AI and the Crisis of ‘Classical Liberalism’

What do you do when the Invisible Hand is building a totalitarian one-world government?

This week, the American Enterprise Institute’s new Council on AI Ethics released a “founding document” outlining its principles and goals in the pages of The New Atlantis. The body’s creation responded to an earlier call by AEI fellow M. Anthony Mills for a “President’s Council on AI,” modeled on the W. Bush-era President’s Council on Bioethics. The Trump White House has not established such an entity, so AEI decided to take the initiative and convene a group of mostly socially conservative thinkers to explore “the deepest questions AI is raising about human meaning and purpose.”

The document begins by announcing that it will avoid addressing “practical issues like job loss, bias, and the tradeoff in AI development between speed and safety,” and will eschew regulatory debates in favor of philosophical reflection. Although the rationale offered for this reticence is plausible enough, it is also difficult not to see it as evidence of the emerging tension within the right between religious and social conservatives on one hand and tech accelerationists on the other. In an article for Compact this week, editor-at-large Greg Conti identified this emerging “cleavage” around AI as “the most momentous determinant of the future of the American right wing.”

The Bush administration’s Council on Bioethics, which like the new AEI council was made up mostly of social and religious conservatives, was distrusted by liberals, who saw it as a means of pursuing a regressive theocratic agenda and stymying projects like stem cell research; Barack Obama disbanded the body soon after assuming the presidency. Today, it is left-wing political figures who are more likely to advocate for moratoria on AI development, while the Trump administration has largely taken a “let it rip” approach, a few deviations notwithstanding. But as I’ve argued previously, progressive AI critics often seem to offer little more than “irritable mental gestures” by way of explaining their stance, eliding any substantive account of the challenges of the new technology. In contrast (as I’ve also said before) religious thinkers—not least Pope Leo—have been more perspicacious.

The AEI is known less as a hub of religious or social conservatism than as a stalwart of unwavering laissez-faire advocacy. On this ground, one might find the profound reservations about AI enunciated by the new Council somewhat surprising. Today’s tech rightists, who usually define themselves as free-market absolutists, are wont to dismiss any qualms about the technology as a stalking horse for a “woke” or “communist” agenda. If you fail to revel in the glories of the latest model, for the e/acc’ers on X, you are probably a useful idiot for the totalitarian “safetyist” regime bent on stamping out all innovation.

Even if it declines to endorse any particular regulations, the new Council’s findings at minimum suggest that there is sound reason for the sort of misgivings that might justify limiting AI development. At the outset, the document cites the bioethicist Leon Kass’s notion of the “wisdom of repugnance,” which holds that “the gut reactions of ordinary citizens to the transformations they see occurring around them are worthy starting points for ethical reflection.” As noted, today’s visceral anti-AI “gut reactions” are often coming from the left of the spectrum, which tends to translate them into proposals like the moratorium on data centers supported by Bernie Sanders and AOC. Kass influenced the attempt at a federal ban on human cloning two decades ago, and it isn’t hard, reading the AEI document, to see how its arguments could be marshaled in favor of a ban on superintelligence.

Clearly, the current AI trajectory is bringing to the fore once again the tensions between those two unsteady legs of the fusionist stool: business and religion. But what is more interesting are the dilemmas AI is creating within the GOP business wing’s standard “classical liberal,” free-market framework. Exhibit A of these dilemmas is the influential work of Dean W. Ball, self-described classical liberal and former Trump administration AI advisor and author of its laissez-faire AI Action Plan. Since leaving the administration, Ball has criticized its confusing and haphazard attempts to rein in the industry, in particular its main bugbear Anthropic. Ball has said that he “opposes literally almost all AI regulation,” but he co-authored a New York Times op-ed earlier this year with his Biden administration counterpart, Ben Buchanan, arguing for “appropriate guardrails,” especially when it came to “catastrophic risk.”

The appearance of the op-ed prompted many of Ball’s erstwhile e/acc allies to denounce him on X as a turncoat, just another woke “safetyist” who wanted to usher in Luddite totalitarianism. Ball’s lengthy defense of his position was revealing, but the following passage particularly stood out:

Fundamentally, I view the state as a kind of tragic necessity, something we must merely tolerate, because without it, no civilization we can conceive of is possible. I think it is possible AI will change this—that “AI,” very broadly conceived, may one day be able to conduct enough of the core functions of the state that new architectures of civilization will become possible. As a “classical liberal,” I am intrigued by this. And yet I am also a conservative, and as a conservative, this notion makes me fearful. The conservative in me is skeptical of top-down projects to remake the world, including projects with a libertarian bent to them.

