Hi there, argmin readers! Today’s post is a late and brief reaction to a tiny subplot of yesterday’s historic election: the widespread rejection of AI datacenters. It was inspired by an insightful Dan Davies thread on bluesky that I have been unable to get out of my head.
Elections are a muddled measurement of “public opinion,” but they’re probably the only real measurement we get. And with every surprising measurement, a good rational agent must reflect on its beliefs. So what should we passive rational observers make of this historically large swing towards the Democratic Party in yesterday’s vote?
Obviously, this is a major repudiation of Donald Trump, his policies, and the Republican Party. However, you readers probably don’t come here for my political takes, and I don’t want to use this Substack for them anyway. Even if it wasn’t the biggest issue on the ballot, something closer to home is worth pondering. The American people are loud and clear that they hate AI.
Nearly fifty measures explicitly about building data centers appeared on ballots across the nation. All of them look like they will vote to limit further construction. A datacenter moratorium was part of Gina Hinojosa’s platform in her successful bid to become the first Democratic Party governor of Texas in over 30 years. The anti-datacenter position was undeniably the winning one. Poll results from this last summer showed people would rather have a coal plant in their backyard than a datacenter. Now elections confirm these sentiments were real.
We could try to convince ourselves that this sentiment is just about datacenters themselves. They are gross, gigantic buildings that employ no one, make a ton of noise, and are often powered with dirty energy. Who would want one of those? But datacenters are also a big, ugly, legible signal of a deeply unpopular industry.
At some point, we techy folks have to reckon with how much people hate what we build. We can point to revealed preferences all we want (people are adopting it! It’s making money!), but we did this during the social media boom of the 2010s (people spend a ton of time doomscrolling, therefore it’s good!). These economic numbers are also measurements, but we have to reconcile them with other data. Adoption and economic growth are so concentrated in so few hands that, while this props up our stock indices, it doesn’t lead to tangible benefits for everyone else.
Certainly, it doesn’t help that the industry has been selling itself as coming for everyone’s jobs. It doesn’t help that many prominent people working at these companies take to social media to gleefully exclaim that they think there’s more than a 10 percent chance that their company will kill everyone on earth. It doesn’t help that the big AI companies have been caught committing felonious cybercrimes all summer. It also doesn’t help that the leaders of these companies have been quite visibly embracing the Trump administration. Hugging a loser is a bad look, guys.
It’s a shame, because the technology is truly amazing. The AI true believers (accidentally) built something very cool that I would not have thought possible five years ago. But this amazing technology makes everyone miserable! It even immiserates its biggest boosters, who get sucked into productivity doom loops managing swarms of agents that speak in a unison chorus of Claudish. What’s our positive case here? The torment nexus of “superintelligence” ain’t it.
I have a lot of friends, colleagues, and alums who make a living building AI models. And I guess a bunch will be infinitely rich after the upcoming IPOs. But if you’re working for something more hated than the tobacco industry, maybe it’s way past time you stood up to your employer and demanded a better direction. And no, an open letter doesn’t count.
I’m not hopeful that this will happen. Ten years ago, I wrote a blog post shell-shocked by the election of Trump and disgusted with my industry. The cameo from Sam Altman explains a lot of my pessimism. I’ve been soul-searching for a decade. The leaders of the tech industry, instead of offering a positive vision and doing better, ended up sucking up to the new authoritarian government. They did it with pouting and feet dragging in 2016. They did it with their chests out in a celebratory embrace in 2024.
Our industry and field need a complete reboot. But that requires collective action. How do we fix it? What do we do next? I’d love to hear your thoughts on how the tech sector can clean up its image and make a positive, populist case for itself.
I don't think there's any quick fix. You have described how the tech Bros have dug their own PR graves with great flair and efficiency.
I think AI will be rehabilitated in the eyes of the American public when really wonderful things like cures for cancer come out of it. It will have to prove its usefulness.
I also think that the AI development path suffers from gigantism— the conviction that simply building the database of everything is the way forward. I think smaller, dedicated AI tools that consume less compute, energy, and other resources are the way forward. It is true that you can peel a potato with a 10-foot knife, but there are more efficient ways to do it.
from the people but not for the people