This piece originally ran in The American Prospect. Read the original here.
In April, the Pew Research Center released polling showing only 17 percent of the American public thinks that so-called “artificial intelligence” technologies, like OpenAI’s signature ChatGPT, will have a positive effect on the country over the next 20 years. People were sharply negative on AI’s impact on the economy, education, the environment, the news, criminal justice, and arts and entertainment.
This is not an isolated finding. A January Axios poll found 72 percent of the public was pessimistic about AI technologies, and polling from YouGov shows that the three most common sentiments toward the technology are cautious (54 percent), concerned (47 percent), and skeptical (44 percent). Those negative perceptions ticked up this spring, while positive attitudes (impressed, hopeful, excited) have all fallen. The best poll for the industry is an outlier from NBC, which still only found voters evenly split on whether they like AI.
Perhaps the most telling poll comes out of Quinnipiac University, which sorted its respondents by income. Households with an income above $200,000 per year—the wealthiest 15 percent of Americans, according to the most recent Census Bureau data—thought AI would do more good than harm in their day-to-day lives by a 3-to-1 margin. But Americans earning below $50,000 thought AI would do more harm than good for them by a margin of 2-to-1.
Quinnipiac’s data confirms what we think most observers can intuit. The inescapable hype around AI technologies over the last few years has largely been an elite phenomenon, with business owners and technology moguls excited to slash their workforces and profit off of robot labor. Those at the bottom sense they’re on the chopping block as usual. Everyone else in the middle seemingly doesn’t quite know what to think, but is tilting toward the negative.
One of the only things Americans dislike as much as AI these days is the Democratic Party. A July Quinnipiac poll found only 19 percent of the public approves of the Democrats’ performance in Congress, the party’s lowest rating since Quinnipiac started polling in 2009. Indeed, 52 percent of registered Democrats disapprove of their own legislative leadership. And why wouldn’t they? These are the people who lost to someone they called a fascist, whose movement they framed as the end of the American experiment in democracy, yet whose hungriest leaders are now following some of the fascist’s hateful cues while trying to squelch the most popular up-and-comers in their own ranks.
One road back for Democrats is to oppose the AI industry. If they simply look at the facts and follow the values they claim to espouse, they should find themselves in this position. AI as we know it today is a planet-burning, cancer-causing, energy-gobbling, alienation-driving, brain-rotting, fascism-enabling, waste-maximizing pox of a technology. There are types of artificial intelligence that are worthwhile, and we could in theory have a valuable industry advancing that technology. But as currently constituted, AI’s rapid proliferation is neither inevitable nor particularly useful, and it threatens to actively pull the country away from every path to achieving meaningful progress.
Moreover, there is reason to believe AI could collapse under the weight of its own contradictions. As writer Bryan McMahon recently argued in the Prospect, AI development through large language models has begun to slow down, precisely as the industry needs rapid leaps in technological achievement to justify the trillions committed to the build-out. People and businesses just aren’t subscribing to AI models in large enough numbers for Silicon Valley to recoup its historic investment. Once investors lose patience and want to see returns on all this sunk capital, Ed Zitron argues, the AI industry is going to crash, potentially taking the stock market with it.
We can’t fully predict the future of financial markets, especially in the absurd, self-aggrandizing world of Silicon Valley venture capital. But regardless, there’s simply no political or ethical upside for Democrats to enable AI mania. Everything the party claims to stand for—workers’ rights, environmental protection, social justice, liberal internationalism, quality education, prudent macroeconomic policy—counsels against it. Making AI skepticism an anchor for a new, better-organized party would not only get out ahead of identifying a new and unpopular public enemy, it could give Democrats a chance to rediscover the “political” side of the American political economy.
HERE’S THE KEY QUESTION: If our policymakers do nothing and let the AI industry proceed as it pleases, what will the country and the world look like in five years? Let’s start with construction costs. McKinsey estimates that building the data centers the AI industry claims it will require by 2030 will cost $5.2 trillion in worldwide capital expenditures just to handle processing loads, with an additional $1.5 trillion for rote IT infrastructure. That’s a total price tag of $6.7 trillion. That’s over 50 percent higher than the total GDP of Japan, the third-largest economy on Earth. Does anyone think that computer programs drawing buxom cartoon girls are more valuable to humankind than all of Japan?
That investment will do little to aid local governments, because 32 states have passed tax exemptions for the data center industry, driving billions in lost tax revenue. Anyone who remembers Amazon’s HQ2 fiasco should recognize these subsidies for what they are: a shameful race to the bottom for the economic benefit of the largest companies on the planet.
Right now, AI is responsible for 1 to 2 percent of all global energy consumption, about the same as the entire airline industry. (Again, does anyone think chatbot girlfriends and cheating on high school essays are as valuable to the species as air travel?) But the industry is projected to consume 21 percent of all human energy by 2030. Some AI evangelists argue that even this is woefully inefficient, and that we need to prepare for the industry to use 99 percent of energy production.
All of this increased demand for energy is helping to cause consumer electricity prices to soar. After a national election centered on the cost of living, utility bill inflation has become a central issue. Democrats seen as weak on inflation would do well to address it and tie it to AI.
Assuming the power needs of AI data centers continue on their current growth pattern, a leading data center in June 2030 will need the equivalent of the full-time energy output of nine nuclear reactors to meet its power demands. That’s for one data center. AI companies need hundreds of these centers to power their products at the scale being envisioned.
