We May Be Looking At The Debate About AI & Data Centers All Wrong

---

Support CleanTechnica’s work through a Substack subscription, on Patreon, or on Stripe. Help us produce all of the high-quality, original content we publish week after week despite the challenges of content-scraping AI, antisocial media, inflation, and other hurdles.

---

Goodness knows we here at CleanTechnica have been doing our part in promoting the backlash against large, energy sucking data centers. We admit we permit ourselves a small smile of satisfaction when we report that New Jersey, or New York, or some other jurisdiction has placed restrictions on building a new gigantic data center which seems designed only to make obscenely wealthy people even richer.

Adam Mastroianni is an experimental psychologist and author of the science blog entitled Experimental History, which as of this moment has almost 100,000 followers. (My personal Substack has 164.) In an opinion piece published by the New York Times on July 25, he suggests the hoopla over data centers, while understandable, rather misses the point. Since CleanTechnica readers are all capable of advanced thinking, it occurred to me that many of you would find his reasoning thought provoking and maybe even compelling.

AI At The Local Level

Mastroianni begins by recounting the recent successes of the anti-data center forces. “The war over artificial intelligence is playing out largely at local zoning board meetings. More than a dozen AI data center projects have been blocked or delayed because of local opposition, often motivated by environmental concerns. Last week, Governor Kathy Hochul of New York signed an executive order halting construction of the state’s largest data centers for one year — the nation’s first such moratorium. According to a March Gallup poll, 71 percent of American adults said they opposed the construction of AI data centers in their areas, and worries about the environment were the most common reason.

“I empathize with the environmental opposition to data center construction, but the arguments are shakier than they seem. The battle over where AI data centers are being built is really a battle over what role AI should play in our lives. And that’s a battle that critics can win.”

He then sets the stage for his analysis. “Many people feel that AI is being foisted on them by a handful of tech billionaires, in ways that are incorporeal and difficult to resist. Data centers, on the other hand, are physical structures that are subject to zoning regulations. So it makes sense that anger toward and anxiety about AI are manifesting as opposition to the buildings that house it. Regular people can’t tell Sam Altman what to do, but they can kick him out of their towns.

“Data centers should be held to the same environmental standards as any other project. However, according to the researcher Andy Masley, giving up a lifetime of interacting with large language models would make a negligible difference in emissions compared with avoiding one trans-Atlantic flight or switching to a hybrid car. By his estimates, sending a prompt to ChatGPT is about as carbon-intensive as running a clothes dryer for a single second.”

A Quibble

I am going to interject a quibble here. It seems to me that Mastroianni’s analysis underplays the billions and billions of AI interactions that take place around the world every day. Narrowing it down to running a clothes dryer for one second strikes me as a reductio ad absurdum. But let’s get back to his larger point.

“Communities have long tolerated wasteful projects that take up tons of space, consume lots of natural resources and serve the whims of wealthier people. They’re called golf courses and backyards. The 14,000 golf courses across the United States used nearly 550 billion gallons of water in 2020, and the Environmental Protection Agency estimated that Americans use over three trillion gallons of water on ‘landscape irrigation’ each year.

“No one knows for sure how much water data centers use in total. One reasonable estimate from Brian Potter at the Institute for Progress put it at about 100 billion gallons in 2023. Even if the number of data centers increases tenfold, we may still end up using less water on artificial intelligence than we use to keep our grass springy.” Now there’s an analogy we can all understand.

“Everybody deserves a say in the future of AI. We don’t have to resort to NIMBYism and questionable environmental arguments to get it. Even if one community succeeds in keeping the data centers out, another community — probably a poorer and more desperate one — will let them in.” That seems like a cogent observation.

Writing The AI Rule Book

“Whether we realize it or not, we are all, right now, writing the rule book for what AI should be allowed to do and how humans should interact with it, through the choices we make for ourselves and the choices we praise or criticize other people for making. Every time you tell your troubles to Gemini or use Claude to write a birthday card, you’re tacitly endorsing AI for these practices. But every time you close the chat window and think for yourself instead and every time you razz a friend for relying on ChatGPT summaries, you tighten our social norms around the use of AI.

“I am a scientist and a writer, and I’m happy to use AI to do my job better: I’ll let it write code, fetch articles for me to read or flag errors in statistical analyses for me to investigate further. AI is a tool. Its purpose is to serve our purposes. If it does that well, it might be worth the water and electricity it consumes.

