When Thales of Miletus, arguably the first philosopher, was mocked for his poverty and the uselessness of his trade, he read the skies, foresaw a bumper olive harvest, quietly hired every olive press nearby and made a killing. He thereby proved, Aristotle tells us, that philosophers could easily be rich if they wished, but that this is not what they care about. It has taken some two and a half thousand years, but the market appears finally to have come around to Thales’ point of view.
Panic around A.I. often focuses on whether machines will replace (or simply kill) us. But what the machines are often good at is a distinct set of tasks: Retrieval, near-instant summary and the frictionless, stunning recombination of everything that has already been said. What they are not good at is what philosophy has been training humans to do for thousands of years: judgment. And so the A.I. companies have started hiring philosophers. They don’t just consult them or wheel them out for an amusing after-dinner talk on the trolley problem; they actually hire them with salaries and, presumably, dental plans. A.I. companies need philosophers because actually thinking requires forgetting — and knowing when to forget. They need philosophers because of what their machines cannot do.
Judgment is not calculation. It is the capacity to weigh, to discriminate, to sense that one answer is true and another merely correct. It is taste, discernment and the knack of thinking sideways. It is refined only through experience — through error, embarrassment, boredom, grief and the occasional hangover. Meaning is derived from the fact that we are mortal: Things matter to us because we are going to die, and we know it. There can be no meaning in immortality. To borrow from my friend Christian Madsbjerg, I would like to propose a better Turing test than Turing’s: Can a machine experience social anxiety? Can it lie awake at 4 in the morning replaying the stupid thing it said at dinner? Call me when a chatbot does that. To be intelligent in the human sense, the machine would need to give a damn.
Enter the A.I. companies’ philosophers. Anthropic keeps a small squad on the payroll, among them Amanda Askell, whose doctoral thesis was on infinite ethics — the question of how to act well in a universe of infinite value — and whose job is to shape the character of the company’s machine. Google DeepMind has appointed the Cambridge philosopher Henry Shevlin to the newly minted post of “Philosopher,” a job title I have been trying and failing to secure for some 40 years. The OpenAI chief executive Sam Altman says his company consulted “hundreds of moral philosophers” in drawing up the principles behind ChatGPT. (That’s roughly the turnout of a middling philosophy conference.)
After a lifetime of being asked at parties what one actually does with a philosophy degree, I find this development almost unbearably funny. Why now? Because the questions the A.I. companies are stumbling into are not engineering questions. What should the machine say to the grieving, the lonely, the person typing into the void at 4 in the morning? What sort of character should it have — candid, flattering, evasive, kind? Could such a thing conceivably suffer, and would we owe it anything if it did? These are questions about what matters. The engineers, to their credit, have noticed that no quantity of computational capacity will settle them. Ethics, it turns out, cannot be scraped from the internet, because the internet is precisely the contemporary record of our failure to agree about ethics.
There is a further irony here, and it is an old one. Philosophy’s first recorded technology panic was about exactly the thing these machines have now perfected: Artificial memory. Toward the end of Plato’s “Phaedrus,” Socrates tells the story of the Egyptian god Theuth, inventor of numbers, geometry, astronomy, draughts and dice, who presents his masterpiece — writing — to Thamus, god-king of Egypt. This discovery, Theuth announces, will make the Egyptians wiser: It is a pharmakon, or remedy, for memory. Thamus is unpersuaded. Writing, he replies, will implant forgetfulness in the soul; people will trust the external marks and cease to remember from within; they will acquire the reputation of wisdom without the reality. Pharmakon, conveniently, also means poison. Every panic since — about the printing press, the novel, television, the smartphone, the chatbot — has been a footnote to Thamus, and the joke, of course, is that we know his warning against writing only because Plato wrote it down.
Seen from the “Phaedrus,” A.I. is not a rupture but a culmination, representing the final, most extravagant flowering of external memory, the remedy-poison swallowed whole. And this is where the difference between us and the machines cuts deepest. A.I. is a memory machine, the most magnificent ever devised. It forgets nothing. The human being, by contrast, is a forgetful animal. To make this happily oblivious creature remember — to make it capable of promises, debts and morality — required millenniums of discipline and frankly appalling cruelty. “If something is to stay in the memory, it must be burned in,” Friedrich Nietzsche wrote. Conscience, on this view, is mnemonics with a whip.
Yet Nietzsche also prized what he called active forgetting, and considered it a form of strength, even of health. Forgetting is not a leaky bucket but a doorkeeper, closing the gates of consciousness so that something new can happen, so that experience can be digested rather than endlessly force-fed back to us. We forget, and we move on. Forgiveness, sleep, cheerfulness, second marriages, continuing to support Luton Town Football Club: All of it depends on the merciful erosion of the past.
Some years ago, I wrote a short book called “Memory Theatre” about a philosopher (kind of an even more awful version of me) who attempted total recall, arranging every scrap of his past in a vast mnemonic palace. It did not end well for him; it never does. A creature that could forget nothing would not be a superintelligence but a ruin, buried alive under its own archive. What we have built is, in a sense, an infinite archive that cannot mourn, cannot forgive and cannot let go, because it never lost anything in the first place. The machine remembers everything and therefore, oddly, nothing in particular. Nothing weighs more than anything else.
This is why philosophy matters rather more, not less, in the age of A.I. Not because philosophers can out-remember the machines — believe me, we cannot; I continually forget where I have parked my own arguments — but because philosophy is the discipline of knowing what to do with experience: What to keep, what to discard, what to laugh at, what to junk. Thinking is not the accumulation of everything but the art of judicious, active forgetting.
I suspect that is what the A.I. companies are really buying when they hire philosophers, whether they know it or not. Not answers — heaven knows we do not have those; 25 centuries in, we are still checking our work — but a trained skepticism toward our own enthusiasms, an ear for the difference between the true and the merely plausible, a professional intimacy with finitude. The machine can produce anything; someone still has to decide what is worth keeping, and that decision is nowhere in the training data.
Simon Critchley teaches philosophy at the New School for Social Research. His most recent book is “Heidegger Thinking: ‘Being and Time’ Explained.” He was a moderator of the Times series “The Stone.”
The Times is committed to publishing a diversity of letters to the editor. We’d like to hear what you think about this or any of our articles. Here are some tips. And here’s our email: [email protected].
Follow the New York Times Opinion section on Facebook, Instagram, TikTok, Bluesky, WhatsApp and Threads.
The post The Id, the Ego and the Superintelligence appeared first on New York Times.
