What Can We DO About It?

 Earlier this week, a group of nearly 200 economists and businesspeople signed an open letter on the future of AI, entitled "We Must Act Now." 

The statement is a model of strategic ambiguity—that sort of deliberate fuzziness as to key details that is often required to reach a consensus viewpoint that a large committee of people might be willing to get behind. 

The benefit of this, as a friend insisted to me, is that it increases the statement's impact by getting a larger group of people on board. 

The downside is plainly that it says less and less at the level of actual content. 

The statement declares that AI's rapid development "could bring risks," but that it could also bring "opportunities." Oh, could it? We needed 200 economists to tell us that? 

We also learn that people "must act now"—hence the title. But—to do what, exactly?

"[T]o build the incentives, guardrails, and institutions needed to steer AI in a direction that complements humans and benefits society."

To build what now? In order to do what was that? And how are we supposed to do that, again? 

I am here reminded of Jim Dixon's embarrassed reflection on an academic article he'd written, in Kingsley Amis's Lucky Jim: 

"Dixon had read, or begun to read, dozens like it, but his own seemed worse than most in its air of being convinced of its own usefulness and significance. 'In considering this strangely neglected topic,' it began. This what neglected topic? This strangely what topic? This strangely neglected what?"  

The economics blogger Noah Smith was not impressed with the economists' statement either. 

After quoting it in full, he observes on his Substack:

"That’s it. That’s all it is. It doesn’t say what our action ought to be, only that 'we must act'. There’s no appendix, no longer manifesto attached below. It just says AI is getting good, AI could be economically important, AI could take people’s jobs and/or make us a lot richer, and that we have to do something to make sure the AI age turns out alright."

The statement, like so many others of its kind, "might(publishers note)be entitled a wraith's progress," as E.E. Cummings once wrote of Joe Gould's never-materialized "oral history"—since it, like so many of its kind, just sort of trails off at the end, without delivering what it purported to promise—namely, a practical proposal for action. 

The problem with this sort of ambiguity is not only that it is empty and contentless—but that it can subsequently be put to sinister uses. If a statement means nothing and everything, then what is to stop someone from invoking it later on, on behalf of their perhaps half-baked or dangerous policy ideas?

As Smith goes on to argue in his blog post, indeed, to the extent the statement is making any concrete suggestions with real content, they might actually be bad ones. 

As Dwight Macdonald argues in The Root is Man, it's likely that the historical Karl Marx would have opposed the Soviet government as it developed under Stalin—just as he opposed the "Gotha program" in his lifetime. But since he never spelled out exactly what his program was supposed to be, he left himself vulnerable to being "misunderstood" in this way. 

"How could his closest followers mistake so grievously his teachings?" Macdonald asks; then answers his own question: it's because "up to then Marx had not put down on paper with any concreteness what he meant by 'Communism'." 

The irony of Macdonald's essay is that he then leaves himself open to the charge of doing the same thing.

When it finally comes time for Macdonald to define his own alternative "radical" humanist vision that will bring "libertarian socialism" back to first principles, he writes with the same ambiguity he faults in Marx. 

"The specific forms of action, and the organizations to carry them out, are yet to be created," he says. "We seem to be in the early stages of a new concept of revolutionary and socialist politics, where we can hope for the present only to clear the ground, to criticize the old methods that have landed us in a blind alley, and to grope in a new direction." 

But this is just sounding like yet another "wraith's progress"! 

Pages earlier, Macdonald was just criticizing Karl Marx on exactly these grounds. He faults him for always claiming that "bourgeois ethics" was about to be superseded by some glorious new future "proletarian" or socialist ethics—but that he couldn't anticipate in advance what these socialist ethics would be. 

How, Macdonald rightly wanted to know, could Marx expect people to follow him if he couldn't even explain where they were going? Why should we assume we want "socialist values" at all if you can't articulate some sense ahead of time of what they might be? 

"How do we know the struggle is worth it," Macdonald rightly asks, "unless we get some idea of what these fine new values are?" 

