The Environment Can’t Afford AI

Time to put on the brakes

Graphic by the author, from public domain elements. Version with transparent background available at Wikimedia Commons.

“Artificial Intelligence” (AI) is a broad subject that can be approached from multiple directions: economic, political, cultural, creative, and environmental. Over the last couple years, I’ve read articles, listened to podcasts, watched videos, and spoken with friends about their experiences with it. Among the most helpful sources have been Robbie Martin at the Media Roots Radio podcast (which he co-hosts with his awesome sister, journalist Abby Martin), and

Brian Merchant

, who puts out Blood in the Machine here on Substack.

Wikipedia includes the following as examples of of “Artificial Intelligence”:

advanced web search engines (e.g., Google Search); recommendation systems (used by YouTube, Amazon, and Netflix); virtual assistants (e.g., Google Assistant, Siri, and Alexa); autonomous vehicles (e.g., Waymo); generative and creative tools (e.g., ChatGPT and AI art); and superhuman play and analysis in strategy games (e.g., chess and Go).

The “generative and creative tools” have been getting the most attention lately because they have the most potential so far to replace human jobs en masse, including artistic work. These fears were concisely summed up in this meme, which made the rounds and that you probably saw:

Robbie Martin has produced several podcast episodes about AI, taking an impressively deep dive into the corporate interests and personalities involved, and also, as an artist and musician himself, sharing the results of his own experiments with various AI tools. His reflections will be of special interest to other creatives.

As a writer myself, I’ve been really skeptical. In researching this essay, I experimented with two tools, Perplexity and DeepSeek, because they are supposed to be helpful for research. Perplexity primarily provides answers to queries by searching the web and summarizing the results, and you can limit the sources to academic work. DeepSeek defaults to accessing only the data it was trained on, which ends at July 2024, but can also search the internet. Perplexity provides a list of the sources it uses but DeepSeek does not.

For my first test, I entered the query “critique of invasion biology” into both because it’s a subject I know well at this point. (I am co-authoring a book on the subject with Nikki Hill, tentatively entitled, “Don’t Blame the Messenger: A critique of the ‘invasive plant’ narrative.”) Both tools gave answers that impressed me for their completeness, but neither one said anything that Nikki and I haven’t already been exploring in our six years of research (except for of introducing us to the term “reconciliation ecology”). DeepSeek’s answer was more complete and detailed overall but the lack of cited sources made it much less helpful from a research standpoint.

I ran them both through several other queries. Geeky stuff related to our book project like “Clements vs. Gleason” and “what is the theory of niche-based competition?” I didn’t really learn anything that I didn’t already know, and didn’t find the cited sources to be that helpful. In this case, for our book, AI had no contribution to make.

For the next couple weeks I used both Perplexity and DeepSeek on a regular basis as my go-to search engine because Google has gotten so enshittified. For simple requests for information like that they were both far superior. So in terms of learning about something by using the internet (rather than reading books, taking courses, finding a mentor, doing an internship, or going for walks and asking the trees, etc.), AI impressed me.

But—

(and you knew that was coming)

—but the environmental impacts of AI are incredibly significant and will only be getting worse.

For that reason alone we should seriously reign it in, if not ban it for most purposes.

Given the multiple crises we are facing—habitat destruction, extinctions, top soil loss, freshwater depletion, chemical and microplastics pollution, and climate disruption—our collective focus as a species should be on reducing our overall consumption. Every expanse of habitat spared, every square mile of topsoil conserved and every acre foot of water saved is something, as is any reduction in the amount of toxins and hazardous byproducts manufactured and spewed into the world. As for the climate, we must also reduce greenhouse gas emissions and undo any land uses that are responsible for the changes that are increasingly complicating life.

Continuing to develop AI will fly in the face of all of that, and exacerbate all the crises.

“The cloud” is not an accurate metaphor for the infrastructure that supports AI and the internet. There’s nothing puffy, whimsical or ephemeral about it. As

Justin Kollar

puts it so well at his Techno-Statecraft Substack:

We’re in the middle of a massive buildout of digital infrastructure. From data centers and semiconductor factories to power grids and industrial zones, governments and corporations are pouring billions into the systems that sustain the digital economy. But this isn’t just about technology—it’s about land, energy, and power in the broadest sense.

In 2023, data centers used approximately 176 terawatt hours (TWh) of electricity, up from 58 TWh in 2014, and projected to rise to a startling 580 TWh by 2028, which would be 12% of total electricity use. That’s a helluvalotta power generation. Where will it come from? “All of the above.”

