Data Centers

 In a podcast interview earlier this week about the current populist backlash against AI data centers, Jasmine Sun and Ezra Klein started off by trying to explain what it is like to actually live in the vicinity of one of these facilities. "What do they sound like?" Klein asked. Sun replied: "they sound like humming, buzzing, whirring."

B-R-R-R-R-R-R-R-R!

What are the machines saying, a hundred of them in one long room?

They must be talking to themselves, for I see no one else for them to talk to.

—as Ernest Crosby writes in his poem "The Machines"—part of his great collection of verses of social protest, Swords and Plowshares (1902). 

Crosby goes on, later in the poem: 

B-R-R-R-R-R-R-R!

What are the machines saying?

They are saying, "We are hungry.

We have eaten up the men and women (there is no longer a market for men and women, they come too high)—"

I (and others) have long made the point that much of the debate today over AI development recapitulates the same anxieties people had during the Industrial Revolution. In both cases, a core fear is the threat of job loss and displacement due to automation. 

As John Dos Passos writes in The 42nd Parallel: "The steamdrill had thrown thousands of miners out of work; the steamdrill had thrown a scare into all the miners of the West."

And Thomas Carlyle had written in the previous century—when machines and automation were at a still earlier phase of development: "The huge demon of Mechanism smokes and thunders, changing his shape like a very Proteus; and infallibly at every change of shape, oversetting whole multitudes of workmen [...] hurling them asunder, this way and that[.]"

The fear of losing white collar jobs (and, if the technology does actually lead to recursive self-improvement in robotics, as many boosters promise, blue collar ones as well) also obviously looms very large in today's backlash to AI data centers. 

But as Sun and Klein point out, people have a host of other concerns about the impact of these buildings on their communities too: the noise, as said at the outset; that ceaseless B-R-R-R-R that people were already mad at machines for as far back at 1902; the impact on the environment through electricity consumption—and other negative externalities. 

But above all, the data centers have come to serve as a metaphor for the imposition of technological change on people who don't see any personal benefit from it. 

As Jasmine Sun puts it in the interview: "The phrase that you hear a lot from A.I. critics is: Why is this being shoved down our throats? [...] It’s A.I. as sort of an avatar for a small group of Silicon Valley billionaires’ ability to impose their vision of the world onto everybody else without their consent."

This too has long been a complaint against the alienating effects of modern industrial development in general. 

I've quoted before, in this context, Dwight Macdonald's rendition of a key passage from Marx and Engels: "[T]his consolidation of what we ourselves produce into an objective power above us, growing out of our control, thwarting our expectations, bringing to naught our calculations, is one of the chief factors in historical development up to now." 

Indeed, this is precisely how many of us experience AI. It is not something we opted into through some sort of democratic process; it is something that a small group of billionaires has imposed upon our lives.

And while many of us—including me—find it interesting and even exciting as a marvel of technological ingenuity, this does not mean we are sure we would prefer it to exist, weighing all the arguments and costs and benefits on each side. 

One possible stance, then—the stance of the AI skeptics and populists—is simply, shut it down then! This is the course Robert Frost advocated in the previous century: 

Even while we talk, some chemist at Columbia

Is stealthily contriving wool from jute

That when let loose upon the grazing world

Will put ten thousand farmers out of sheep. 

[...] None should be as ingenious as he could,

Not if I had my say. Bounds should be set

To ingenuity for being so cruel

In bringing change unheralded[.]

But, Klein and Sun go on to observe, there are downsides to this approach as well. AI developers will reliably fire back: Look, we're scared of this technology too! We're just as frightened and alarmed about the future job loss as you are. But if we don't build it, someone else will! Do you want all this power to go to China? 

And this is the point at which the debate about AI starts to sound less like 19th and early 20th century arguments over automation and much more like mid-twentieth century debates about nuclear power. 

In the Ezra Klein interview, Jasmine Sun explains that there are essentially two arguments that AI developers will use to defend their decision to create a technology that they freely confess could prove highly dangerous and destabilizing to society: 

One is this sense of techno-determinism. It’s: Superintelligence is going to be built inevitably. There is no way it’s not going to happen. If it happens, [...] I want to maybe make it happen in the least bad way. [...]

