The AI Backlash: Why the Public Turned on AI's Winners

Transcript

Sam: About three in four voters in three swing states think AI will make billionaires richer. About one in five thinks it will do the same for them.

Alex: And most of those voters aren't even scared for their own jobs. That's the AI backlash: why the public turned on AI's winners.

Sam: Before we start, I want to tell you about something I found on Hacker News this week, because it's been rattling around my head. It's called Sphinx. It's a git hook, so a little gate on a developer's push, and when an AI agent has written the code, Sphinx stops the push and asks you a question about what that code actually does.

Alex: Hang on, it quizzes the human? Before the code ships?

Sam: It quizzes the human. It works with Codex, Claude Code, Cursor, plain git. Your answers never go back to the agent, so you can't just ask the robot for the answer. And there's a mode called yolo that lets pushes through, but every skip gets logged.

Alex: A skip log. That's the most honest feature I've ever heard of. Somebody knew exactly how this goes.

Sam: What got me is that I run a big team, and my junior copywriters are shipping AI drafts all day. I'd pay real money for something that makes them say out loud why the draft is right before it goes to a customer.

Alex: And that's the engineering version of my week. My reviewers are drowning. The code is fine until it isn't, and nobody on the team can say why it works. I'm going to read that repo tonight.

Sam: You will not. You'll read it late, get excited, and rewrite it in Rust.

Alex: Okay, that's a fair shot, and I'm not going to dignify it. Welcome back to Dan's AI Intel, the show that helps a curious person actually understand the AI revolution, not just keep pace with it. Each episode takes the one question that matters, digs past the hype and the fear, and asks what it means for the economy, for politics and for the race between the labs. I'm Alex.

Sam: And I'm Sam. And this week the question is a strange one. People are using AI more every month, and liking it less every month. So who, exactly, is the public angry at?

Alex: We'll start with a name for the movement, and a very bad week in April. Then we go to the building that turned an abstract technology into a neighbour. We'll cross the Pacific, because the mood flips there. We'll look at which companies are absorbing the anger, and whether the obvious suspect, prices, is even the right one. Then a trip to nineteenth-century England, a detour into money and happiness, and we finish on whether there's a deal on the table.

Sam: And if you're enjoying the show, hit follow in whatever app you're in. It's one tap, it's free, and the next one just shows up.

Alex: Before we get into it, though, I want to plant a flag. My honest starting position is that this anger is aimed at the wrong thing.

Sam: Oh, here we go.

Alex: Hear me out. People use these tools. My engineers use them, the whole under-thirty crowd uses them, half the country opens one every week. What people hate isn't the product, it's the press tour. It's executives talking about replacing work on a stage. I'm a builder. I think shipping beats announcing, and my view is that if the gains show up in people's actual lives, the mood follows.

Sam: And I think the part that matters is the exact opposite. I sign off my team's AI tool budget, I know every line to the cent, and I'm the person who has to tell a whole department what this means for them. The anger isn't a misunderstanding of the product. It's a very sensible reading of who keeps the money. If it's about power, no better product fixes it.

Alex: So you think it's rational, and I think it's misdirected.

Sam: I think you think the problem is communications.

Alex: That's a little unfair. And a little accurate. We'll see who moves.

Sam: So, the name. Everybody's heard techlash.

Alex: Which is old now. The Economist used it more than a decade ago, and it made the Oxford Dictionaries word of the year shortlist a few years after that. But techlash was about privacy and social media and monopoly. What's forming around AI is broader, and more political.

Sam: And the word that's stuck this year comes from a writer, Jasmine Sun, who put out an essay in April called AI Populism's Warning Shot. Her idea is that people don't see AI as something you adopt. They see it as an elite political project you resist.

Alex: And that essay landed after the worst week the industry has had. A young man threw a Molotov cocktail at Altman's house in San Francisco. A gate caught fire. He was arrested an hour later at OpenAI's headquarters.

Sam: The New Republic reported he'd written a manifesto calling for violence against AI executives, and called himself a butlerian jihadist, which is a Dune reference, the war on thinking machines.

Alex: Days earlier, somebody fired a volley of shots into the Indianapolis home of a city councillor, Ron Gibson, who'd backed a local data centre project. There was a note on the door. It said no data centers.

