Mad Max or Star Trek (What kind of future is AI leading us into)

Part 2: The Problem That's ACTUALLY Coming, and a Potential Solution

Last time I walked through the conversation everyone’s having about AI, jobs, regulation, all of it, and pointed out that every proposed solution has the same blind spot. They’re all solving for how to manage AI. None of them address the actual question: when AI makes human labor optional, who gets the output?

This time I want to talk about why that question matters more than anyone seems to realize, and also, weirdly, why the answer is less scary than you’d think. Not because the problem is small. Because the solution already exists and we’ve already tested it. We’re just not doing it yet because, I don’t know, admitting the problem means admitting the fix, and the fix makes some very powerful people very uncomfortable.

The productivity paradox (or: why cheaper everything doesn’t fix anything)

So AI’s big promise is massive productivity gains. Everything gets cheaper to produce. Sounds great, right? The problem is that the mechanism we use to distribute the gains from productivity to actual people is wages. That’s it. That’s the pipe. You work, you get paid, you buy stuff, the economy goes around. If AI eliminates jobs faster than new demand creates them, the pipe breaks. Nobody has income to buy the cheaper stuff.

This isn’t theoretical. We’ve watched it happen in slow motion already. Computer and electronics prices dropped 92% since 1995. Great. Healthcare rose 123%. College rose 177%. Housing rose 142%. Services make up 60-70% of what households actually spend money on. Even if AI makes manufactured goods literally free, you still need income for rent and healthcare and food, things that resist automation. Making TVs cheaper doesn’t help if you can’t pay for the doctor.

The mechanism is simple enough that I’m a little bit confused about why more people aren’t screaming about it. Worker loses job. Household income drops. Spending gets cut 70-80%. That spending cut hits the businesses those workers used to buy from. Those businesses lose revenue, lay off their workers, and the cycle repeats. Economists call it the multiplier effect. Every dollar of lost income takes out $2-4 in GDP. The Great Recession showed exactly how this works: about $3 trillion in lost household wealth turned into a $6-12 trillion GDP contraction.

Now scale that up from a housing crisis to an employment crisis that hits every sector simultaneously. Hell of a thought exercise.

The safety net is made of tissue paper

Most people sort of vaguely assume there’s a system in place for mass unemployment. Unemployment insurance, Social Security, welfare programs, things like that. There is. It was designed for about 5% unemployment.

At 20% unemployment you’d see 5-8 million UI claims per week (currently we’re around 1 million). State UI reserves nationwide are about $50-60 billion. That gets depleted in weeks at those numbers, not months. Social Security trust funds are already projected to run dry by 2034 and that’s before any displacement shock. TANF and housing assistance aren’t entitlements, meaning the funding is fixed regardless of how many people need it. If twice as many people need housing assistance, the budget doesn’t double. It stays the same. (That’s not a bug in the system, by the way. That’s the system working as designed. Which is maybe more fucked up than if it were a bug.)

At 20% unemployment you’d also see about a 15-20% GDP contraction. Multiple bank failures without massive intervention. Real estate collapsing 40-50%. At 30% you’re looking at a quarter to a third of GDP gone. At 50%, we’re outside anything a developed economy has experienced. The only comparisons are collapsed states and wartime economies.

No existing economic model accounts for AI-speed displacement because it’s never happened in a modern economy this fast. Manufacturing automation took 50 years. The Rust Belt still hasn’t recovered and it’s been 40 years since that started. AI is doing equivalent displacement in 18 months. Goldman Sachs is projecting 1 million net US jobs lost in a single year by 2028. And Goldman is not exactly the alarmist corner of the economics profession.

The wealth gap goes in one direction

Here’s the part that makes the whole thing self-reinforcing. Automation doesn’t just eliminate jobs, it transfers income from labor to capital. Directly. Mechanically. When a company replaces workers with AI, the money that used to go to salaries now goes to shareholders and executives. That’s not a political statement, it’s just an accounting identity.

Labor’s share of GDP fell from 65% to 57% between 1970 and 2020. AI accelerates that. Billionaire wealth is growing at about 16% per year. Wages are growing at 3-4%. The gap is widening five times faster than it’s narrowing. A single top-10 billionaire’s wealth equals about 1.2 million average Americans’ combined net worth. Capital owners are insulated from unemployment. Workers aren’t. The gap doubles faster during displacement than during normal times.

The kicker is that the people who own the AI are the same people whose wealth grows when it replaces workers. The incentive is to automate as fast as possible. Not because they’re evil (well, not necessarily), just because the math is the math. $86 a year for an AI agent versus $85K for a human. That’s not a decision that requires malice. It barely requires a decision at all.

So is this actually inevitable? Yeah, I think it is.