According to MIT, the training process alone for GPT-3 consumed 1,287 megawatt-hours of electricity, enough to power 120 single-family homes for a year. Companies release new models every couple of weeks. When they do, the previous model is discarded. Even more energy must then be generated to train the even more complex new model. The scale of this industry’s waste is sickening. Meta AI’s GPUs alone consume as much power annually as nearly 341,000 average American homes.
AI evangelists try to cast the energy needs as a good thing, saying it can’t help but fuel the green transition. But the challenge of green energy has never been insufficient demand for electricity. And given the immediacy of the industry’s hunger for power—and Republicans’ (all too literal) scorched-earth renewable-energy policy— new solar and wind plants aren’t the winners of the AI boom. Instead, in Salt Lake City, retired coal plants are coming back online just to fuel data centers. Elsewhere, Microsoft hopes to reopen the Three Mile Island nuclear reactor and channel 100 percent of its output, enough to power 800,000 homes, into data centers.
Due to the use of fossil fuels, heavy AI build-out is already causing higher incidences of asthma, cancer, and other respiratory diseases. The public-health costs from data centers increased by 20 percent just from 2022 to 2023 and are projected to reach over $20 billion a year by the end of the decade.
Those costs are not borne equally; AI data centers are increasingly being located in densely populated, low-income, majority-minority areas. One study showed that nine of the top ten counties with the greatest public-health costs from AI are low-income and majority-Black. The most acutely impacted communities can have household-level costs that are 200 times the amount of the least impacted communities. The sacrifices of frontline communities are not hypothetical—they’ve already begun.
As Earth becomes less and less hospitable each year, as entire species are being eradicated faster than we can count, as residents choke from noxious fumes in the atmosphere, and as we face a chaotic worldwide effort to rapidly transform our entire energy grid … do we really want to risk it all for the sake of chatbots whose main proven function is facilitating cheating at school?
EVER-PROLIFERATING DATA CENTERS are also imposing more hidden costs, ranging from undermining critical thinking and social skills to exacerbating drought and other climate disasters. More and more research, including from OpenAI, is finding that significant use of chatbots is linked to both diminished reasoning skills and social isolation.
In the last two months, we’ve gotten reporting that Meta AI internally sanctioned its chatbot to engage sensually with children (and also help users create arguments about innate racial superiority and spread medical misinformation), and on ChatGPT egging on an adolescent’s suicide. In both cases, software that is increasingly being deployed and marketed as a form of social connection was brazenly endangering people’s mental well-being. Both companies made a deliberate decision to turn a blind eye to that behavior until public outcry.
This is not the first or last time AI has reportedly played a role in inciting suicidal behavior. In October of 2024, the mother of a teenager in Florida blamed her son’s suicide on a program called Character.ai, which allegedly caused him to drift away from social connections and family. And just a couple of days after the story about ChatGPT being sued for wrongful death for pressuring a teen not to leave out his noose so that he wouldn’t be stopped, The Wall Street Journal reported that a mentally unstable man was incited to kill his mother and himself after ChatGPT reassured him that his neuroses around his mother poisoning him were real.
As journalist Molly White reported, OpenAI also hosted and promoted a “Looksmaxxing” bot that pushed toxic incel-inspired gender ideals and aggressively encouraged invasive cosmetic surgery for men. As White wrote, these chatbots “parrot extreme ideology around gender dynamics, sex, and dating; promote pseudoscientific beliefs; and potentially drive vulnerable or young users toward extremist communities.”
Similarly, while some AI therapy programs maybe are useful, they exist in a landscape of disastrous programs that all too often wind up magnifying feelings of loneliness and low self-worth.
The causal evidence for ChatGPT use undermining critical thinking is even stronger. A randomized control trial from MIT found that repeated use of the chatbot resulted in progressively worse writing, research, and idea generation.
THE AI INDUSTRY’S CURRENT TRAJECTORY is one of disastrous consequences. People aren’t happy about it. Democrats have an opportunity to do the popular and right thing, but it will cost them—perhaps permanently—their already failing political marriage with the titans of Silicon Valley. If they stand with AI, then the bipartisan commitment to allowing the common interest to be run over roughshod will be reaffirmed, and Democrats will confirm that they are more interested in chummy community with the elite than in listening to or governing for the public.
Particularly at a time when cost of living and the oligarchy are the issues turning people out, it’s hard to think of a better fight to pick than taking on the wealthiest executives in America over raising people’s utility bills.
There will be opportunities at all levels of politics for Democrats bold enough to take on the industry’s influence apparatus. Local and state candidates can work to oppose giveaways to data centers their communities do not want. Democratic candidates in states with large tech industries—Arizona, California, New York, Virginia—can capitalize on promising to directly police firms’ conduct. At the national level, the DNC, DSCC, DCCC, and congressional candidates can talk about the need for congressional hearings and legislation, as well as conducting oversight of the industry’s executive branch regulators.
With the launch of a new PAC spearheaded by figures like Marc Andreessen, it looks like the AI industry is getting ready to copy the cryptocurrency political playbook and spend huge money in the 2026 midterms to try and suppress any criticism of the industry’s conduct. The American public has realized they don’t like AI, and now the industry itself is recognizing that dynamic. The question is whether Democrats will notice and capitalize on it or whether they’d rather continue competing for the title of least popular group in the nation.
Image credit: “OpenAI logo with magnifying glass (52916339167)” by Jernej Furman from Slovenia is licensed under CC BY 2.0.