Sam: That's grim. And I want to be careful here, because those are extreme outliers. A handful of people. The real evidence isn't in the violence, it's in the polling.

Alex: Good, because the polling is the interesting part. In September, Data for Progress surveyed thousands of likely voters across Iowa, Michigan and Ohio. Majorities in all three states had an unfavourable view of the AI industry, and about three in four disagreed that AI executives care about people like them.

Sam: But here's the part that stopped me. Only about two in five of the working voters were worried about their own job. Roughly half rarely or never used AI at all. And among the people who said AI had touched their lives, far more called it good than bad.

Alex: So it's not personal experience, and it's not job fear.

Sam: It's about who profits, who's watching and who's in charge. And that's what you said you didn't buy, Alex. People aren't confused about the tool.

Alex: I'll take the hit on the first round. But notice what the data doesn't say. It doesn't say the product is the problem, and it doesn't say builders can't do anything.

Sam: Noted. You're still wrong.

Alex: Provisionally, partially wrong.

Sam: Can I push on one thing? Does the violence poison the well?

Alex: For the people who were never going to listen, yes. It lets the industry say it's all cranks.

Sam: And that's the trap. If the industry decides the story is a few cranks, it never has to read the polls.

Alex: Same mistake as a company deciding every angry customer is a troll. My support team learned that one the hard way.

Sam: Mine too. The loudest complaint is rarely the most representative, but the quiet ones just leave.

Alex: And the quiet ones are what a poll is.

Sam: Should we call it a Luddite thing? Because I've heard that a lot.

Alex: Neo-Luddite misreads it, and I'll come back to the actual Luddites, because it turns out I have feelings about them. Most of the critics aren't trying to smash the machines.

Sam: And the phrase AI populism earns its keep because it puts the fight where the data says it is, between a public and an elite it thinks is capturing a technology.

Alex: Which tells you what would answer it. Not a better product. A different deal.

Sam: And there's another reason it's not just a mood. Building AI at the scale the labs plan needs land, grid connections, water, tax treatment, and the permission of elected officials. Every one of those runs through voters who are souring.

Alex: I run an engineering org, so I'll say it in my terms. I've been planning compute like it's a purchase order. It's not. It's a permit, and a permit is a political object.

Sam: And the labs are heading to public listings, which turns ordinary investors into owners, right when the industry's own political money is pouring into the midterms.

Alex: So the mood is now a constraint on the build-out. That's why this is an episode and not a tweet.

Sam: And why I stopped skimming the polling emails my analytics team sends me.

Alex: You were skimming them?

Sam: I was skimming them like a person who thought this was a sentiment story. It's a supply chain story.

Alex: For most people AI is something on a screen. A data centre is a building the size of several football fields at the end of your road, pulling as much power as a small city. That's where the populism got a face.

Sam: And the scale of the pushback is new. A tracker called Data Center Watch counted dozens of projects, worth over a hundred billion dollars, blocked or delayed in just the first few months of this year. That's roughly as much as all of last year.

Alex: Opposition groups more than doubled since the end of last year, and they're in nearly every state now.

Sam: Hundreds of state bills in the first weeks of the year. Moratorium proposals in over a dozen states. And a second tally added tens of billions more in the disrupted column, though that one mixes delays with outright cancellations, so I'd hold it loosely.

Alex: Here's the one I keep coming back to. Virginia. Data Center Alley, the cluster that hosts more of this than anywhere. The Washington Post and the Schar School polled registered voters there in March. Back in 2023, about seven in ten were comfortable with a new data centre nearby. Now it's about one in three.

Sam: Fewer than half as many. And support for the tax breaks used to lure them collapsed right alongside.

Alex: Same poll, a majority said data centres hurt household energy bills, and only a sliver said they help. Though a majority still credit them with creating jobs.

Sam: Which is what makes it so interesting. They're not saying there's no upside. They're saying the upside isn't theirs.

Alex: And the comparison that got me was nuclear. A few years back, only about a third of Americans said they'd be comfortable with a nuclear power plant nearby. Data centres are approaching that level of local tolerance.

Sam: A reactor and a server farm, neck and neck.

Alex: Not saying it's a perfect comparison.

Sam: It's an excellent comparison for a dinner party.

Alex: It's a decent one for a board meeting, too. Whatever I write in a capacity plan, the answer that matters is whether the town says yes.

