AI is making everything worse

Here’s a breaking news headline that seems to be rocking Washington this summer: AI doesn’t seem that popular with the American public.

Now I’ve sat in a lot of rooms and conferences over the last few years filled with the most enthusiastic AI boosters from the government, private sector, tech giants, and frontier AI companies, and there’s no shortage of friends and people I respect who have gone to work on some of the hardest and most exquisite problems of the AI age. Professionally, I’m in Signal chats, email lists, and salon dinners where tomes like “Situational Awareness” and “AI2027” and analysis from experts like Fei-Fei Li are name-dropped with abandon. There are a lot of interesting and exciting questions to debate about AI in the long-term.

But in the short- and medium-term, there’s been a marked change in recent months about how AI is being discussed in those conversations and spaces. Many of those conversations sound like the final scene of Robert Caro’s book THE POWER BROKER, where the ministers to Robert Moses, who physically remade greater New York in his image, whine and wonder about the “ingratitude of the public toward great men.” Why weren’t New Yorkers grateful for all that Moses had done to improve their lives? Why, today, isn’t the public more warmly embracing the exciting revolution of AI?

To wit: “Republicans and AI CEOs are in full panic mode watching politicians and the public turn on [data centers,] the physical engines of AI growth. In private, they tell us they can't find a compelling message to shift opinion fast enough.”

Much of the AI-boosterism media is treating this as some strange pearl-clutching issue — “oh no, data centers have a bad reputation! It’s threatening our investment in the future!”

Politicians and AI consultants seem focused on the “one neat trick” idea, imagining they’re just a couple smart talking points and slick slide decks for local city council meetings away from turning around the industry’s reputation. People misunderstand how wonderful data centers are!

No, they don’t. They recognize exactly what’s happening here.

I suggest that the problem is both more obvious and more fundamental: The daily lived experience of the AI boom sucks.

Here’s the basic problem: AI has arrived in American life and communities across the country as a giant turd mountain. No one wants it; it makes everything worse when it arrives; and it’s been pushed and backed and cheerled by the absolute worst people in modern American life.

That’s not a landscape where a couple improved talking points turn around an issue. As that Axios article explained, “Data centers are as popular as spent nuclear waste, and more broadly have become a proxy for feelings about AI.”

All of this feels like it should be less surprising to the AI industry than it appears to be.

Sure, someday it might make a whole lot of things better, but the ordinary Americans’ experience with AI is as if after the invention of electricity, every factory worker was assigned to start each workday by receiving an electric shock, and newly founded electrical companies went around to rural farms tasering farm children involuntarily and then dropping a dead electrocuted hog down the family water well.

It’s not hard to imagine that Thomas Edison might have struggled a bit popularity-wise amid such a roll-out.

AI is being forced on Americans in their personal lives and professional careers, and — for now — it seems to be making everything worse. Let’s review the top-line experiences of how an ordinary American is interacting with and reading about the impact of AI:

It’s pillaging your knowledge and life’s work.

We’ve lived through years of headlines about how the backbones of these frontier AI companies — what are known as “large language models” or LLMs — have been built by thieving everything ever written, said, or created by anyone, from blog posts and books to photographs, movies, and music. It’s not entirely clear that the AI training systems were outright illegal, but it seems clear to virtually any sentient being that it was morally wrong and underhanded. AI feels like a parasite feeding off humanity, a comparison made explicit by one wildlife photographer whose work was scraped to generate realistic-looking photos that mimic what he’s spent his life creating: “This harvesting of our work, scraping and impersonating our styles, it’s the same as a parasite that sucks the light out of its host without invitation.” Now, having turned the internet into a giant fetid wasteland of slop, scam websites, and wooden-sounding LinkedIn posts, the industry is turning to hoovering up and destroying old books that contain words not yet poisoned by AI.

It’s going to steal your job.

It’s not hard to see why students at commencement exercises across the country booed mentions of AI this spring. Optimism is fading at a generational level. We’ve seen years of dire warnings about millions of people being put out of work by AI. “It seems very likely to a large number of people that we will get massive unemployment caused by AI,” said the “Godfather of AI.” Hundreds of tech executives are warning of “large-scale job displacement.” And one study shows 99 percent of corporate executives think AI will lead them to layoffs in the next two years! (Given that I’m sure that 99 percent of corporate executives wouldn’t agree that Joe Biden won the 2020 election or that climate change is real, we should take “99 percent” unanimity seriously.) Now, there’s a lot of emerging evidence that the “AI job apocalypse” may not arrive, at least as quickly and broadly as once featured, but all manner of industries are facing existential questions about once-stable careers, from truck driving to accounting.

