Ethics Won’t Save Us From AI

When it comes to new technology, including artificial intelligence, it’s easy to succumb to the temptations of doom or boom. Ours is the best of times, ours is the worst of times — the debate is predictable.

Others, seeking to escape bifurcation, propose a third way grounded in some principle, method, or procedure of neutral universality that avoids deadlock while delivering a rational solution acceptable to all, at least in principle. This motivation is understandable, even praiseworthy. But such projects inevitably fail: a reasonable account of the dangers and promises of technology cannot be neutral, because any principled argument about benefits and harms must be grounded upon a robust, thick notion of the human good, even though thick commitments are rarely acceptable to all.

One such third way — with philosophy and classical liberalism as universal methods — was proposed earlier this year by Brendan McCord in a manifesto of sorts titled “Existential Pessimism vs. Accelerationism: Why Tech Needs a Rational, Humanistic ‘Third Way.’” The piece was published on the Substack of the Cosmos Institute, McCord’s new Austin-based nonprofit, which is dedicated to articulating this third way for our current technological moment. Unfortunately, the manifesto’s view of philosophy is too thin to deliver what he believes it must do. A better alternative is on offer.

A successful tech entrepreneur, McCord has immersed himself in the study of philosophy “to understand the good life and the good society,” he writes in the newsletter post. “Only if technologists seek this wisdom,” he declares in bold italics, “can we ensure human flourishing in the modern technological age.” To that end, he has launched a variety of projects for tech entrepreneurs to philosophically discover the “mean between two extremes,” given the “urgency” of a moment when “humanity is setting foot” on the “new continent” of AI.

McCord, who is especially focused on the tech world’s varying moods about artificial intelligence, outlines several schools of doom and boom before suggesting philosophy as the “rational, humanistic” mean between “existential pessimism” and “accelerationism.” In his description, existential pessimism views AI apocalyptically, as an existential threat, although the three types of pessimism — rationalism, long-termism, and effective altruism — have differing fears and assessments of risk. Briefly, rationalism wishes to preserve human life and mitigate suffering while facing the challenge of creating and maintaining AI friendly to humans. Long-termism conceives of AI as an existential threat to human life in the future. Effective altruism judges even the smallest chance of AI doom unacceptable, since that small risk multiplied by infinite negative weight — extinction — is infinite risk. The boom side, accelerationism, is a “techno-optimistic” mood celebrating AI’s promise to “unlock man’s potential to overcome himself and become something higher, better, and more refined…. Technology replaces religion.”

Agreeing with the boomers, McCord thinks the future is not determined, so we need not passively await our fate but should responsibly direct ourselves and civilization. If we act well and intelligently, AI will “make us safer and wealthier” and “help us better flourish.” At the same time, progress is not guaranteed, and the “godlike powers” of AI require that we direct it to humanistic purposes while avoiding “diminished expectations for human excellence, a loss of meaning and ordinary purpose.”

How are we to be free, responsible, and flourishing in our use of AI? These are the relevant questions, to be sure, and far more fruitful than boom or doom simplicity, and certainly not questions AI can answer for us. So McCord correctly saves a spot for philosophy and other humanistic disciplines. As Leon Wieseltier noted in a 2013 New Republic article, “The question of the place of science in knowledge, and in society, and in life, is not a scientific question…. It is not for science to say whether science belongs in morality and politics and art. Those are philosophical matters, and science is not philosophy.” McCord makes a roughly similar point, arguing that “only through philosophy … can we liberate ourselves from the dogmas and prejudices that motivate our judgments,” and, moreover, the inability of philosophy to reach settled conclusions “might amount to the only genuine knowledge available to us here and now.” Uncertainty and inconclusiveness are signs of wisdom.

So far, so good. There is wisdom in a Socratic sensibility, a sense of philosophy as therapy of desire, tempering our dogmatic confidence and intellectual conformity. But Socrates’ method of question and refutation is just as likely to end in confusion as in progress toward knowledge. The early Platonic dialogues reveal Plato at his most Socratic, a mode diminishing in the middle and later dialogues, but the texts invariably end in puzzlement, wonder, or an admission of not knowing. Book I of the Republic is a fine example, with Socrates defeating definitions of justice provided by his interlocutors before concluding that “I know nothing. So long as I do not know what the just is, I shall hardly know whether it is a virtue or not.”

McCord understands reason in this way, as a neutral instrument, a referee rather than a player, very much in the skeptical mode of Socrates separating sense from nonsense: “The clarity we need now cannot be built on unmoored passion and devastating memes; reason must sit in judgment and act as an umpire among the various competing views.”