Sam: And this one lands for me, because my sister Claire lives in a small town east of Atlanta where one of these is going up on old farmland. She and her husband own a few acres close to the site. She goes to the county meetings. She reads the minutes. She's the only person I know who reads minutes for fun.

Alex: What does she say?

Sam: That every promise gets repeated at every meeting, and none of them shows up next to her power bill. Georgia Power has publicly pledged that residential customers won't pay for data centre growth. They've talked about a rate freeze. But customers say their bills have climbed by tens of dollars a month since 2023, and critics point at a February filing about gas plants for data centres in a couple of counties.

Alex: And how much of that is the data centres is contested.

Sam: It is. The utility says one thing, advocates say another, and nobody can sort that out from the passenger seat. But the point isn't who's right about the bill. The point is a family next to the fence has decided they're paying for something that isn't theirs.

Alex: That's the cleanest version of the whole populism argument. When the costs are local and visible and the gains are distant and private, consent evaporates.

Sam: Okay, but who pays for that? Because that's what Claire asks, and nobody at the meeting answers it.

Alex: Which is why the move companies are starting to make is the interesting one. Microsoft launched what it calls Community-First AI Infrastructure in January. It promises to cover the utility costs its sites create and limit water use. No independent oversight, though.

Sam: And OpenAI's April policy paper floated the same idea: data centres paying their own energy costs and bringing local jobs and tax revenue.

Alex: Every one of those promises is the industry quietly conceding the point. You're offering a town a share of the gains, not only a share of the noise.

Sam: Which is what I've been saying. A bargaining position, not a misunderstanding.

Alex: I'll let you have that one for now.

Sam: Think of it like a neighbour who throws a huge party. The music's loud, the street's full of cars, and then he hands you the cleanup bill.

Alex: And you weren't invited, and you've noticed he's very rich.

Sam: That's the vibe in the minutes.

Alex: Microsoft's pledge has a gap in it, though. Nobody outside checks it.

Sam: A promise with no audit is a press release in a hard hat.

Alex: You've been saving that one.

Sam: Since the morning I read the announcement.

Alex: It's now at the top of national politics. In March, Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez introduced a bill for a national moratorium on new AI data centres until federal rules exist.

Sam: Which has very little chance of passing. And Interior Secretary Doug Burgum called a moratorium a surrender flag to China.

Alex: Then in April, Maine's legislature passed the first statewide pause on new data centres above a modest size. Governor Janet Mills vetoed it, and not because she disagreed. It lacked an exemption for a half-billion-dollar project in the town of Jay. She said she still favoured a moratorium.

Sam: Think about that. A governor vetoes the pause because it would've hit one project in one town, and says she still wants the pause.

Alex: And in Wisconsin, the Marquette Law School poll found the share saying data centres' costs outweigh their benefits went from a bit over half last October to about seventy percent in February and a little higher by July. The swing came from Democrats and independents. Republicans were already sceptical.

Sam: So it isn't one party's issue. That's the part that scares me, as someone who's supposed to be calm and strategic. It's bipartisan and it's getting stronger.

Alex: Yeah. I go quiet when I see a curve like Wisconsin's.

Sam: This is the part that confirmed it for me. Pew has asked the same question since 2021. Does the increased use of AI in daily life make you more concerned or more excited?

Alex: Back then about a third said more concerned. Now it's over half. And the excited share has shrunk to under one in ten.

Sam: And the shape matters. The jump came in a single year, the year generative AI hit the mass market, and concern has never come back down. This year, for the first time, a majority of Americans under thirty said they were more concerned than excited.

Alex: Gallup's survey with the Walton Family Foundation found Gen Z's excitement dropped in a single year, and anger now sits above excitement.

Sam: Which is my daughter's generation. Maya is a junior, she's starting to think about colleges, and she uses AI every day to quiz herself and plan essays. She's also the angriest person at our dinner table about it. Both of those are true at once, and I used to think that was a contradiction.

Alex: It isn't. It's the whole story. Pew found most Americans expect AI to mean fewer jobs over the next couple of decades, and that share has grown. Quinnipiac found a majority expected more harm than good in their daily lives, up sharply in under a year.

Sam: And NBC had far more voters viewing AI negatively than positively, a net rating slightly below Immigration and Customs Enforcement. Pew had it... no wait, that was NBC. I keep mixing the two up.