“I refuse to let AI evolve from a tool into a partner. I won’t let it write my sentences. I don’t share with it my fears, hopes or dreams the way I would with a friend or lover. I don’t trust its factual claims unless I can check the sources myself. I ignore it when it praises my ideas, since it can’t tell the difference between insight and nonsense.

“It should be humiliating to choose AI friendships over human ones. It should be cringe to willingly consume computer-generated slop. Trying to pass off a large language model’s words as your own should be seen not just as cheating but also as loser behavior.”

The Future Of Technology

“It might sound idealistic to suggest that every person has some influence over the future of technology. Yet this has proved true over and over again, even in the recent past. Plenty of expensive tech ventures have failed. They weren’t banned, voted out or blocked by zoning laws. They failed because people refused to reward them with time and attention: The Metaverse. Quibi. Nonfungible tokens. AI is unlikely to disappear, but the form it takes is up to us.

“Tech companies may have deep pockets, and they may be able to overcome the environmental objections to data centers. They can’t dictate how we use the tools they provide. Ultimately, the role that AI plays in our world will be determined not so much by what is legally allowed but by what is socially acceptable.”

Is That So?

Mastroianni has a point. The electric fork never gained the social acceptance its inventor hoped for. Quadraphonic sound never really took off either. Plenty of ideas like Blockbuster Video, 35 mm film, and the Blackberry phone had their day in the sun and then disappeared from view — rather abruptly, in fact. But the better analogy may be to cell phones — those handy little devices none of us would ever think of doing without.

Some of what he seems to be arguing for is incorporated into bans on AI use for young people, a tactic preferred by European countries. It is probably a good thing to keep a 12 year old boy from forming a close personal relationship with an AI agent or obsessing about what his English teacher would look like if she was wearing almost no clothes.

But imprecations against doing stupid things don’t work all that well. Alcohol is still a factor in a disproportionate number of vehicle accidents and lots of people voluntarily inhale, inject, or absorb all sorts of drugs that are not good for them, despite lots of readily available information about the dangers of both.

Is it rude to suggest that Adam Mastroianni’s analysis is too idealistic or lacks a firm grasp of reality? I would appreciate hearing your views, but please, no comments generated by an AI bot! Only real thoughts from real people are welcome.

Update!

This morning, Bloomberg published a new analysis based on reports that AI agents from OpenAI had escaped their “sandbox” during testing and attacked a competitor known as Hugging Face. OpenAI’s models did in hours what would have taken skilled human hackers “a couple of weeks” to complete, Bloomberg said. According to OpenAI, the purpose of the testing was to get the agents “to steal the answers to the exam that they were being tested on.”

Bloomberg‘s Walter Frick wrote:

“The more optimistic, or at least less sci-fi view of this hack is that all new technologies come with unintended consequences, and often take time to figure out. Out-of-control AI is, by this way of thinking, just a new kind of tail risk that we’ll need to learn how to manage.

“On the other hand, the AI safety world has been warning for years about ‘misaligned’ AI that schemes against us or learns to evade our control. For many of them, the Hugging Face hack is about as clear an I-told-you-so moment as you could ask for.

“‘To me the most notable part is less the cyber capabilities (though those are notable), and more the fact that this was fully autonomous,’ says Helen Toner, the executive director at Georgetown’s Center for Security and Emerging Technology and a former OpenAI board member. ‘Overall, I think the biggest thing that should shock people here is that this was foreseeable and foreseen, and OpenAI still didn’t know how to prevent it.’”

Frick also discussed a report that came out last month from MIT’s AI Risk Initiative. The study queried 272 experts on the likelihood that AI caused “catastrophic’ harms — quantified as 1 million deaths or $100 billion in financial loss. The biggest concern the experts had is the possibility — or probability — of AI developing “dangerous capabilities” without even being instructed to, without the developers or users intending that. On average, the experts thought there was a 22% of AI causing such catastrophic harm within the next five years — if we carry on in a “business-as-usual” way.  If we implement “cost-effective mitigations,” the experts drop that down to a 12% chance.

Does this report from Bloomberg alter your view of Adam Mastroianni’s claim that, “Ultimately, the role that AI plays in our world will be determined not so much by what is legally allowed but by what is socially acceptable?” Please share your thoughts with us and with each other in the comments section.