October 18th, 2026
This post is one of my advice & arguments pages about the harms and hazards of the AI Hype Movement.
“As we approach the coming jobs cliff, we’re entering a period where a college isn’t going to be worth it for the majority of people, since AI will take over most white-collar jobs. Combined with the demographic cliff, the entire higher education system will crumble.”
This is the kind of statement you don’t hear that much from sub-CEO-level #AI boosters, because it’s awkward for them to admit that the tech they think is improving their life is going to be disastrous for society. Or if they do admit this, they spin it like it’s a good thing (don’t get me wrong, tuition is ludicrously high and higher education absolutely could be improved by a wholesale reinvention, but the potential AI-fueled collapse won’t be an improvement).
I’m in the “anti-AI” crowd myself, and I think the current tech is in a hype bubble that will collapse before we see wholesale replacement of white-collar jobs, with a re-hiring to come that will somewhat make up for the current decimation. There will still be a lot of fallout for higher ed (and hopefully some productive transformation), but it might not be apocalyptic.
Fun question to ask the next person who extols the virtues of using generative AI for their job: “So how long until your boss can fire you and use the AI themselves?”
The following ideas are contradictory:
If you’ve ever taken a macroeconomics class, you’ll have studied the theory of pricing based on supply and demand. Although there are a lot of problems with this theory, and at minimum the famous quote “The market can remain irrational longer than you can remain solvent” comes to mind, it’s interesting to apply the basic principle of supply and demand to the future that the AI boosters imagine. In their imagination (and in the claims they keep making, such as “Large language models have PhD-level intelligence”) a chatbot (or perhaps an “agentic AI” built out of chatbots) will be able to do the same job as a real human, across a variety of jobs, especially things like software programming.
In economic terms, if the output of a chatbot can really stand in for the output of a human worker, then the two goods are part of the same market, and the introduction of chatbots represents a huge increase on the supply side of that job market.
The result, assuming demand remains constant, is that the price of the good (i.e., the wage for workers and/or chatbots) will fall. Now in theory, a concomitant increase in demand could offset that effect, leaving the price the same or even higher if the demand increases faster than the supply. The particular shape of the supply and demand curves matters too, but the point I’m trying to make here is actually even simpler: from the macroeconomic perspective, if the “product” of a chatbot or agentic system is indistinguishable from the “product” of a human worker, then the price of the two things will be the same.
Put simply: the wages of jobs that AI can replace will become equal to the cost of paying for the AI. Right now of course, token costs are heavily subsidized, and the big companies like Anthropic and Open AI are losing money on every token. They’re doing this so that they can try to entice businesses to switch to their service believing it will be cheaper, and prices will come up to profitable levels once the clients are hooked and have fired their human staff. That’s still not even working well due to problems with quality (see this article on hiring humans to clean up after AI mistakes). But in the long run, unless LLMs are actually recognized as being far inferior to human capabilities (which they are), the effect will likely be to depress human wages substantially while the AI token costs come up to match or nearly match those lowered wages for comparable amounts of work.
At the macro level, what does this do to the economy? You’ll have a bunch of humans who cannot support themselves financially and have to seek new careers. The money that they would have been earning (and then spending, making the entire economy go) will instead be siphoned off by one of a few massive tech companies as token costs. Unlike the humans who would spend their paychecks on groceries, local businesses, and rent, the tech companies will spend that money on electricity, water, and more infrastructure investment. Electricity and water are costly not because lots of human labor is required to produce them, but because they depend on scarce natural resources. So money spent on those things also goes to big businesses who gatekeep those resources, not to lots of wage-earners who would then immediately spend it on groceries. Similarly, infrastructure investment employs a few skilled workers for construction and data center setup, but the lion’s share of the cost goes towards the companies managing those things, as well as investments in real estate and things like taxes: places where money is destroyed or hoarded, not spent again. So all of these imply a huge slowdown in the velocity of money, which combined with a huge unemployed population, is a recipe for disaster.
As an anarchist, I can be hopeful that that disaster might have some silver linings in terms of getting masses of people to understand more about the downsides of the hierarchies they’ve accepted and believe are necessary, but it’s still a frightening prospect. It’s very different from the “AI will take over the world” bogeyman that the AI hype movement itself loves to use as a distraction, but it’s a very real threatening economic consequence that could happen if the boosters are even somewhat right about LLM capabilities, or even if they’re wrong it could still happen as long as they can convince client businesses that the models are more powerful than they actually are for a while.
This is incidentally one of my strongest motivations for writing these articles: that it really matters exactly how many people get hoodwinked how hard by the AI hype movement, in real economic terms.