Alex: You said Pew and meant NBC. The five-year question is Pew. The ICE comparison is NBC.

Sam: Thank you, I'll put that on the wall.

Alex: And you run a team that writes copy.

Sam: Which is why I have an editor.

Alex: The Stanford AI Index put the cleanest gap on paper. Almost three in four AI experts expect AI to be good for how people do their jobs, against fewer than one in four Americans.

Sam: And on the economy the gap is just as wide. Most experts expect a positive effect. Hardly anyone else does.

Alex: And I'll admit this is the one that makes me uncomfortable, because by that definition I'm an expert. There's a post on Hacker News this week quoting a Meta engineer. He said he's never felt this uncomfortable doing his job. The code feels like a black box, and the pace of shipping is moving so fast that checking every line isn't possible.

Sam: So even inside the building the consent is thin.

Alex: Even my own engineers say it, in quieter words. The people who understand this best are less sure they control it. That's a cousin of what voters are saying: the system around the tool doesn't feel like it was built with them in mind.

Sam: And here's the paradox you need to say out loud, Alex. Use is rising. NBC found over half of voters had used an AI platform recently. Gallup found half of Gen Z uses it weekly.

Alex: So what I said at the top, that the anger is aimed at the wrong thing, doesn't survive this. If it were about the product, usage would fall with approval.

Sam: You're conceding.

Alex: I'm conceding that a better product doesn't fix the mood. I thought if the gains reached people the mood would follow. But the mood has already come apart from the gains people can see in their own hands. That's a reason, and it moved me.

Sam: Write that down. Did the room hear that? Alex Rivera changed his mind, with a reason.

Alex: Don't get used to it. I haven't said who writes the deal.

Sam: Now here's what complicates the story, and I love that it does. This isn't a global slide into pessimism.

Alex: Ipsos runs the biggest annual cross-country survey. The share saying AI products offer more benefits than drawbacks actually rose in a year, even as the share saying it makes them nervous edged up too. Their latest edition is even wider, and excitement and nervousness are about level worldwide, often inside the same person.

Sam: And the geography is stark. In China, Indonesia and Thailand, around four in five adults see more benefit than harm. In Canada, the United States and the Netherlands it's about two in five.

Alex: Pew's survey across a couple dozen countries had about half in the United States, Italy, Australia, Brazil and Greece more concerned than excited. South Korea was among the lowest.

Sam: And trust in your own government to regulate it. In Singapore and Indonesia, most people have it. The United States came dead last of every country surveyed.

Alex: There are exceptions that make me suspicious of any tidy story. Germany and France actually warmed since 2022. And India is the one to watch, because nervousness there rose faster in a single year than anywhere else surveyed.

Sam: So what's the pattern?

Alex: Where AI arrives with rising incomes, trusted governments and a feeling of national ascent, the public is optimistic. Where it arrives on top of stagnant wages, low trust and visible inequality, the same technology reads as a threat. The backlash tracks the social contract more than the software.

Sam: That's the line I'd put on a slide. And as a marketer I hate that no amount of copy fixes it.

Alex: The physical fights are global too. A New York Times investigation traced them from Mexico, where residents near a Microsoft site reported water outages lasting weeks, to Ireland, where campaigners won limits on new data centres around Dublin.

Sam: In Chile, Google dropped plans for a data centre in Cerrillos after protests about water use in a drought-stricken area. And in London, PauseAI and Pull the Plug led what MIT Technology Review called the city's biggest anti-AI march yet, a few hundred people through King's Cross. In San Francisco a similar crowd marched past the offices of Anthropic, OpenAI and xAI.

Alex: A few hundred. I want to be straight about that. It's the biggest, and it's a few hundred.

Sam: Which is why I think the polls matter more than the marches.

Alex: Agreed. Marches are a headline. A poll in Virginia is a permit.

Sam: So who's absorbing this? Because nobody tracks each AI company with the same question over years. The public formed views faster than the pollsters could measure them.

Alex: Piecing it together, OpenAI is the best known and the most exposed. In the battleground survey, a majority of voters in all three states viewed it unfavourably. Altman is unknown to about three in five voters, but the ones who know him are negative. His net rating was around minus twenty in every state.

