3:38
In December of 2023, Google executives were hawking their wares at the annual global climate change summit in Dubai. There, at COP28, Google claimed that artificial intelligence could supercharge climate action ["I think AI has a really major role to play in addressing climate change"]. From Google's standpoint, it seemed that the only thing holding the world back from eliminating emissions was just finding the right efficient solutions. And it wasn't just the likes of tech giants like Google that were selling a future of AI salvation,
the whole conference seemed to be drinking in the promise of AI as the silver bullet to saving the world ["Because, ladies and gentlemen, AI is the future."] But is it? Will artificial intelligence pull us out of climate chaos? Today, we dig deep into the cogs of artificial intelligence. We'll unearth what artificial intelligence means through concepts like large language models while trying to decipher whether AI is a savior or a destroyer of the planet and our jobs. We'll explore all the way down to bowels of capitalism, to understand the driving force behind the AI boom. And there we might find a sliver of hope,
a potential path to build a better, more just, and ecologically sound world for all. What is AI? ChatGPT exploded the internet. A chatbot that could seemingly converse, understand, and know what you're saying. The future had arrived. Machines were now sentient, or so it appeared. And soon after, AI tools that could create images or even videos out of thin air, blossomed. Twitter, TikTok, and YouTube were soon awash with AI-generated images, scripts, and videos. But, what exactly is this new branch of artificial intelligence? To grapple with that question,
we have to take a step back and understand that artificial intelligence has quickly become a catch-all for numerous programs and algorithms that often employ high-level pattern recognition and prediction to appear as if they can learn and are sentient. The idea and the application of Artificial Intelligence have been around for decades, but the reason why it has exploded into the public eye in the last three years is because of significant advancements in Large Language Models. Drawing on innovations in machine learning and deep learning, both of
which leverage high-level pattern recognition and neural networks, researchers and companies like OpenAI feed programs vast sets of data on which the program trains to understand patterns and anomalies. Based on that training, these programs can then predict and generate whole sentences, or even whole papers. Or as cybersecurity expert Jeff Crume explains ["think about this is a little bit like the autocomplete when you start typing something in, and then it predicts what your next word will be. Except in this case, with large language models, they're not predicting the
next word. They're predicting the next sentence, the next paragraph, the next entire document."] Essentially, Generative AI programs wield machine and deep learning to ingest vast amounts of words, video, images, or audio from large data sets often scraped from the internet as a means to predict and generate what comes next. In their simplest form, this is how the AI we're seeing right now, like ChatGPT, MidJourney, or even AI-powered search engines are functioning. This is not sentience, but instead a very powerful pattern recognition and prediction program. While this technology has some amazing applications, unfortunately, it does not exist
in a vacuum. Training and using AI has very real material consequences. To increase the accuracy and fidelity of these models, for example, computing power demand rises exponentially. As this paper summarizes: "For a linear gain in performance, an exponentially larger model is required, which can come in the form of increasing the amount of training data or the number of experiments, thus escalating computational costs, and therefore carbon emissions." In short, more intelligent AI means more energy demand and emissions, but just how demanding will AI be?