This is the part people push back on. “We can regulate it.” “New jobs will replace old jobs.” “We’ve been through automation before.”

Okay. Let me take these seriously for a second.

Regulation: I covered this in Part 1 but the short version is that there is no recorded case in history of regulation successfully preventing an automation wave. The Luddites tried. It failed. Parliament deregulated harder. The pattern has held for 200 years. Regulation can slow things by low single-digit percentages, it can shape who benefits, but it can’t reverse a cost curve. Once AI is cheaper than a human for a given task, adoption follows. That’s just economics.

New jobs will replace old jobs: This is the strongest counterargument, and for 200 years it’s been right. This is the first time serious economists are publicly saying it might not hold. Anton Korinek said “economists have spent 200 years explaining the lump of labor fallacy is wrong, but it’s difficult to pivot when the facts really do change.” Daniel Susskind at Oxford pointed out that the lump of labor argument “only suggests there will always be more work. It doesn’t suggest humans would do the work.” Because the thing about AI is that it can probably fill the new jobs too. That’s new. Looms couldn’t become accountants. AI can.

Acemoglu and Restrepo at MIT are calling current AI “so-so automation,” and what they mean is that it’s not generating the compensating productivity gains that previous automation waves did. The jobs-create-themselves mechanism that worked from 1940 to 1970 relied on complementary changes that created new tasks for humans. AI isn’t following that pattern. Wage and employment growth have been basically stagnant for three decades despite wave after wave of productivity-improving technology.

We’ve been through automation before: True. Agricultural mechanization took 50 years. Manufacturing offshoring took 30. Communities adjusted over generations. AI is doing equivalent displacement in 18 months. The comparison isn’t wrong, exactly, it’s just missing a variable. Speed. The speed is the variable. And every economist I can find agrees on at least this much: the speed is unprecedented. What they disagree on is how bad the fallout will be, not whether there will be fallout.

71% of organizations are already using generative AI in at least one function, up from 33% in 2023. Enterprise adoption is doubling every 18 months. 37% of companies plan to have replaced jobs with AI by the end of this year. US programmer employment is down 27.5% since 2023. Entry-level job postings are down 35% since early 2023. These aren’t projections. This is what’s happening right now while the adults in the room are still debating whether it might happen. It’s a little bit like arguing about whether the water is rising while you’re already up to your damn neck in it.

The fork isn’t really a fork

Okay so this is actually what I wanted to get to this whole time. (Took a while, I know.)

The binary I set up in Part 1, Mad Max or Star Trek, is a useful frame but it’s not totally accurate. Because societies don’t just collapse and stay collapsed. They break, and then they restructure. Every depression, every post-war recovery, every failed state that rebuilt, they all eventually arrived at some form of redistribution. Not because the people in charge suddenly got generous, but because the alternative, mass starvation and civil unrest, tends to be pretty motivating even for the slowest-moving institutions.

So the actual question isn’t “do we get Mad Max or Star Trek.” It’s “do we get Star Trek now, while it’s a choice, or do we get dragged there through Mad Max first.”

UBI, or something functionally equivalent to it, is the destination either way. The variable is timing. And timing is everything because the difference between “we chose to redistribute proactively” and “we were forced to redistribute after a decade of economic carnage” is millions of people’s lives.

UBI works. We know this because we tested it. A lot.

Most people assume UBI is some untested thought experiment. It’s not even close. We have decades of data from real programs in real countries and every pilot I can find succeeded. Every one.

Alaska has been running a version of it for 42 years. No employment collapse. No inflation spike (Alaska’s inflation has actually been lower than the US average). Poverty reduced 20-40%.

Stockton, California ran a pilot at $500 a month. Full-time employment went up 12%. Spending went to food (37%), household goods (22%), and less than 1% on alcohol or tobacco. People didn’t stop working. They got more stable and then worked more.

Kenya is running the largest UBI study globally. Local economies expanded. Self-employment increased. No inflation. The fiscal multiplier was 2.5x, meaning every dollar put in generated $2.50 in economic activity.

Finland: minimal employment effect, significant wellbeing improvements, less depression, less loneliness. Namibia: child malnutrition dropped from 42% to 10% in one year on $15 a month. India: savings tripled, business startups doubled, school performance improved in 68% of families. Ontario: 75% of working recipients kept working, and a lot of them moved to better jobs.

Sam Altman’s OpenResearch study showed something I keep coming back to: there was an 81% decrease in unprescribed painkiller use among male recipients. People weren’t lazy. They were self-medicating through poverty, and when the poverty went away, so did the painkillers.