Sam: And in the Axios Harris Poll 100, the big corporate reputation ranking out in May, OpenAI fell by double digits, going by Quartz's reading of it. And it turned partisan. Republicans now rate it well above Democrats, where two years ago there was almost no gap.

Alex: Its president, Greg Brockman, is a major funder of Leading the Future, the super PAC fighting state AI rules. Which matters, because the company's most visible faces and its political money point the same way voters already distrust.

Sam: Anthropic is the interesting one for me, because it sells itself as the safety-first lab. It's reportedly committed tens of millions of dollars to Public First Action, a group pushing for guardrails.

Alex: And it didn't buy a premium. About half of battleground voters hadn't heard enough to rate it, and the ones who had leaned negative. Being the pro-regulation lab hasn't bought it goodwill yet.

Sam: Meta carries an older deficit. Two-thirds of Americans viewed Zuckerberg unfavourably in Pew's survey last year. Musk is polarising rather than simply unpopular, split sharply by party, and Tesla's reputation actually rebounded this year.

Alex: Microsoft, the company most tied to OpenAI, dropped more than twenty places in the ranking. And then there are the firms that kept their standing. Nvidia sat near the very top. Apple, Alphabet and Amazon scored very good.

Sam: Because they sell things people already understand.

Alex: Reputation follows who a company appears to serve. The labs whose product is AI itself, whose founders talk openly about replacing work and whose valuations depend on it, are taking the anger.

Sam: And as someone who has to choose which of these vendors my brand sits beside, that's not abstract. I do not want my pet food company's name next to a lab voters think is eating their jobs.

Alex: That's a real stake. Mine is the opposite. I need these vendors to still be standing in three years, and a lab that needs public consent to build and public money to list is a vendor risk I'd never put on a slide before this year.

Sam: I notice the same thing from marketing. The brands people forgive show up in their life as an actual thing, not a promise.

Alex: Shoes get forgiven. Models don't.

Sam: Shoes don't give interviews about replacing your job.

Alex: Mine would be an odd company if they did. Anyway, that makes it a vendor risk.

Sam: So your board should be reading polls.

Alex: My board is going to be reading polls. Here's the intuition I hear everywhere, including at my own dinner table. AI is sucking capital out of everything else and pushing prices up for everyone.

Sam: And I'll own it, that was mine too. I assume the worst about pricing. If a bill is up, it's the data centres.

Alex: The evidence backs half of it. The capital is real. The four biggest cloud companies, Amazon, Microsoft, Alphabet and Meta, guided this spring to roughly seven hundred billion dollars of spending this year, up from about four hundred billion last year.

Sam: And ING estimates AI investment accounts for roughly a third of US growth this year. The IMF credits it with holding up growth worldwide.

Alex: So it's propping the economy up. But this year's price squeeze has another main cause. Inflation hit three point seven in September, but strip out food and energy and it's well below that. Energy was up around a fifth, gasoline by about a third, after Iran's Revolutionary Guard declared the Strait of Hormuz closed in March.

Sam: And electricity, where data centres really do add pressure, rose about four percent. Under the headline.

Alex: So the fuel shock is the big story. How much of the electricity move is data centres, versus fuel costs and grid upgrades, is disputed, and public data can't cleanly separate them.

Sam: So my worst-case pricing theory is half wrong.

Alex: Sam Okafor, assumes the worst about pricing, learns the worst is mostly oil.

Sam: I'm taking the correction and I'm still angry. Because the perception isn't irrational.

Alex: Go on.

Sam: Mark Zandi at Moody's estimates that over three years, real spending by the top fifth of households grew by about four percent, and everyone below them saw close to zero. And the Federal Reserve's numbers show the top one percent hold about half the stock Americans own, directly or indirectly. The bottom half holds almost none.

Alex: So the boom shows up in share prices, and most people don't own the shares. It's a party with a guest list.

Sam: And the payoff hasn't reached most firms either. A February study from the National Bureau of Economic Research asked thousands of executives in four countries. About nine in ten reported no effect on productivity or employment over three years.

Alex: That one lands on my desk. I defend an AI budget to a board, and the straight answer to what we got for it is we're still learning. Not a great answer.

Sam: So AI isn't causing the cost-of-living crisis. It's making a small group very rich during one, and its benefits to everyone else are a forecast.