It gives you the wrong answer when you seek help.

Despite the vast scale of uncompensated intellectual property theft, AI is often bad at answering questions. We’ve all experienced the weird “hallucinations” when you ask Claude, ChatGPT, or even Google for an answer and it … just makes something up. That’s hardly an edge case — one BBC-led study found that the “AI assistants misrepresent news content 45% of the time – regardless of language or territory.” Trusting AI turns out to be terrible. And there’s growing evidence that it’s actually not helping students, even before you get into the rampant cheating scandals. There’s a stunning new Chinese study that shows a wide (negative) divergence between students who use AI and those who don’t. As international relations professor Nicholas Grossman wrote this week, “Evidence keeps accumulating that LLMs are a powerfully useful technology for people who already know what they're talking about, and a dangerous outsourcing of cognitive function for people who don't.”

It’s making your life harder, not easier.

Sure, I’ll stipulate that in certain specific fields — coding! — AI is making some huge leaps, but all the talk of “10X” or “100X” productivity boosts are turning out to be vaporware. Perhaps they’re ten percent, not 10X. Perhaps an hour a week for some workers. Companies that just a year ago were screaming at employees to adopt AI are now scaling back as gains fail to materialize and costs skyrocket; the ROI isn’t there. Even what little gains and tools might be useful are being offset by the slop our coworkers and bosses are creating. As one summer study showed, “For every hour an employee spends getting a useful output from artificial intelligence, they spend another hour making it usable.” Maybe AI is helping some coworkers somewhere, but more of us can relate to the headline about the State Department presenting at a global conference using a map that mislabeled every country in Africa or the Walmart workers who now have to babysit their new AI systems and correct its errors, or the teachers and professors having to redo their curriculums because so many students are cheating.

And all of that is being overlaid with how many companies are investing in AI “customer service bots” that are even more infuriating to deal with than the phone trees of yore. Time and again, we’re being shuttled into seemingly inescapable Kafka-esque loops of screaming “operator” or “let me speak to a human!” when trying to interact with all sorts of companies where we’re actually the paying customer. Oh, you thought customer service was bad a couple years ago? Surprise! It’s worse now, thanks to AI: “Customer service hits a new low.”

It’s making bad people’s lives easier, not harder.

Too often, the most prominent use-cases and adoption of AI seems to be by people intent on invading privacy, enabling surveillance, and generally making public life worse. It is enabling evil by bad people. The two dominant use cases I see almost daily are storylines about Meta’s AI-powered “pervert glasses” making harassment easier and dozens of headlines about Flock’s AI-powered cameras being misused by cops to stalk women. (Here’s a particularly egregious case, where a cop met a woman on a movie set, entered her car into a stolen-automobile list, and then used Flock’s AI-powered cameras to find her and pull her over, almost causing a wreck in the progress.)

It’s hard in this moment to win the competition for “most hated tech company,” but surely near the top of that list are both Flock and Palintir, companies whose public messaging seems to be “imagine how wonderful the world would be if we used AI to stalk every human simultaneously at all times”? There’s a nationwide backlash against Flock cameras, with them being vandalized or cut down, and my Instagram feed has been filled with all manner of “hypothetical” videos about what it takes to blind or cut down a camera. You want to talk about moral panics? Here’s a fun fact for you: More cops have been arrested for stalking women using Flock cameras than there are trans athletes in sports.

It’s making your electricity more expensive.

At a time when Americans are worried particularly about pocketbook issues, they’re watching their electrical costs rise dramatically in many parts of the country — entirely because of AI usage. A few sample headlines: “AI data centers trigger massive 'irreversible' 76% electricity price spike in largest US region” (Yahoo); “AI Data Centers Are Sending Power Bills Soaring” (Bloomberg); “Electricity prices rising by double the rate of inflation. Data center demand means no relief ahead, analysts say” (CNBC). As a federal mandated independent power watchdog wrote, “The price impacts on customers have been very large and are not reversible.” The electrical bill is something that American households buy every month and, with a few years in a house or community, you get a finely tuned sense of what’s “normal.” Now, those bills are soaring — and there’s a clear culprit each month when families open their electrical bill. Something they didn’t ask for, something they don’t have any control over, is stealing money from their pocket.