Another is: This technology might kill us all, but it also might be really amazing. It might produce superabundance for everybody. We might be immortal. It might double everyone’s life span, cure all diseases, bring the cost of every consumer good — housing, energy, whatever — to near zero, and that would be utopia.

 People made these same arguments about atomic power in the 1950s: it could remove any constraints on the human energy supply, they said. Plus, there is the same "arms race" mentality that is driving AI proliferation today: if we don't build it, this dangerous technology will still exist—it will just all be concentrated in the hands of the Russians and the Chinese. 

But here, Dwight Macdonald's arguments against nuclear-powered modern development, in his classic tract The Root is Man, seem equally relevant to AI. 

As for the "techno-determinist" argument, Macdonald asks: 

"What becomes of the chief argument of Progressives,—that out of present evil will come future good—if we now confront the possibility that there may not be a future? [... I]nstead of manufacturing solar systems, man seems more likely to destroy his own little globe. And our sufferings, far from being for the benefit of those who are to come, are more likely to remove the first condition of their coming: the existence of an inhabited earth." 

And as for the argument from "superabundance," Macdonald has this to day: 

"[W]e, too, may perish in the next war because atomic fission is the latest stage of scientific discovery, and Progress depends on the advancement of science. But a simple-minded person might see in such modern truisms as that [...] atomic fission holds ultimate promise of the Abundant Life [...] a similarity to that promise of a better life in Heaven on which the Catholic Church banks so heavily."

Macdonald's conclusion from this is that we should in fact consider the Frost option: pause technological innovation. Set "bounds [...] to ingenuity." Do things on a smaller scale. 

It is clear, in short, that Macdonald would have favored a moratorium on data center construction, just as he favored a moratorium on atomic research and even on the methodology of mass production. 

"If automobiles cannot be made efficiently in small factories," he writes, "then let us make them inefficiently. If scientific research would be hampered in a small-unit society, then let us by all means hamper it." 

Instead of fixating on some future superabundant "utopia" that never seems to arrive, but which the determinists are always telling us will some day arrive to justify all our suffering and sacrifice—Macdonald urges us to embrace the "here and now." Let the morrow tend to itself. Sufficient unto the day is the evil thereof. Don't sacrifice present good to the future, because the future is unknowable, and all that talk of the superabundant utopia to come may be just so much pie in the sky. 

The analogy to nuclear power, though, perhaps suggests reasons to think that both the "determinists" and the pessimists about new technologies may sometimes be off. It's possible that neither the booster nor the doomer is correct and that a third option ultimately proves true: the new technology actually ends up changing less in human life than either side believed it would.

Here we are more than eighty years after the splitting of the atom, after all, and we have neither a utopia of superabundant cheap energy nor a barren, charred, uninhabitable hulk of a world. 

Indeed, more than half a century after the rise of civilian uses of atomic energy, people are still struggling to get nuclear power plants built. And this is not because we lack the technology, but because they are profoundly unpopular, and no one wants to live next to one—much like data centers. 

Nuclear power did not in fact solve all of humanity's energy needs and usher in utopia. It hasn't even eliminated our dependence on fossil fuels!

It quite possible too, then, that from the standpoint of fifty years on, both our present hand-wringing and our techno-optimism about AI alike will seem equally hyperbolic, and for the same reasons that atomic power ultimately proved far less transformative than many people had feared or hoped: 

It's actually hard to build things; and many of the barriers new technology faces stem not from hard limits in the domain of intellect or ingenuity but from the unwillingness of human communities to tolerate the new technology's presence in their neighborhoods. So much may really come down to that ceaseless B-R-R-R of the machines talking to each other after all.

As Klein puts it in the interview above: 

My view for a very long time has been that a lot of people in these [AI] companies overrate how much of the bottleneck in scientific and human progress is raw intelligence. [...T]he world is full of friction. [...]And just the amount of the world that is slowed down by: We don’t have any good ideas. We are out of ideas. Versus: It is hard to organize things amid humans, with raw materials, in bureaucracies, in organizations. [...] Intelligence is important, but it is not everything. And I think anybody who has been in organizations knows it’s actually less than you think it is.