Alex: Which is exactly the condition for a fairness backlash. And the costs land on the young first. My son Diego graduated in May. He's working shifts at a running store and applying for first jobs, and the rent takes more than half his pay. His loan payments start in November.

Sam: Oh, Alex.

Alex: He's fine. He's resourceful. But it's why this finding doesn't read like a statistic to me. Stanford's Digital Economy Lab found employment of workers in their early twenties in the most AI-exposed occupations is now nearly a fifth below where it would have been, and the gap has widened since the study's first version.

Sam: A ResumeBuilder survey of recent graduates found about two-thirds had been ghosted after interviews. That's one survey, so I won't lean on it. The New York Fed had recent-graduate unemployment well above the national rate at the end of last year.

Alex: And Indeed's economist called the market low-hire, low-fire. Nobody's being let go, nobody's being let in.

Sam: Which is what I said last time. The first rung is missing.

Alex: And last time I said the economic part is fixable. I'm less sure after today. If the first rung is gone, the fix isn't a product. It's a different arrangement of who gets hired and who gets the gains.

Sam: A different arrangement. You're talking like a populist.

Alex: Talking like a father who owns a spreadsheet.

Sam: So you promised to come back to the Luddites.

Alex: I did, and before you say it, this is not a tangent.

Sam: It's the beginning of a tangent. You get a certain voice.

Alex: It's a historically relevant tangent. In 1811, textile workers in Nottinghamshire started breaking the knitting frames that were undercutting their trade. They weren't against machinery as such. They went after the machines owners used to cut wages and skill. Parliament made frame-breaking a capital offence the next year, and the mass trials at York ended in executions and transportation to Australia.

Sam: And the name became an insult for anyone afraid of technology, which is a misreading of what they fought about.

Alex: They fought about distribution. And what they lived through has a name. The economic historian Robert Allen called it Engels' pause, after Friedrich Engels, who described the misery of Manchester's workers in the middle of the century.

Sam: Give me the number.

Alex: Over sixty years, output per worker in Britain rose by nearly half. Real wages rose by about an eighth.

Sam: Sixty years.

Alex: Sixty years where the gains went to profits. Then in the next sixty, wages finally outran output. Historians still argue about the exact figures because the wage series are reconstructed, but the shape is accepted. Allen's explanation is that owners captured the return, and wages caught up only once capital piled up enough to make labour scarce and employers had to bid for it.

Sam: Picture a bakery that doubles its ovens and keeps the wages where they were. The bread piles up, the owner gets rich, and the bakers wait.

Alex: For a couple of generations.

Sam: And you can feel the modern version in my building. Nobody's smashing frames. They just quietly stop trusting that the tool is on their side.

Alex: Not sabotage, withdrawal.

Sam: And it's rational. If the draft you write with the tool is really the tool's draft, whose promotion is it?

Alex: That's the Luddite question in a hoodie. Who owns the skill once the machine has it? So the public's pessimism isn't ignorance. It's an accurate read of the first phase of a general-purpose technology.

Sam: And the backlash is not a detour.

Alex: On Daron Acemoglu and Simon Johnson's reading, it's the mechanism. Their book, Power and Progress, runs that pattern across a thousand years. Technology raises output, but gains get shared only when the direction of innovation and the split of proceeds are pushed away from a narrow elite.

Sam: And they later shared a Nobel for their work on institutions.

Alex: The Luddites lost their fight. Workers' gains came later, as their bargaining power grew. The question for AI is whether the bargaining happens in decades, like then, or faster.

Sam: The computer age rhymed with this, didn't it?

Alex: A milder rhyme. In 1987 Robert Solow wrote that you can see the computer age everywhere but in the productivity statistics. When productivity did arrive in the late nineties, it was shared unevenly. The Economic Policy Institute calculates that US productivity has risen by close to ninety percent since 1979, while a typical worker's hourly pay rose by about a third.

Sam: And you were in the middle of that.

Alex: Back in oh-one, I was building a start-up and I'd have told you every dollar pouring in was infrastructure for a shared future. I meant it. The infrastructure was real. The sharing took a while.

Sam: The man who had a spectacular failure lecturing us on how history rhymes.

Alex: That was a different failure, thank you. Later and louder.

Sam: Let me say what it means for Monday. I can't fix who gets the gains of an industrial revolution. But my team asks me every month whether they're on the right side of this, and I can't say yes because the tool is good.