It is obvious enough why there might be a tension between conservatism and any scheme to develop superintelligence. What seems more noteworthy here is that Ball defines his fascination with AI “conducting the core functions of the state” as an expression of his classical liberal sensibility, but then goes on to worry that this vision also amounts to a “top-down project to remake the world.” (And the thing is: It obviously is that.) So the question is: Just how could a “classical liberal” be seduced by such a scheme? Isn’t classical liberalism, not just conservatism, intrinsically hostile to “top-down projects”?

What Ball seems to be alluding to in the line about “new architectures of civilization” is an idea articulated decades ago by philosopher Nick Bostrom, whose 2014 book Superintelligence exercised a profound impact on the entire AI industry and the culture around it.* In it, Bostrom summarized a concept he had already been discussing for some time: the singleton, which he defines as “a world order in which there is at the global level a single decision-making agency.” In the chapter where he discusses this notion, he is exploring a hypothetical scenario that anticipates, in broad strokes, the current AI arms race as its main participants conceive it. Bostrom asks: “Will one machine intelligence project get so far ahead of the competition that it gets a decisive strategic advantage—that is, a level of technological and other advantages sufficient to enable it to achieve complete world domination?” The resulting triumph, he suggests, would set the stage for establishing a “singleton.”

With that in mind, we can define the AI era crisis of classical liberalism as follows. Its standard position is to demand minimal government interference in private industry on the grounds that letting competitive enterprise prosper will make everyone richer; any such intervention, it is claimed, is a slippery slope toward totalitarian statist tyranny. The problem today is that the leaders of the most vaunted private industry regularly announce that their technology, rather than making everyone richer, may instead create a “permanent underclass,” necessitating things like UBI, a point recently reiterated not by “woke” Anthropic types but by tech right figurehead Elon Musk; moreover, influenced as they are by Bostrom, many in the industry also believe that what they are creating is an instrument of technocratic world government of the sort classical liberals warn us against. Hence, as Ball recognizes, the laissez-faire approach to the current AI trajectory turns out to also amount to a “top-down project to remake the world.” Oops!

If you have the sang froid of a Nick Land, none of this is likely to bother you too much. But I would guess most classical liberals don’t—which is perhaps why AEI is taking a stance closer to the conservative one on stem cell research than the conservative one on fossil fuels. My further suggestion is that this impasse might be an opportunity for classical liberals to rethink their fundamental account of the relationship between human flourishing and the state, and much else—but that’s probably too much to ask for now.

*Update: In subsequent interactions here and on X, Dean Ball clarified to me that he “do[es] not believe a Bostrominian singleton is a likely outcome,” and went on to explain his views at some length; his reply to this post is worth a read. As I attempted to clarify to Dean, I didn’t intend this post primarily as a critique of his work; if anything, the passage I quoted seemed, as I said, “revealing” to me because it laid out very cogently some of the tensions I was trying to explore, which in more glib articulations of AI accelerationism are usually passed over entirely. In his reply to the post, Dean said that the best precedent for thinking about AI was the Industrial Revolution, in which “the world was remade in a way no one could ever have quite predicted, much of it was done to people rather than by them, and ultimately, it ended up for the better. Yet a reasonable person at the time, including perhaps myself (or at least the conservative in me) would not have been crazy to oppose them.” He went on: “And I ask myself: what should the classical liberal of the time thought about the Industrial Revolution? Should they have welcomed the rearchitecture of the world? Hence the tension.” This is all well put.

Support Compact

Thanks for reading this Substack, a project of Compact magazine. If you enjoy our work, please help us keep doing what we’re doing—and ideally, expand.

Specifically, there are a few big ways you can help us right now:

• If you’re not a subscriber to the magazine yet, click here to take advantage of an unbeatable limited-time deal for new subscribers.

• If you already subscribe, please consider donating to our summer campaign. Every tax-deductible donation helps us meet our goal.

• Share our stuff with your friends and followers, online and off!

This week in Compact

• Greg Conti reviews books by Laura Field and Harvey Mansfield

• Nathan Pinkoski looks back on Brexit, ten years on

• Michael Tracey makes the case against podcasts

• I assess the triumph of the “mini-Mamdanis”

• Ashley Frawley argues against “mental health awareness”

• Germán S. Díaz del Castillo examines Mexico’s oil strategy