Here is a possibility worth holding in mind, just for a moment. What if humans are something better than machines? For that matter, what if it isn’t close?
In a way, the thought sits uneasily. For about 500 years, the scientific method has existed in a state of almost-continual triumph, while humankind has endured a triple fall as a consequence: first from the center of the universe (Copernicus), then from the center of the world (Darwin), and finally from the command of their own minds (Freud). Upon each of these revolutions, and at a thousand points of scientific inquiry between, our pride has received another debilitating shock.
Nor has it been much of a battle. For instance, not long before the debate on evolution, we received an equally devastating proof that humans were not separate from the natural world, but a part of it—cell theory. It was less controversial only because it was irrefutable; a child with a microscope could see that a stalk of grass and the skin from his thumb had the same basic structure. In this sense, even the famous fight over evolution was really just a slower rearguard defeat.
And yet, a single, unconquerable backstop to this series of scientific conquests remains: consciousness. As the philosopher Thomas Nagel famously summarized the problem: Why is it “like” something to be alive? Why are we here, aware, rather than nowhere, being nothing? Researchers over the past few centuries have tried obsessively to answer these questions. Somehow, nevertheless, we are not one iota closer to a definitive solution than the cavemen were.
This is the chasm that the fevered marketers of artificial intelligence have convinced much of the world that they will soon effortlessly leap. In fact, the clear likelihood is that they are not just wrong, but memorably wrong, hilariously wrong. At least, that is one conclusion a reader might draw from Michael Pollan’s searching new book, A World Appears.
Pollan has always been headed in this direction. The central concern of his work has consistently been ingestion—what crosses the threshold between the world and the self. First, his landmark works on eating helped reshape the American diet (“Eat food. Not too much. Mostly plants,” he advised); more recently, he has been interested in psychedelic drugs, once more anticipating his subject’s emergence into mainstream discourse. Consciousness is the logical final destination for this project, and the subject of his new book: everything that a person takes in from the outside, and what that point of intersection means.
[Read: How to have a ‘don’t-know’ mind]
A World Appears begins with a forthright admission that after a great deal of reading, numerous interviews with leading scientists, and extensive personal experimentation, Pollan has arrived at no concrete views about his topic. As he writes, there are currently at least 106 competing hypotheses of consciousness, comprising 22 physicalist accounts (physicalism being the belief that the “mind” is nothing but a quality generated by the physical matter of the brain) and “no fewer than eighty-four non-physicalist theories.” Such a profusion of competing ideas, he dryly observes, is “a pretty good indication that the field is flailing.”
He guides us through that welter in four stages, each representing an ostensible escalation in complexity. The first is one of his favorite subjects, plants, which he initially takes to represent the most rudimentary form of consciousness. But even on that point the ground shifts beneath his feet, as it were: Plants, he reports, can “integrate information from more than twenty distinct ‘senses,’ including all five of ours.”
From there, he moves into the book’s finest passages, about feeling. Feeling, Pollan convincingly argues, actually precedes computation as a necessary condition of consciousness. (One of his most compelling interview subjects, the neuroscientist Antonio Damasio, believes that feeling has been neglected because male scientists long considered it too “feminine” to seriously study.) As he notes, “It is one of the paradoxes of computer science that the ‘higher’ capabilities we once thought of as uniquely human—reason, language, intelligence—have proved easier for machines to master than the more elemental capabilities we share with animals, including feelings and emotions.”
The third section of the book tracks thought, through the lens of Pollan’s attempt to record his own stream of consciousness; the fourth, most mystical one, is about the self—whether it exists, and what might constitute it when we know our physical selves to be continually changing. It culminates with Pollan, 71, meditating in a cave in Santa Fe, making peace with the insoluble nature of his search.
I can think of more lucid and arresting introductions to this subject than A World Appears, which conceptualizes these dense abstractions in a sincere but labored fashion (for instance, Consciousness and the Novel by David Lodge, or the first part of John Searle’s Mind before it becomes too speculative). Yet Pollan’s real genius—the word is not too strong—remains intact. That is his uncanny ability to scent the direction in which the culture is headed. He did it with food and psychedelics, and now, though A World Appears focuses on AI only intermittently, he has done it again. By patiently mapping the problem that many of the creators of large language models claim, either cynically or foolishly, to be on the verge of solving, he brings this technology—which has come to dominate recent headlines, financial markets, and political debates—into a far more realistic light.
[Read: The useful idiots of AI doomsaying]
“Just about any place you push on it,” Pollan concludes, “the computer-as-brain metaphor breaks down.” I laughed out loud when I read one of the many examples he cites in support of this argument: “A recent study demonstrated that a single cortical neuron can do everything an entire deep artificial neural network can.” AI is an exciting and useful tool, but I don’t think that disparity is something OpenAI CEO Sam Altman is about to crack at the lab.
Pollan is understandably chary about the potential romanticism that lurks behind his conclusions. He’s a science writer, after all, working from evidence—and historically, those resisting scientific revolutions have sometimes descended into superstition, pseudoscience, and hate. There is a direct line of misappropriation from the theory of natural selection to the eugenics of Nazism and Jim Crow. Moreover, he admits, in Silicon Valley, any doubt about AI “can get you branded a specieist.”
But his caution misses something crucial. Computing began as a scientific revolution, to be sure, but these days it is primarily, exhaustingly, an economic one, wrapped in an aura of utopian mysticism. The chieftains of AI reject humanism not because it is anti-scientific, but because it is anti-business; workers are expensive. That’s why the recent marriage of big tech and right-wing politics might strike some as a relief. It’s simply more honest. Indeed, tech itself has become as spiritually reactionary as the political movement assimilating it—think of the Tolkienesque names, the space fantasies, the romantic nativism of the memes shared with equal enthusiasm by Donald Trump’s administration and Elon Musk.
What Pollan demonstrates is that AI is not incidental but fundamental to this violent alteration. That’s because, however hard it gets sold as a new beginning, this technology seems more like an end point—our final arrival, after 500 years, at the specific problem of what science and technology cannot do, cannot achieve, cannot solve.
The panic at this potential failure is central to the hysteria over AI. We’ve banked quite a lot on materialism, maybe too much. The decline of religion has left many people without beliefs through which we can touch transcendence. To what do we owe consciousness, if not God? The conquest of Mars and the achievement of the singularity are, like the nationalism resurgent across the globe, daydreams that offer a taste of that old comfort. Because AI truly does threaten to change our earthly conditions so radically, its purveyors are correspondingly grandiose in their rhetoric. Yet their heedless actions demonstrate only a belief that we are here in a finite place, with nothing sacred or divine in us—nothing that AI can’t re-create on a silicon chip. By that line of thinking, our only real task in this life would be to grab what we can, and laugh at the guy we took it from on the way out of the door.
A World Appears, with its admirable syncretic blend of empiricism and wonder before the limits of empiricism, steals back for humanity some of the sensation of miraculousness that this era has largely outsourced to technology. In the book’s introduction, Pollan describes a research project that tried and failed to answer the question of how “a particular piece of animal tissues generates the feeling of being alive.” That enduring mystery is what prompted Pollan to write this curious, compassionate book. Always to seek the answer, never to find it: That, of course, is what it means to be human. Some people find this fact terrifying. But there is also a pure exhilaration in standing on that last precipice, face-to-face with the question that exists beyond all other questions—which is to say, God.
The post The One Thing AI Will Never Do appeared first on The Atlantic.