Alex: And I can't tell my engineers to be patient for sixty years.

Sam: Yeah. That's the whole problem with it, as a boss. There's a strand here I didn't expect. The grandest promise for AI isn't only abundance, it's flourishing. Everyone's rich and fulfilled. Behind it sits an assumption that wealth and happiness move together.

Alex: For individuals, they mostly do. A few years ago, Matthew Killingsworth, Daniel Kahneman and Barbara Mellers settled a long dispute in the journal PNAS. The older claim was that happiness stops rising at a modest income. Re-analysing a big sample of real-time mood reports, they found it keeps rising with income, even at high incomes. Only for the least happy fifth does extra money stop helping, because what makes them unhappy is something money can't fix.

Sam: For whole nations it's weaker. The economist Richard Easterlin found, across dozens of countries, that over a decade or more rising national income didn't bring rising life satisfaction. Betsey Stevenson and Justin Wolfers dispute that.

Alex: But the common explanation is that people judge their lives against other people, and they value security and control as well as consumption. When everyone gets richer together, the comparison doesn't change.

Sam: And if AI lifts the average but concentrates the gains, and people lose control over their work and their towns, you could be richer on paper and unhappier in fact.

Alex: What people object to is a loss of standing and say, not a lack of gadgets. That one sentence explains the county meetings and the rent at once.

Sam: Abundance alone doesn't buy consent. Security, dignity, a fair share. That is not a marketing problem.

Alex: You say that like you've tried.

Sam: I've tried. Everyone in my job has tried.

Alex: The industry's leaders once offered an answer. Years ago, Altman published an essay called Moore's Law for Everything. It proposed an American Equity Fund, financed by an annual levy on the market value of large companies and on private land. He estimated it would pay each American adult thousands of dollars a year within a decade.

Sam: And last year, Dario Amodei, who runs Anthropic, warned that AI could eliminate half of entry-level white-collar jobs within five years and push unemployment into double digits.

Alex: The evidence on basic income itself is more modest than either the dream or its critics suggest. OpenResearch, a lab Altman funded, gave a thousand low-income Americans a thousand dollars a month for three years. Recipients worked a little less, spent more on essentials, helped relatives, and their mental-health gains faded after the first year.

Sam: We went deep on that in our episode on universal basic income, number thirty-one, back in July. The short version: cash reliably eases poverty, but it can't replace the purpose and status that work gives.

Alex: What's collapsed isn't the appetite for sharing. It's faith that the gains will be shared voluntarily. Years earlier, a Gallup-Northeastern survey had Americans split almost down the middle on a basic income for workers displaced by AI.

Sam: In June, a Verasight poll found about seven in ten backed requiring large AI firms to hand half their stock to a public wealth fund. And most still did when the idea was attributed to Bernie Sanders. Just Capital found about two-thirds think everyone should get a direct financial benefit from the wealth AI companies generate.

Alex: And in the battleground states, majorities including Republicans supported a basic income funded by a tax on AI companies.

Sam: Meanwhile OpenAI's April paper, Industrial Policy for the Intelligence Age, floated a Public Wealth Fund, higher taxes on top capital gains and on sustained AI-driven returns, and pilots of a shorter work week at full pay.

Alex: Critics noted it didn't say how the fund would be financed.

Sam: So a fund that isn't funded, from a company whose own president bankrolls a super PAC against state rules. That's not shipping, Alex. That's a press release.

Alex: I'd defend it more if it had a repo. Though I'll defend the instinct. A public wealth fund is at least an admission the question is on the table. For years the industry's answer was that growth would solve it.

Sam: Growth solves nothing if the person it's growing for never sees it. I've watched a company's revenue climb for years while my team's budget stayed flat.

Alex: That's a CFO problem as much as a distribution problem.

Sam: It's the same problem at every scale. The lesson is the one Acemoglu and Johnson draw. Shared prosperity is negotiated, not gifted. The public has absorbed that faster than the industry.

Alex: The dream hasn't died. It's changed hands, from a gift the builders would give to a claim the public intends to make.

Sam: Which brings us to the midterms, on the third of November. The two industry-aligned networks, Leading the Future on one side and the Anthropic-funded Public First on the other, had spent tens of millions by late September, according to a New York Times tally.