Fossil fuels, for one. Utility companies across the nation that were planning to reduce fossil fuel usage are reversing themselves and or are now seeking to build new fossil fuel infrastructure, some specifically for AI. Greenhouse gas emissions from data centers tripled between 2018 and 2024 and are expected to “skyrocket” according to the MIT Technology Review. Microsoft and Google are among the corporations “quietly abandoning” emission reduction commitments they formerly made. As Wired put it: Generative AI and Climate Change Are on a Collision Course.

Nuclear too. Microsoft is reopening the notorious Three Mile Island plant in Pennsylvania. Google is planning to build “mini” nuclear reactors. Despite the “green” veneer that some people try to give nuclear energy because of its low carbon emissions during normal operation, environmentally it’s terrible. Uranium is highly dangerous at all stages of extraction, refining and use. Thorium, though touted as a safer alternative, is still dangerous to health and nature. How to store the waste produced from both is an unanswered question.

So-called “green” energy will be part of the mix, but if worldwide trends continue, it will simply be adding to our overall energy use, not replacing fossil fuels. Also, as I am always pointing out here, the environmental damages necessary for “green” energy are themselves atrocious, and affect areas that would otherwise left alone. From the mining of rare earths and other critical minerals, to the manufacturing of components, and the maintenance of facilities, all sorts of damage and pollution happens, including from the use of fossil fuels, which are absolutely necessary to the process.

According to Scientific American, “The AI Boom Could Use a Shocking Amount of Electricity.” In a “worse case scenario,” data center expansion could lead to a “10 fold increase” in energy use by that sector.

That’s just powering the data centers.

Water is needed for cooling them. According to research published in Nature, US data centers used 1.7 billion liters of water day, over half of it potable water. “A medium-sized data centre (15 megawatts (MW)) uses as much water as three average-sized hospitals, or more than two 18-hole golf courses.” Current water use by US data centers is estimated to be equal to that of 2,000,000 families. The increased demand for water is anticipated to lead to crises around the world, since other industries, like agriculture, also anticipate a greater need. Ironically, researchers in Australia used AI to predict where water wars will break out in the future.

Water is also required to manufacture semi-conductors.

Then there’s all the upgrades that will need to be done to the electrical grid, which is already aging, to support more data centers. Besides replacing obsolete portions of it, many new miles of transmission lines will be needed, as well as substations, transformers, and other equipment. Setting all this up will take materials like copper, steel, timber, etc., and fossil fuels to build it all.

All these resources, all the degraded habitat, all the the pollution—for what? AI umpires for baseball? Judging beauty contests? Bad marketing? Just to make us more stupid?

But besides the silly stuff, there is what Justin Kollar referred to as “power in the broadest sense” which I interpret as including the immense economic and social power that will be accumulated by corporations and oligarchs as AI becomes more ubiquitous. At the time of this writing, early in Trump’s second administration, it seems apparent that the tech overlords are attempting to concentrate such power in their own hands. It’s a new “merger of state and corporate power” (h/t Mussolini).

Indeed, observers like Brian Merchant explicitly draw a line between AI and the consolidation of power by Big Tech. Others speak of technofeudalism.

Honestly it’s a fucking nightmare.

We need to ask some serious questions about what’s happening and where it’s going. The sort of questions we didn’t ask when the iPhone was released, for example, and look how much that invention altered life, mostly for the worse.

As

Brett Scott

of Altered States of Monetary Consciousness (Tagline: “100% AI Free!”) writes: “Tech doesn’t make our lives easier. It makes them faster.” He goes on:

We don’t just live in any economy. We live in a mega-scale corporate capitalist economy, and in such a setting technology is never used to save time. It’s used to speed up production and consumption in order to expand the system. The basic rule is this: technology doesn't make our lives easier. It makes them faster and more crammed with stuff.

Resistance is deeply, profoundly necessary, if only to keep our own selves sane. Part of that is spending time away from screens, around non-human living creatures whether plant, animal or other. How do they feel about these changes? Do they want this stuff? Resistance is also about fostering in-person relationships with our fellow humans as much as possible. I could go on, but I will give the last word to

Livana

, who writes the Visions for a New Earth Substack, and who spoke what is in my heart when she said:

Tech won’t save us. Human relationships built on love and reciprocity will. Mutual aid and parallel networks of care will. Connection to and reverence for the wild will. Encounters with soul will.