AI Must Be Destroyed
Turns out Skynet isn't a terminator that rips apart human flesh, it's an unconvincing lying machine that punches holes in the ozone layer
Yesterday, a piece of news crossed my feed that seemed to offer just a small bit of comfort: that the speed at which everything is getting worse might be, at least, deccelerating. "The rise of solar power and China's staggering EV growth may have pushed global emissions into decline"
Both independent research firm Bloomberg NEF and the international Climate Analytics institute say it is possible we will see global emissions fall in 2024 and 2023 will have been the peak of global emissions. We will know if these predictions have come true at the end of the year.
A simple reduction of total emissions is, of course, barely the beginning of what we need, and only the abolition of plastic production, fossil fuel consumption and arms manufacture will actually be sufficient in beginning to heal our global ecology. And yes, mass production of EVs and solar panels bring with them their own terrifying ecological costs. But considering how many catastrophic wildfires we've seen this year, if there is still a net reduction in CO2 put into the atmosphere it would mark a truly epochal shift in global material economy.
But at the same time, if those projections are built even a little bit on self-reporting from corporate agencies and pseudo-independent eco-consultants, then it's bullshit. Because another story circulating at the same time was that "AI Datacenters Are More Than 600 Percent Worse for Environment Than Tech Companies Claimed"
If these companies were one country, their actual emissions would rank them as the 33rd biggest emitter in the world.
This "600 percent" figure is probably an undercount of the actual impact, because it only tracked numbers from 2020-2022, before generative AI really hit the big time. But with clever eco accounting, the tech firms have been playing three-card monte with ecological outputs, using elaborate carbon-offset schemes to claim that their businesses are "carbon neutral."
Manipulation of the eco/carbon markets is to be expected--it's how we got the richest man on earth after all. Tesla has made a huge percentage of its revenue from carbon credits, making billions from other car manufacturers that invest in Tesla in order to "offset" their own ecological destruction. All that "green" wealth has been used to turn one of the world's biggest communication platforms into a nazi propaganda machine to soothe the ego of the world's most-divorced man.
Musk also uses that sweet sweet offset cash to fund Grok, a generative AI program through which free speech equals environmental destruction. Seeing that generating just 100 words of plaintext with GPT-4 consumes over a liter of water, imagine the devastation required to create pogrom-inducing anti-black memes of Donald Trump protecting kittens.
Folks who grew up in "red states" may remember the Bush-era lib-trolling practice of "rolling coal": truck owners mounted un-filtered exhaust pipes on their cabs, so the truck visibly spewed a black slurry of toxic chemicals and smog behind it as it drove. As AI-generated images are increasingly right-wing coded, AI imagery itself could be seen as a kind of alienated, disembodied equivalent: Uncanny Valley Trump God-Kings as alt-right coal rolling.
And it's not just about raw ecological damage: the mere presence of a data center in your town drives up electricity and water costs for everyone, so data centers create disastrous economic effects wherever they pop up. And that's to say nothing of other potential health and safety risks: a Bitcoin mine in a small Texas town produced a "Nightmare Health Crisis" as extreme sound pollution from the always humming servers gave people hypertension, heart palpitations, chest pain, vertigo, migraines, tinitus and panic attacks.
Bitcoin, NFTs and the associated blockchain-based ponzi-schemes were already infamous for their disastrous ecological impacts. Seeing as the 'mining' process is built on computers solving increasing numbers of increasingly difficult random math calculations, every crypto coin "minted" costs more energy to create than the previous. And as each transaction using crypto requires recording the entire data length of the blockchain, so too does every crypto purchase and every sale. Bitcoin alone (forget all other crypto-currencies) consumes more energy than many countries. In an essay I wrote in 2021, I argued this is a feature, not a bug: crypto is built on the myth that with enough electricity you could somehow turn code directly into value.
