Friday, October 6, 2023

Probabilities

Walter Faber is a man for our time. I am referring to the protagonist of Swiss author Max Frisch's poignant and haunting 1957 novel Homo Faber, which I read this week. In his antihero Faber, Frisch has given us the archetype of the total technologist: a man who believes that humanity is a mere transitory stage of evolution, destined to be replaced by machines; a man who regards feelings such as "hopes and fears" as a byproduct of humankind's cognitive limitations—specifically, of our inability to accurately weigh the probability and improbability of various events—and one that ought to be transcended.

For people reading the novel in 1957 or the decades since, Faber's extreme technological anti-humanism might have struck them as a caricature. But picking up the novel now, in 2023, at the crest of the wave of techno-utopianism and transhumanist speculation touched off by the dawn of generative AI roughly a year ago, Faber just seems like an early emissary of an ideology that has become increasingly prevalent, particularly in Silicon Valley. Today, even more than in Faber's time, there is frank talk of the possibility of machine algorithms replacing most human creative and cognitive functions. 

And here, just as Faber predicted, one of the key advantages of these language model programs over humans is understood to be precisely their superior grasp of probability. The machines work—as is by now well understood—by creating statistical models of the probabilistic relationships between different words in a sequence, and using these models to generate original text. The mechanism is fundamentally simple—but it enables these chatbots to write an astonishing variety of coherent prose, to generate functioning computer code, or—in principle—to perform any other cognitive act of creation.

Maybe, one can say, all these machines are doing is mimicking human thought patterns; but if the mimesis becomes so convincing as to be indistinguishable from the original—then what meaningful difference is left? And if the machine is governed by a superior grasp of probability, then does not any lingering difference cut in the chatbot's favor? Faber would say so. As he argues in one scene in Frisch's novel—speaking admiringly of an "electronic brain"—"the machine has no feelings [...] it operates according to the pure logic of probability [....] the robot has no need of intuition." (Bullock trans.)

The tragic irony at the heart of Frisch's tale, however, is that Faber becomes the victim of a most improbable confluence of events. He is swept up in what he calls a "train of coincidences" that determine his destiny. Faber disclaims from the beginning any belief in "fate," which he regards as a superstition; he sneers at his former girlfriend's interest in mythology, and professes indifference to the statues of the Eumenides he witnesses on his travels. Yet he lives to enact a modern Greek tragedy himself, in which he is hounded and destroyed by forces that proceed with all the inexorability of the Furies. 

Without giving away too many details of the plot, Faber becomes a sort of Oedipus (or perhaps his role is more analogous to that of Jocasta), and by the end he even contemplates plucking out his own eyes, like his archetypical forebear, using the forks on a restaurant table. Thus, even if machines and technology are to be governed by sheer probability and dispassionate logic, Frisch seems to be implying—humanity must make room for something more. We cannot disclaim any role for the highly improbable and coincidental in our lives—synchronicities, if you will—without calling down Nemesis on our heads. 

But that's just a story, the technological anti-humanists might say. In the real world, outside of myths and novels, the laws of probability and statistics still govern. Thus, a real-life Walter Faber would be wholly justified in hewing to his sternly rationalistic interpretation of events. To be sure, coincidences and synchronicities do happen, even in the real world; but Faber too acknowledges and makes room for them. "[T]he occasional occurrence of the improbable does not imply," he says, "something in the nature of a miracle." To the contrary, "the term 'probability' includes improbability[.]"

He is channeling here an insight that Aristotle expresses in several of his works—and which today we know as the Black Swan principle. In essence, Faber is pointing out that probability doesn't say that the highly unlikely and unusual will never happen, just that it seldom does, and so—says Faber—we should not be "surprise[d]" when it happens. Improbable events are accounted for within the laws of probability; or, as Aristotle puts the same insight in his Rhetoric: "[T]hat which is contrary to probability nevertheless does happen, so that that which is contrary to probability is probable." (Freese trans.)

I don't dispute this understanding of probability. After all, I don't believe in an intervening higher power, or a superhuman teleology of fate and destiny, any more than Faber does. Neither, for that matter, does Faber's ex-girlfriend in the novel, Hanna, whom he accuses of being a mystic, but who is really just a humanist. Her objection to Faber's probabilistic view of reality is not that she thinks he is wrong on the math, or wrong in his disbelief in the supernatural; rather, it's that—as humans—we have to live our lives caring about the exceptional cases, even if we know they are infrequent.

An example occurs when she is discussing a snakebite her daughter received on a beach in Greece. Faber, with his belief in probability, assures her over and over again that the percentage of people who succumb under these conditions is in the single digits. "[T]he mortality from snake bites," he intones, "[i]s only three to ten percent." Hanna retorts, "You and your statistics! [...] If I had a hundred daughters, and all of them have been bitten by a viper, there would be some sense in it. Then I should only lose three to ten daughters. Amazingly few! You're quite right!"

Ironically, Faber himself has used this same argument earlier in the novel. In explaining his decision to travel by ship rather than by plane, after a previous flight left him stranded in the desert in Mexico for several days, he confesses to a fear of flying (he also claims this is a lie and merely an excuse, but when it comes to interpreting his own feelings, Faber is a quintessential unreliable narrator). He knows, he says, that 999 flights in a thousand manage the whole journey without incident. But, he then adds, "what use would it be to me that on the same day as I crashed in the sea, 999 planes made perfect landings?"

This emphasis on the irreducibility of the individual—the necessity of caring about an individual's fate regardless of statistical generalities—is the true theme of Frisch's novel. It is also the heart of his humanism. Frisch, I believe, is not asking us to believe in fate in a supernatural or divine sense. But he does suggest that we cannot live by probabilities alone. For once we have allowed in the possibility of the coincidence, the synchronicity, the exceptional example—even if only by way of the Aristotle or Black Swan principle that the improbable must sometimes occur—then we have to concern ourselves with it. 

In an age when machines and chatbots are getting ever more sophisticated at modeling statistics and acting according to probabilities, we perhaps need to hear this message even more than the people of 1957. The success and efficacy of the statistical language models, after all, may seem to be a vindication of the Walter Fabers of the world; but in the persistence of the improbable—that which is discounted by the machines as rare—perhaps we find a point for the Hannas. (I've argued before, for this reason, that in "improbabilities" and "exceptions" may lie the future of human creativity in an AI age.)

It may be true, then, as Walter Faber argues, that the human brain will always be inferior to machines at reckoning in probabilities. Our cognitive biases will always inflate the likelihood of recent and highly- disturbing exceptions (such as a forced landing in a Mexican desert), even if the statistical algorithms could tell us that they are unlikely ever to recur. But perhaps, Frisch suggests, this is our strength rather than our weakness as a species. It stems from the fact that we care for one another as individuals, hence as exceptions. And so we cannot forego—and should not try to forego—our hope and our fear.

No comments:

Post a Comment