It’s using up your water and cooking our environment.

At a time when Americans are beginning to experience what living in a rapidly shifting and changing and hotter climate feels like, data centers are making a vast landgrab to steal vast quantities of water. Sample headlines: “AI is draining water from areas that need it most” (Bloomberg); “A data center drained 30M gallons of water unnoticed — until residents complained about low water pressure” (POLITICO). Consumer Reports has a whole website tracking how AI is vacuuming up America’s water and raising electrical prices. It’s giving real “I drink your milkshake — I drink it UP!” vibes. The “productivity gains” from AI are actually costing more in carbon emissions than it’s saving.

It’s killing children and flooding the world with nonconsensual porn.

How’s this for a PR problem: Wikipedia has an entire page dedicated to “Deaths linked to chatbots.” It’s a recounting of a whole series of horrors, including children who were led to suicide by chatbots, and there’s plenty more harm to calculate, as, again, terrible people are using AI tools to victimize children with AI-generated porn. Grok, Elon Musk’s Nazi-friendly AI system, has overrun X with seemingly illegal images, a problem that has overwhelmed regulators and legislators.

Oh, and it threatens to end the world.

As I wrote earlier this summer, the AI leaders themselves are bandying about probabilities like there’s a “25% chance” of AI going “really really badly” for humans. And, as I also wrote earlier this summer, there’s increasing evidence that despite such risks, the AI companies aren’t doing a particularly good job policing their own systems. What would we think if GM CEO Mary Barra was going around the pundit circuit saying she thought that there was a 25 percent chance that the new Chevrolet Silverado would wipe out all of humanity? Or Boeing CEO Kelly Ortberg was showing up on CNBC and saying he was confident there was only a 25 percent chance that everyone who flew aboard a new 787 would die?

Beyond all of this, there’s more!

People can sense how the AI race is driving and exacerbating the K-shaped economy, where the wealthy get wealthier and the already struggling struggle more, and helping to power the already giant gaps on income inequality. As a Washington Post article this morning says, “Evidence is piling up that the AI boom may instead be widening the gap between America’s economic aristocracy and everyone else. Data shows that the people making the most use of AI are concentrated in richer urban areas, while disadvantaged regions have lagged behind. Affluent Americans are on a spending spree thanks to investment portfolios fattened by an AI-addicted stock market. And some economic experts are warning that AI could widen the income gap between American workers and business owners that has caused anxiety for a decade.”

The fact that, on the Post website, those two sentences are interrupted by a prompt, “Get concise answers to your questions. Try Ask The Post AI.” helps underscore how AI feels at every turn that it’s being foisted unwantedly on us as unwitting experiment subjects in daily life:

Have you ever tried to use “Post AI”? It’s horrible!

And then, after all that, we get to the AI boosters themselves. In an age where Americans are beginning to feel political outrage against a corrupt oligarchy that seems to be literally stealing from us, where everything in American life seems to be an insider-trading grift, the AI industry is a perfect and obvious villain. It doesn’t hurt that most of the players in this space are the literal people, like Mark Zuckerberg and Elon Musk, who seemed to have wrecked American life over the preceding decade and now are promising that this next moment will be even more disruptive. Zuckerberg just published a particularly dystopian manifesto about AI, and Musk is investing millions to help Republicans win the midterms and turning AI-boosterism into a partisan issue.

AI’s image problem isn’t the data centers. Or at least it isn’t ONLY the data centers. The image problem for AI is AI.

I can see — and believe there will be! — all manner of scientific and medical advances that can come through AI. With time and patient experimentation, we’ll be able to use it for drug development, medical research, climate research, genetics, and more, but today, right now, I struggle to point to a meaningful experience ordinary Americans are having where AI isn’t contributing to the worsening of daily life.

If you’re looking for a political solution, there’s a relatively straightforward one: Try to show one area, any area, where the ordinary human experience is being made better by the roll-out of AI.

GMG