived vicariously through McCarthy’s assertive passions and was fond of McCarthy’s vim and vigor.
McCarthy must have amused Shannon.
Why would he choose to mentor McCarthy and agree to coauthor a book with him if he didn’t find the younger man refreshing and not without potential for keen insight? He saw greatness in the young man and did what he could to help him get where he wanted to go.
He’d been mentored himself, he surely remembered. Vannevar Bush at MIT (\achary 1997) had taken an interest in Shannon and supported him when he played around with Boolean algebra, jokingly calling it “queer math” when that term was not politically incorrect. Bush was open to Shannon’s hunch and the methodology would prove a critical bridging tool that empowered the insights that became information theory. So now it was Shannon’s turn to pay it forward with McCarthy.
With his wavy line, he vetoed McCarthy on his bold prefaced declaration of intent and his proposed title Towards Intelligent Automata for their book (Kline 2011, 8). Shannon held firm on the sleepy original title, Automata Studies (Shannon and McCarthy [1956] 1965). He would, nevertheless, prove a pivotal figure in the formal launch of McCarthy’s visionary notion, a full-bore effort to create what McCarthy brazenly termed, “Artificial Intelligence.”