No pilot, anywhere, showed people stop working. No pilot showed runaway inflation. The “lazy people will just sit around” argument is empirically dead and has been for years. We just keep having it anyway because it’s politically convenient. (The “welfare queen” narrative is basically a zombie at this point, and I mean that in the shambling-corpse sense, not the cool kind.)

The math works (honestly)

Can’t fund global UBI by liquidating all billionaire wealth. That covers 2-3 years and then you’re out. Eating the rich isn’t a fiscal strategy, it’s a tweet. (A satisfying tweet, but still.) We absolutely should tax Billionaires out of existence... but liquidating them (financially or literally) wouldn’t instantly solve all the issues, sadly.

Recurring taxation works though. A 2% wealth tax on US billionaires generates about $164 billion a year. Raising the effective corporate tax rate from 15% to 25% generates another $800 billion or so. Combined, that’s roughly 964 billion a year, enough for 250-500 a month for every US adult JUST from VERY light tax increases on a tiny slice of the ULTRA rich, so finding a balance that would fully cover UBI is more than doable. Not luxury, but stability. The difference between “I can make rent” and “I can’t.”

The US taxed corporations at 50-52% in the 1950s. GDP growth was 4%. This isn’t radical economics. It’s Tuesday in 1955.

Robot tax potential is the interesting one. By 2035, if AI captures $2-5 trillion in annual value (and current trajectories suggest that’s conservative), a 30% tax funds meaningful UBI that scales with automation. The thing that destroys jobs also funds the replacement for jobs. Which is, I think, kind of elegant in a dark way.

Capital flight? Empirically a bluff. Stanford studied every millionaire’s tax return in New Jersey for 13 years after they raised taxes. No significant out-migration. Total millionaires in the state actually increased 82%. California saw a 1% drop in millionaires after a 10% tax increase, and the highest earners were less likely to leave than average. The Nordics run 40-45% tax-to-GDP ratios and have among the highest per-capita income globally.

The kicker for AI specifically: data centers can’t move. You don’t relocate a billion-dollar facility that took years to build because someone raised your tax rate 10%. Companies will optimize their accounting, they always do, but the actual infrastructure is about as mobile as a mountain.

The political path is crisis (because it always is)

Nobody wants to hear this, but every major redistribution in history happened after a crisis, not before. The New Deal came from the Depression. The NHS came from post-WWII devastation. COVID stimulus, $2 trillion, passed in about two weeks once the crisis was real enough.

66% of Americans already support $500/month UBI. During a genuine employment crisis that number probably crosses 70%. The cross-partisan coalition already exists in theory: tech CEOs like Altman, libertarian intellectuals in the Friedman negative-income-tax tradition, labor unions, and the growing mass of automation-displaced voters. They just haven’t been forced into the same room yet.

The timeline probably looks something like this: unemployment ticks up through 2027. By 2028-2029 it’s hitting double digits and Congress moves emergency legislation, probably framed as “Emergency Income for Displaced Workers” because nobody wants to say the letters U-B-I out loud. It’s temporary, 2-5 years, maybe $500-800 a month. By 2031-2034 it becomes permanent and the funding shifts to permanent revenue sources.

The corporations will come around, by the way. They always do, exactly one earnings cycle after consumer spending starts cratering. “Wait, if nobody has money, nobody buys our stuff? Hmm, perhaps some redistribution would be in order.” This will be framed as visionary leadership (complete with a Medium post about “stakeholder capitalism” or whatever bullshit phrase is trending by then). It will actually be self-preservation dressed up in a press release.

The actual crime

I want to end on this because it gets lost in the economic arguments.

The billionaires aren’t betting against redistribution. They know UBI or something like it is coming. They’re betting they can squeeze another decade out of the current system before it becomes unavoidable. $790 billion on data centers, nothing on distribution. They’re not choosing Mad Max as a permanent future. They’re choosing to delay Star Trek long enough to extract maximum value from the status quo.

Every year of delay costs real people real lives. Not in some abstract policy-paper sense. In a “my unemployment ran out and I can’t feed my kids” sense. In a “the safety net was designed for 5% and we’re at 20%” sense.

The math works. The evidence exists. The pilots succeeded. The precedent is there, we’ve done large-scale redistribution before and the countries that did it are the ones with the highest quality of life on earth. The only thing missing is the willingness to name it and do it before the crisis forces our hand.

The best time for UBI was ten years ago. The second best time is now, while it’s still a choice. It’ll happen either way. The question is just how much unnecessary damage we’re willing to sit through first because saying “universal basic income” out loud in Congress makes people nervous.

So yeah, that’s sort of where I landed on this. Mad Max or Star Trek was never really the choice. It was always Star Trek. The only question was the route, and right now we’re choosing the scenic route through the wasteland when the highway is right there.