But as we see, as soon as it comes time for Macdonald to articulate his own program, he retreats as well into quite similar clouds of nebulosity. He says: I can't really get into specifics yet; I'm just trying to "grope in a new direction"; I'm just criticizing what came before. 

In other words, he pulls out a version of that Paul Nizan line that annoyed me so much and that seemed to me so laden with totalitarian potential: "Will I be asked what human life would be made of and what it would be like [after the revolution]? I do not see it clearly yet. I am groping my way[.]" 

I've had enough with people "groping their way" toward things—insisting I go along with them while refusing to tell me where we are going. 

To quote David Mamet: "What would have prevented them from a clear statement of their goals had those goals been realizable and laudable?"

This too, plainly, is part of the problem with the AI statement. If its many signers knew what they were doing and where they were proposing to lead us, why couldn't they just come out and tell us? Why retreat behind such opaque ambiguity? 

Is it because their actual plan is unsympathetic? They want to regulate so as to corner the market and box out competitors? I doubt it; there were a lot of people with very different incentives who signed on.

More likely the reason they don't tell us what their plan is, is that they don't have one. 

They, like us, don't really know what to do about AI. 

They, like us, are perhaps partly still hoping the whole thing just blows over somehow and turns out to be not that big of a deal. It "could bring risks," they say. But then again, it might not. 

They like, us, are disturbed by the prospect of mass unemployment caused by technological displacement, and think that something ought to be done to ward off this danger, but they aren't sure what. 

And so, like us, they resort to hand-wringing. 

Which I can hardly fault them for—it's all I can think to do about AI too. 

But if the point of this statement was to in some sense offer an "expert" perspective on the issue, we can well wonder what experts are good for, if they don't actually tell us anything we didn't already know, and they are just as helpless before the Moloch of the new machine as we are. 

Macdonald was well aware of all these problems. He knows full well he has left his readers with little more than a "wraith's progress" when it comes to providing an actual radical program or "basis for political action," as he puts it. 

"When I first began politics," he writes, "readers used to ask me all the time what they could DO?" and he cites one sardonic reader who comments on what a "damn good article" The Root is Man really is, but then confesses it left him with no idea how to respond to the question of "what can we DO" about it—except maybe something to do with founding "cooperatives." 

In the endnotes, in which Macdonald added his political afterthoughts, appended some years later, he rather endearingly and disarmingly admits that he just doesn't really know what to DO about any of this either. He says he never found a new "basis for political action"—"and I can't say I'm still looking very hard for it." 

(These notes date from the period when Macdonald, as he repeatedly says, basically "lost interest in politics"—though fear not; we would regain it in the '60s, when he would go on to share a stage with Robert Lowell and Norman Mailer at anti–Vietnam war protests, etc.)

Macdonald, to his credit, even entertains the possibility I just raised above—namely, that perhaps he doesn't know what to DO about things because there is no actual answer to the problem. 

"We must learn to live with contradictions," he writes, "to have faith in skepticism, to advance toward the solution of a problem by admitting as a possibility that [...] it may be insoluble. [... I]t is better to admit ignorance and leave questions open[.]"

And as he wrote later on, about the early Cold War, "Perhaps there is no longer any solution to these agonizing problems." 

Perhaps that, too, I say, is why the AI statement is so devoid of content. They reveal no grand plan for regulating or "steering" AI—because there is none to be discovered. 

They wring their hands about it because that's all any of us can do think to do about it. 

I'm reminded here of that passage in Martin Amis's Einstein's Monsters, about another potentially world-changing or world-destroying technology: "I am merely going on about nuclear weapons; I don't know what to do about them."

Perhaps the economists' statement doesn't ever tell us what we can DO about AI, then—because there is no answer to that question. 

To quote the immortal words of James Thomson (a.k.a. Bysshe Vanolis): 

Perhaps all the oracles are dumb or cheat

    Because they have no secret to express;

  [...] none can pierce the vast black veil uncertain

  Because there is no light beyond the curtain;

    [...] all is vanity and nothingness.