What does AI mean for the planet? While the chief sustainability officer at Google described how AI would help solve climate change at COP28, data centers across the world were guzzling fossil fuels at a breathtaking pace. Artificial intelligence isn't magic, it has material costs. Its neural networks, its data sets, its learning process all rely on metals like copper mined in hyper-exploitative conditions, on data centers guzzling fossil fuels, and on the millions of exploited "ghost workers" doing the grunt work of the digital age. And this demand for energy, materials, and labor has ballooned thanks to the explosion of
generative AI technology. While it's hard to pin down exact numbers because big tech firms work hard to obscure those numbers, there are open-source AI tools out there that are similar enough to make some guesses on just how demanding generative AI is, especially when it comes to energy. This paper did exactly that. They extrapolated from the energy requirements of the open-source AI tool BLOOM, and found that GPT-3, which is what ChatGPT is based on, had an energy demand of 1,287 MWh to train, causing roughly 500 tonnes of carbon dioxide emissions. Admittedly,
that number is small, the equivalent of the energy requirements of 100 American households. But this technology is in its infancy, and that is just for training the one Large Language model. With every tech company salivating at the chance to integrate AI into their products, that number will only compound and grow. Like in the case of Google searches. Google claims that the average energy consumption of their search is.3 Wh, but when AI processing is used for search instead that number could jump as high as 6.9-8.9 Wh. When considering that Google processes 9 billion searches a day,
that spike in energy demand would be devastating. Indeed, a 2024 energy outlook paper from the International Energy Agency claims that "By 2026, the AI industry is expected to have grown exponentially to consume at least ten times its demand in 2023." Even the CEO of OpenAI admitted that artificial intelligence is an energy guzzler: ["…AI is going to need a lot of energy"] Energy consumption is just the start, however. Freshwater demands will also rise as corporations build more and more data centers to keep up with the AI boom. And data centers,
the physical manifestation of AI, need water because they get extremely hot- similar to how your computer heats up when it's running a particularly intensive program. To offset this, data centers use a combination of liquid cooling and air conditioners to maintain an optimal temperature for the servers. This requires a lot of energy, but also, importantly, a lot of water. Water to the tune of 700,000 liters for training just one AI model. Taken together, global AI use is projected to "account for 4.2 - 6.6 billion cubic meters of water withdrawal in 2027." That's
the equivalent of half of the United Kingdom's yearly water usage. This could spell disaster, particularly for the environments and people surrounding data centers. Like when Microsoft, which now owns OpenAI, decided to place a data center in the Arizona desert. According to a report from the Atlantic, Microsoft's AI data center expansion in Goodyear, Arizona could use "56 million gallons of drinking water" every single year." Which is the equivalent water usage of 670 families in the surrounding town. If more data centers continue to get built as expected to power ChatGPT and other AI tools, the question
of freshwater could quickly become one between keeping AI churning or keeping families alive. But if we believe the words of Google's chief sustainability officer, these impacts don't matter because AI is going to help "solve" climate change. This is a patently foolish idea. Climate change is already solved. We know what we need to do. We need to stop all fossil fuel production and extraction, while rapidly installing wind and solar generation. The problem is not a lack of technology. As Stanford engineer Mark Jacobson lays out in paper after paper, we have
all the technologies we need to drastically reduce emissions. We can achieve a 100% renewable world and divorce ourselves from almost all fossil fuel use today. The problem that AI needs to solve, then, is economic and political. We have all we need for an energy transition, but fossil fueled capitalism's many tendrils are blocking that transition because it is an existential threat for some of the biggest corporations in the world. AI won't help to dismantle capitalism or stop climate change, indeed, if it's any indication, it's currently being wielded to deepen the oppression, exploitation, and destruction caused by capitalism.
And while we wait around for some future promise of AI saving us from the destruction of fossil capitalism, the energy demand for generative AI will continue to skyrocket. Making the energy transition even more difficult. Not to mention, AI is also being rapidly deployed throughout the fossil fuel industry to streamline exploration, refineries, and even pipeline logistics. As Priya Donti, an expert in the field argues: ["the applications of AI that are uh counteracting climate action so AI is being used to accelerate oil and gas exploration and extraction it's also a big driver of things like
targeted advertising that change how we consume or of how we consume information about climate online."] The is no doubt that some of the AI hype around solving climate change will come true. But while perhaps helpful down the road, championing AI as the solution to climate change distracts and actively obstructs us from tackling the mammoth tasks we need to be addressing right now. Perhaps instead of Microsoft "spending more than $10 billion on cloud-computing capacity in every quarter of late" in what a semiconductor analyst called the "largest infrastructure buildout that