Alex: And none of their ads mentioned data centres. NBC found the early ads didn't mention AI at all.

Sam: When an industry's own political money avoids naming the industry, it has understood how toxic its brand has become. And Data for Progress found three in four battleground voters less likely to back a candidate funded by AI super PACs.

Alex: It changes what winning means for the labs. Compute still matters. So do sites that can be built, power that can be secured without raising neighbours' bills, and a brand voters don't distrust.

Sam: Which favours the companies with old trusted consumer brands and their own energy deals.

Alex: And it penalises pure-play labs that need public money for listings and public consent for construction. We covered the valuation side of this in our AI bubble episode, number forty-nine, back in August.

Sam: The most likely path, on the current evidence, is a negotiated bargain. State rules making data centres pay their own grid costs, community benefit agreements, disclosure and safety laws, some public stake in the gains.

Alex: A harder backlash is plausible if the midterms reward candidates who run against the industry. Moratoria and a sceptical Congress. And the gentlest path, where the productivity gains reach paycheques fast enough to change the mood on their own, is the least likely near-term. Nothing in the firm-level data or the junior-hiring numbers says it's imminent.

Sam: And the signposts are specific. Do candidates who campaign against AI money win? Do more states pass data-centre rules, or does one finally impose a moratorium? Do the labs' listings include any mechanism for public ownership? And does entry-level hiring in AI-exposed jobs recover?

Alex: A clear yes on the first two says the harder road. A yes on the last two says the gentler one.

Sam: And on Monday, for me, the move is boring. I stop buying tools that promise to replace a person, I buy the ones that make my person better, and I say so out loud to my team.

Alex: Mine is to treat power and permission as line items in every build plan, not assumptions. And to stop talking about replacing work in rooms where my engineers can hear.

Sam: That one's free.

Alex: It's the cheapest item on the roadmap.

Sam: And this is where we end up not agreeing. Say yours.

Alex: Here's the thing. I came in thinking the anger was misdirected, and the use-up, approval-down pattern took that away from me. It's a claim on the gains, and it's a fair one. But I still think the industry can write part of the answer before a legislature does. Builders who put a site fee and a share of the gains on the table first will end up with better towns than the ones who wait to be forced.

Sam: I don't believe that. Not because builders are bad people. Because every dividend a town like that one has seen arrived after the meetings, not before. A voluntary gift is something you can withdraw. A law, you can read. I came in saying no product fixes this and I'm leaving there. You moved me on one point: prices. I was blaming the wrong bill.

Alex: So I moved on the cause, you moved on the bill, and we split on who writes the deal.

Sam: And I'd rather split honestly than agree early.

Alex: That's the most Sam sentence ever said.

Sam: It's on a mug. If you take three things from this, the first is that the backlash isn't about whether AI works. People use it more every month and approve of it less, and that gap is the sign it's about power.

Alex: Second, the anger is concentrated in the West and among the young. It's sharpened, not caused, by a price squeeze that comes mainly from oil. And it has a physical target: the data centre in somebody's town.

Sam: Third, history says this is the first half of a longer cycle. Britain's workers waited sixty years for the revolution to reach their wages. The question is whether the bargain gets struck in years or in generations, and whether the industry helps write it.

Alex: I hope you came away understanding a little more about where this is heading. It's a genuinely strange moment, and the horizon on what anyone can know is very short, and that's what makes it worth following closely.

Sam: One honest note on how this is made. It's AI-generated. Dan builds a custom stack of AI tools to research, analyse, verify and illustrate the questions worth understanding, mostly to learn them himself, and he publishes it for anyone who'd like to follow along. AI-assisted, fact-checked, worth a second look.

Alex: Before we go, one useful thing you can do: follow the show.

Sam: Whatever app you're listening in, there's a follow button or a plus sign.

Alex: One tap, and it's free. You get each new episode the moment it lands.

Sam: And for a small independent show, a follow is the biggest single thing that helps it reach other people trying to make sense of all this. So if this was worth your time, tap it now.

Alex: One more thing. The Virginia poll that flipped from seven in ten to one in three, and the British chart where wages trail output for sixty years, are laid out as pictures on connectiveshift.com.

Sam: They're free, they take a minute, and seeing the two curves next to each other makes this whole episode click.

Alex: Thanks so much for listening.

Sam: We'll see you next time.