This, however, is a reason far too thin to distinguish good from bad. Reason is here essentially calculative and instrumental. When presented with an already determined goal, technical rationality accomplishes the goal, somewhat like the current capacities of machine learning. But instrumental rationality is incompetent to resolve disputes about the meaning, content, or value of goals and purposes themselves. As Thomas Aquinas notes in his commentary on the Nicomachean Ethics, scientific intellect does not “establish but only beholds” the order it seeks to know in the real world, logic forms only an internal coherence of meanings, and technical reason arranges things through “the mechanical arts.” By contrast, there is a distinct type and form of intellect, that of practical reason, the domain of prudence, which considers and guides “the order of voluntary actions,” those human acts springing from our will in keeping with the truth of what ought to be done. Prudence, however, must have an account of the human good — a thick account, one with real and genuine content — in order to direct us to, or in service of, the good. Absent a view of the good, the domains of science, logic, and technique are directionless, bereft of a point or goal.

McCord wants philosophy to guide our choices about AI, but he views reason as a technologist views reason, as instrumental, a tool dealing with the means, although his means is precisely that of the skeptical umpire. But an umpire who understands only the physics of a ball in motion, even if thoroughly and entirely logical, is incompetent as an umpire. An umpire must know the ends and purposes of the game, and commit to the game as worth pursuing as part of human activity. McCord views reason almost like a large language model, but an LLM knows nothing about what ought to be. Isn’t the challenge facing us determining what ought to be? If we are to govern our flourishing, we’ll need an account of flourishing itself, which entails real commitments and determinations about human nature and the purpose of being a human at all.

It’s not accidental that McCord resonates with the Enlightenment and classical liberalism, which saw that the power and light of reason grow as humans emancipate themselves from custom and prejudice and the mediation of tradition, authority, and religion. But absent those inherited meanings, reason is simply an instrument of calculation, a technique of analysis — it’s an artificial view of intelligence, meaning it is alien or false to the human condition. Actual human intelligence, particularly in its moral role, operates only while embedded in what traditions we have received and embodied in concrete judgments of value. McCord rather blithely suggests that the past looks benighted compared to the present, and gives credit to the “twin forces of classical liberalism and science” for our advance. But these forces drew power not from their substance — they eschew inherited, thick substance — but from their procedures. They are thin by design. Science is skeptical, and its early modern version defined itself through rejecting any thick notion of the good inherited from Aristotelian philosophy in favor of the neutral, dispassionate view from nowhere. Similarly, the modus vivendi of liberalism works by not having views about God, religion, human nature, and the human good, bracketing those concerns to our private lives while favoring neutrality and procedures in the public square. Reason, for McCord, has an amazing power to sort, but it is entirely procedural.

The English political philosopher Michael Oakeshott describes such a thin view of reason in Rationalism in Politics:

There are some minds which give us the sense that they have passed through an elaborate education which was designed to initiate them into the traditions and achievements of their civilization; the immediate impression we have of them is an impression of cultivation, of the enjoyment of an inheritance. But this is not so with the mind of the Rationalist, which impresses us as, at best, a finely-tempered, neutral instrument, as a well-trained rather than as an educated mind.

Rationalism tends by its inner logic to privilege the principle of the empty head, which assumes that objectivity depends on us having no previous views, no commitments, and no authority, but only the unfettered power of the light of reason to guide us.

This is not how reason actually works, nor could it ever work in this rationalist fever-dream fashion. Rationality is tradition-constituted if it is to have any meaning, content, or substance; it does not stand above the fray of debate and conversation looking down in judgment. Rationality, as thinkers such as Roger Scruton, Hans-Georg Gadamer, and Alasdair MacIntyre correctly argue, operates only through — rather than upon — those connected arguments, disagreements, crises, and advancements that constitute a tradition. Absent an inherited tradition — a culture — rationality remains an empty power, a faculty, a machine capable of work but without material upon which to work. Tradition precedes us as prejudgments we inherit, but those prejudgments are the horizon within which we think, and without which we have nothing to think about, leaving only empty formalism. Tradition, quite obviously, is not infallible, and some forms of common sense are common nonsense, but it remains the case that the starting place of thought is not the power of rationality itself but the content of a culture and its pathways.

The view from nowhere, the tennis umpire in his high chair watching rather than playing and contesting the match, is all but doomed to fail, just as the Enlightenment moral project, McCord’s inspiration, was doomed from the start. The Enlightenment attempted ethics without teleology, disdaining telos — the human good — as a superstitious and irrelevant remnant of the medieval schools. Liberalism and rationalism thought we would progress if ethics bracketed substantive views about the human good or human flourishing, a perspective entirely alien to Aristotle and medieval philosophers. (Unsurprisingly, McCord’s self-created syllabus or canon of philosophy entirely skips the Middle Ages, leaping from the ancient directly to the modern world.) In After Virtue, MacIntyre views the liberal, rationalist project as incoherent. Ethical reflection, properly understood, is an exercise in understanding humans as they are, and humans as they could be if they actualized their telos, with ethics as the way to move from how we are to how we could be.