humanity has ever seen," we could, I don't know, direct those resources to building out infrastructure like a 100% renewable grid and dismantling the fossil fuel industry. At the end of the day, machine and deep learning, and AI in general are tools. The problem is that these tools are being wielded not for the well-being of the many, but instead for the enrichment of the few-an enrichment that is reliant on the destruction of the planet and the oppression of the masses. AI under capitalism For the digital artist, image-generating AI could spell extinction. Ever since the precipitous rise of image-generating models like DALL-E and MidJourney, making a living off of one's artwork has never seemed so precarious. Why would a
company pay an artist an hourly wage when they could quickly generate a similar image with AI and just pocket those wages as profit? Indeed, I'm even guilty of this. I've used image-generation tools for my thumbnails in the past. So, AI is unequivocally bad, right? Well, there's a key element missing from this analysis. AI is bad under capitalism. The acceleration of artificial intelligence accelerates the precarity of work under capitalism. This is crucial to understand. Capitalism functions based on surplus value, which often presents itself as profit. Capitalists-the
owners of companies- create that surplus value by paying workers as little as they can get away with while simultaneously trying to work them as long and hard as possible. The foundational trick of capitalism is minimizing the cost of labor while maximizing the amount of commodities produced. In short: exploitation. And technologies, whether it's the fossil-fueled steam engine or in this case, energy-guzzling AI are wielded to increase the efficiency and speed of workers, while also making their jobs more uncertain. An increase in worker productivity means that
laborers create more products for lower costs, and having to compete against AI-powered machines means workers are willing to take pay cuts just to keep their jobs. We're already seeing huge multinationals like Amazon wield artificial intelligence for exactly this purpose. The company employs sensors, data tracking, and AI to work warehouse employees harder and faster. Simultaneously, the long march of AI-empowered robotics fueled by massive sums of energy and emissions is pushing Amazon employees out of jobs. Amazon is not
alone, however. Corporations will increasingly use AI in an attempt to outcompete their adversaries, assert monopoly control over the market, and squeeze workers even harder. As Marxist Mark Morley writes: "The best AI for generating images, text and for solving problems, is and will continue to be developed by enormous monopolies like Google and Microsoft, with the best engineers, best hardware, and biggest databases. They will use their monopolistic position to make monopoly profits of course, and the technology's advantages, namely in speeding up and cheapening production, will be used by other corporations to lay off
some workers, and to drive down the wages of others. This technology is also already being used to speed up labour, and thereby increase the rate of exploitation." Ultimately, under capitalism laborers are chained to the forces of the market- selling themselves and their creative works to get money so they can place food on the dinner table and secure a roof over their heads. As such, capitalism forces the worker-artist, and indeed all workers, to ensure that their labor can be sold and is valuable in order to make a living. Put another way by Morley,
"Under capitalism…the artist's existence is precarious and subordinated to the vagaries of the market. They must jealously protect their exclusive right to the sale of their artwork, otherwise their livelihood risks being destroyed." As a result of the relentless force of capitalism's drive for profit, AI, which could have such beautiful applications, is instead turned into a tool that destroys the ability of the most creative of us to support ourselves. But AI doesn't just stop there in a capitalist economy. It corrupts our very ability to stay
healthy. It's now being used to streamline health insurance. In the relentless pursuit of maximum profits, the medical insurance company UnitedHealthcare employed an AI algorithm called "nH predict" to process insurance claims. The model is trained on the data of millions of other patients, and uses that foundation to predict medical outcomes and discharge times. And based on those projections, a patient can be denied care. However, the data and how the AI model was trained were faulty. One lawsuit against UnitedHealthcare claimed that the AI program had a
90% error rate. As a result, a slew of denials rained down on those who needed medical help, especially the elderly. And those denials meant the immiseration of thousands just because a faulty algorithm and company deadset on profit accumulation deemed them not worth the money. In a sense, all of this exploitation and destruction makes sense. Why should we expect this new technology to somehow save us, when the development of most technologies under capitalism, from the steam engine, to the cotton gin, to the conveyor belt, have been used to further
the exploitation of workers and the planet as a means to maximize the profits of capitalists? Again Morley explains: "Capitalism lays its hands on a revolutionary technology whose real potential is to harmonize and rationalize production and to enhance the creative powers of humanity, and instead uses it to further discipline the worker, to throw more workers on the scrapheap, to make the artist's existence even more precarious, and to concentrate more and more power in the hands of gigantic corporations." So, then, AI is not necessarily the problem. The