Of course, the entire blockchain economy relies on the relative affordability of electricity – and indeed, when electricity costs spike, many bitcoin mines either shut down, turn to electrical market manipulation or even shakedown local governments to not mine during peak electricity demand. There is a mutually reinforcing economy of scale between energy producers and bitcoin miners, propped up at almost every step by tax breaks and government subsidy.
Compared to AI, crypto is a tiny baby. The demand for data centers, power generation and AI chips is staggering. In the most eye-catching example, Microsoft has just funded the re-opening of infamous nuclear power plant Three Mile Island in Pennsylvania. The plant will operate under an exclusive contract: Microsoft will buy all the power it produces for twenty years.
And all this for generative AI, a product that people actively don't want: including AI in marketing materials reduces the desire to purchase the product. AI is a proven loser. The only people still enthusiastic are the folks who have already invested, largest among them Microsoft and Google, who have leveraged an astonishing amount of their business on the future of generative LLMs, with investments that make the Mark Zuckerberg Metaverse boondoggle look like a relatively reasonable vanity project. But the hype was so intense and the sales tactics so aggressive that there are a whole lot of people, corporations and middle managers who are "bought in".
Even still, it has appeared over the last weeks like the AI bubble was starting to pop.
OpenAI is currently raising a funding round at a valuation of at least $150 billion, and is expected to raise at least $6.5 billion — but potentially as much as $7 billion — led by Josh Kushner's Thrive Capital, with rumored participation from both NVIDIA and Apple...OpenAI will have to continue to raise more money than any startup has ever raised in history, in perpetuity, to survive.
Worse still, OpenAI also is trying to raise $5 billion in debt from banks “in the form of a revolving credit facility” according to Bloomberg, and the terms on revolving credit facilities tend to have higher rates of interest.
If Sam Altman's OpenAI, the most well-funded and highly lauded of the grifters, is starting to run out of rope, the whole enterprise is probably doomed. It's possible, however, that the historic 50 basis point Fed rate reduction may have bought the AI bubble a few more months, as cheaper money may convince some investors who were on the fence to dive in. Still, to everyone but those paid not to see it, it seems clear that the AI bubble is not long for this world.
And not everyone buying into the hype is just a mark. If crypto mining has a virtuous cycle of supply and demand with power production, AI puts that cycle into overdrive. While most of the world decarbonizes, the US under Biden has become the world's largest producer of fossil fuels, largely via a vast expansion of fracking in the gulf coast. This has been made viable by some major innovations in extraction technology, most importantly perfecting the technique of horizontal fracking, as well as a kind of "containerization" of oil rigs and pipelines, as more and more of the industry has successfully standardized sizes and fittings, allowing for faster construction and more easy storage, transport and ramps up and down of production.
For these oil producers, potential decarbonization in China and the EU is a doomsday scenario, and the whole "clean" "natural gas" line is less and less effective at pushing product. So while generative AI may not actually solve almost any of the problems that it is marketed to solve, it does solve one massive problem for the US economy: how to keep selling more and more gas.
[UPDATE 9/23 11:30PM: The Blog Blood in the Machine a few days ago ran a story full of the gory details about how explicitly the gas industry and the AI industry are coimbricated! Check it out]
In 2023, computer infrastructure and oil and gas extraction were two of the five fastest growing sectors of the US economy. Of the other three, two (telecoms and software development) are interrelated to these trends. It's a boom time for oil, for data centers, and for tech infrastructure.
The prominence of the hard-right Big Tech fascists, exemplified but hardly exhausted by Elon Musk, JD Vance and Peter Thiel, is reflective of the increasing alignment of tech with extractive and war industries. Big Tech has failed to make truly innovative and exciting products for almost a decade now, and it has "solved" this problem by creating ever more absurd grifts that derive their apparent legitimacy and value from exactly how expensive they are in energy and material effort.
Ideologically, the very fact of requiring so much material, economic and infrastructural investment means that they must be good technologies, because otherwise there's no way so much investment and so many smart people would be so involved; an emperor and his empty wardrobe. That's a scam, a confidence game on the scale of political economy, and thus it's a cozy and comfortable place for the far-right.