Now, along with teleology, jettison any and every thick conception of the good caught up in teleology: what is left? Humans as they are, that is, not yet flourishing, not yet well, not yet wise, not yet good — which is self-evidently the very condition requiring moral and ethical reasoning to resolve. But where, then, does morality come from? Morality is obviously not related to our telos, since that has been jettisoned. And morality cannot be derived from human nature as it is: precisely what needs improving is that facticity, and so it cannot be a normative source or goal. From what does it come, then? Reason? But reason is merely a power, an instrument, lacking content in and of itself. There remains only a powerful tool whirring in circles on nothing and about nothing.

Brendan McCord, quite admirably, has devoted serious effort to reading philosophy. He’s launched and sold successful tech startups, is involved in a variety of think tanks, university programs, and more, all while devoting hours a day to reading — a sort of great books program, as it were. But this approach to reading runs aground in the same way that MacIntyre views the Enlightenment moral project. The great books disagree with each other, and recognizing this disagreement is as likely to foster relativism or nihilism in readers as humility, wisdom, or serious commitment — let alone the drive to prospect through disagreements to the truth.

Wisely reading these books requires having a tradition, a narrative, into which to place, weigh, and evaluate arguments. Tradition isn’t a trap, it’s not entirely determinative of our reading, and texts often challenge our tradition, pushing us to develop and grow it so it stays alive and sprightly. To approach texts without a way to order their claims in terms of weight, value, truth, and purpose is to be adrift, not knowing up from down, near from far, left from right, light from heavy. The merely calculative, value-neutral aspects of reason, such as logic or analytical power, are incapable of determining value or worth.

So while I find myself offering a cheer for McCord, it’s a warbled, muted cheer for an incomplete project. I’m all for a third, sober path; I’m all for reason; and I’m all for philosophy taking its rightful place in the future of humanity. But reason as a pure umpire, reason as rationalism, reason as science and liberalism — such reason lacks the ability to constitute a third way between doom and boom because it is devoid of content. McCord cannot simply do philosophy as a study of thought thinking thought or a collection of books chronologically arranged.

Philosophy is going to need to do something far more robust: develop a metaphysics, ethics, epistemology, anthropology, and politics coherent in themselves, coherent with each other, and corresponding to human nature as it actually is. McCord is going to need some content, and it needs be a content up to the challenge of the time.

Here is a suggestion: returning to a tradition that developed and continues to develop all those areas of content, which is the tradition explicitly rejected by the early modern scientists and classical liberals for whom McCord cheers but who have proven almost entirely impotent to address or resolve the debates between pessimism and accelerationism. I speak, of course, of the Aristotelian tradition.

In the absence of that tradition, or even a rival tradition with content of its own, I do not find myself sympathetic with McCord’s understanding of hope. He writes:

Although reason is the medium through which philosophy is practiced, hope is the foundation upon which the most powerful philosophy rests: hope in humanity’s ability to uncover powerful truths that lead to our flourishing. The leading figures of the Enlightenment — Bacon and Descartes, the philosophes, the rationalists, and the positivists — ​ushered in the modern age with hope. Their quest to “relieve man’s estate” would have seemed daunting, improbable, even impossible.

This, it seems to me, conflates hope and optimism, or hope and confidence. Hope, in the classical sense, trusts in the agency of another, especially God. In that account, hope in is quite distinct from hope that. One hopes in another agent at work, willing what you will, caring for what you care for. Hoping that is a wish or a confidence that some desired outcomes or states of affairs will occur or be accomplished. I hope in God to save me, but I hope that my team wins the World Series, hope that my child will recover from an illness, and hope that the human future is brightened by wise use of AI. But this is either wish or confidence — optimism — rather than hope in its older, more venerable sense.

With McCord, I hope that a mean between boom and doom will be found and the human future with AI will be bright, and I hope that philosophy will play its rightful role in bringing about that future. But I cannot hope in philosophy accomplishing this, given the ghostly non-substantiality of his philosophy. Logic, analysis, and rationality are unable to determine the good and can just as likely cause our doom as our boom; they are valueless tools willingly in service of any master. Unless and until we are willing to reason about the good in a committed, substantial way, ethics is doomed to fail. I hope it is only our ethics and not we ourselves so doomed.

Of course, I must admit that the Aristotelian tradition is a tradition among others, and not persuasive to everyone. I think it is true, reasonable, and the best account. Not all do, and some, perhaps many, will find it false, outdated, or incoherent, opting instead for another tradition of rationality. I admit this, and it poses no challenge to my account, admitting, as I do, that there is no view from nowhere, no reason as umpire calling scores. If we are to use philosophy to shape the human future, we will have to do so not as rational subjects but as humans, who, whether we know it or not, live according to a vision of what is good. Aristotelianism is reasonable, to be sure, but it is not philosophy just per se. But there is no such thing as philosophy just per se anyway.