problem is that it's reflecting the extractive and exploitative nature of capitalism. So, could we wield machines and deep learning as tools to help facilitate a world on the rubble of capitalism? Can Artificial Intelligence be used for the better? To escape the tyranny of profit, and the endless exploitation of workers and the planet that comes with it, we need to build an ecosocialist economy. One that focuses on producing goods for use and needs rather than exchange and profit- a democratically planned economy that is guided by feeding, housing, and caring for people and the planet instead of padding pocketbooks. For
those steeped in the anti-communism of imperial core countries like the United States, economic planning is a boogie man- a utopian pipedream of leftists that would immediately fail in the real world. Supposedly, the market is the only way to facilitate the vast flow of goods across the globe to those who need it. There are just too many inputs, changing too quickly to properly plan and react to the needs of everyone-let alone the planet. But here's where AI, and specifically machine and deep learning systems could actually come into play. These technologies could help
facilitate and strengthen a democratically planned economy. If there's one thing these models are really good at, it's assessing vast sums of data quickly and understanding patterns, which is perfectly suited for understanding production workflows and supply chains globally. Indeed, if Amazon and Walmart are already using automation and AI to streamline and plan their vast companies, why is it so inconceivable to do so in a democratically planned ecosocialist economy? Again, Morley argues: "There is no reason sensors could not be integrated into the economy
as a whole to provide real time data about what is being consumed, and in what proportions, where, and what equipment is in danger of breaking down and therefore needs to be fixed in good time. The German software giant SAP has already developed an AI-powered application called HANA, which is used by companies such as Walmart to plan all their operations harmoniously using real time data." Deep learning algorithms could act as a crucial tool in a modern, socialist planning toolbelt. Helping to make sure people all over the world get what they need when they need it while simultaneously drawing down carbon emissions to zero. But
crucially this planning AI must be run by the people for the people and planet. It will simply be a tool to make sure everyone gets healthy, delicious food, excellent healthcare, comfortable living spaces, and relaxing leisure time to enjoy being with friends and family. A tool to build a world for people, not technology and profit. But right now, artificial intelligence is accelerating fast. It seems that each month brings new developments and scarily sentient models. To make matters worse, the highly antagonistic political climate has made it hard to decipher the extent of AI's impact, like in the case of Elon Musk's recent acquisition of X by his AI company xAI.
This could mean the potentially disastrous proliferation of AI on a platform already riddled with misinformation and hate speech, but honestly, it can be hard to sort past the media's tendency to overhype or catastrophize developments in AI, which is why I've been using this video's sponsor Ground News to peer past the curtain of media bias and stay up to date with the rapid progress of Artificial Intelligence. For this story, of the more than 50 sources compiled, over 30% leaned left while less than 25% leaned right. Comparing the headlines, the majority in the center are relatively similar. (show examples of headlines).
However, with Ground News's Bias Comparison, we can see how the contents of these articles differ (show Bias Comparison feature, located under primary story headline). For example, left-leaning outlets framed the acquisition with heightened scrutiny of Musk and painted X as a "pro-Trump" machine, while right-leaning outlets portrayed the deal optimistically, highlighting the "blockbuster" nature of the transaction and the potential for innovation by infusing X with more AI. I've been using Ground News for a long time because of tools like bias comparison that helps make sense of Elon Musk's role in the U.S. government. Ground News is a website and app that collects over 50,000 media sources into one place and lets you compare how
headlines are being covered across the political spectrum. Every story comes with a quick visual breakdown of the political bias, factuality, and ownership of the sources reporting - all backed by ratings from three independent news monitoring organizations. The layout of Ground News allows you to quickly compare headlines and articles to see what information is emphasized or left out. I especially like the Blindspot feed, which highlights stories that are disproportionately covered by one side of the political spectrum. Especially when it comes to news about climate
change, I think it's really important to see what articles on both sides of the political spectrum are missing. Which is why I use Ground News literally every day. You can go to ground.news/occ to get 40% off their Vantage Subscription, which includes a feature called my news bias. It's basically a dashboard that visualizes your news diet over time. It shows your top news sources, whether you engage with different perspectives, what topics you're interested in and a lot more. Go to ground.news/occ or click the link in the description to support an independent news platform working to make the media landscape more transparent.