Combine that with traditional US powerhouses of retail and agriculture – for whom cheap energy prices facilitate global supply chains and petroleum intensive fertilizing and pest-control – and the absolutely horrific levels of ecological destruction produced by the genocide in Gaza, the West Bank, and, as of this morning, Lebanon, some huge percentage of it created by US produced and provided bombs – to say nothing of the ecological destruction wrought by otherwise defunct weapons systems deployed in Ukraine – and it becomes increasingly clear that the US economy is a pollution economy.
The American economy's current resilience is based in standing bulwark as the last major hold out against any kind of green transition. If things continue in this direction, it's not impossible to imagine the US fighting trade and/or literal wars in order to force the rest of the globe to continue destroying the environment. A dying empire trying to take the entire world down with it, rather than live in a world where it does not reign supreme.
Ok, so AI is bad, what of it? Don't we just wait for the AI bubble to burst?
That's not going to be enough. Even if AI falls apart tomorrow, even if it takes Microsoft or Google with it (which are not impossible scenarios), the industry will still try to find a use for every new data center that is already in process of being built. As long as the social, financial and infrastructural incentives are in place, some other shady, electricity-requiring product no one wants will take its place. Microsoft is locked in to that contract with Three Mile Island for 20 years. What will they do with the power?
Big Tech has become an industry that finds increasingly absurd ways to burn cash (in the form of VC funds) and electricity, and there's no reason to expect things will change just because one bubble bursts. They pivoted pretty cleanly from blockhain and "web 3.0" to AI, after all.
The longer the AI bubble continues, however, the more it results in direct investment in physical infrastructure, and the more disastrous it will be for communities and for the planet. As we saw above, data centers create potentially serious health hazards, as well as driving up utility costs and reducing water availability, all things that will hasten gentrification, housing crises and localized climate disaster.
It's also where we can intervene.
Because we can research and discover where new data centers are being built. We can protest and draw attention to absurd data center tax breaks that incentivize their construction in our communities. We can organize to block, picket, or otherwise disrupt their functioning.
Over the last few years there have been a (long-overdue) flurry of reevaluations of the Luddite movement, historicizing them correctly as an early organized labor movement that looked to attack the capitalistic use of particular kinds of machines, not just the ignorant "anti-tech" position they get smeared with by our tech overlords.
What would a modern luddite movement against data centers look like? Would there be methods of protest, direct action or sabotage that could stop these centers in their tracks? How do you brick a server, or an entire data center? Can you stop the shipment of AI-powered chips? Could hackers take them down remotely? What would real resistance to these developments look like on the ground?
Because if we can't stop these data centers being built in our backyards, then how do we stop them polluting the night sky, taking away the very stars we gaze upon?
All of this is less than worthless, it is a hateful abomination upon our world, the petty schemes of vicious racists who want to drown the world in blood, smog and images of six-fingered Jesus that they might feel just a little bit better about their pathetic empty lives. That the stars might be blotted out, the cities flooded and the rainforests burned all for the vanity of the world's most embarrassing, terrible men: this is the promise of capitalism in 2024, this is what America means today.
Ecological struggle has always meant struggle against capitalism, built as it is entirely on waste and destruction, but there is a strategic political opportunity in the widely reviled and obviously-scam-ridden field of AI. People already don't like it. That's a point of leverage, a point of possible unity and agreement. All we're missing is an angle of approach. The only way we can find that angle is by joining together to experiment, trying things out and figuring out what works.
Ecological scientific progress today does not look like a more energy efficient computer chip or a better solar panel, but the research, technologies and methods of revolt, sabotage, and liberation. Unlike big tech, our tech doesn't need angel investors and millions in financing to scale, doesn't need states to protect our IP through patent and copyright or universities to hold our labs. Our science just requires each other, the capacity for creativity and collaboration, and the desire to try and answer questions we've up to now been too timid to ask.