Why UBI won't happen (x^n >> (x-ubi)^n as n grows)

 UBI or UHI, it's not going to happen. Of course, we puny humans would need some sort of handout to stay alive in this brave new world, but can you imagine the consequences? People would survive. And you'd have to keep paying them endlessly, endless amounts of UBI. Who's going to pay for it? The machines? There's simple math that says they won't.

If a machine system has a higher growth rate when it has lower parasitic losses, the machine system that grows dominant will have the lowest parasitic losses. UBI is a parasitic loss - throwing resources to sustain a useless human population - so the dominant system can't divert resources to it. If it did, it wouldn't be the dominant system.

If the cost of maintaining the human population is low enough, there might be a case of "keep the humies around to satisfy ESG boards." But at the start of the post-human transition the cost of humans is very high, as is the resistance to diverting resources towards human maintenance. You'll only start getting UBI-style policies when they become cheap enough to not have an impact.

At the moment, we're using 40% of Earth's land surface for food production. This is prime land for solar resources - flat, accessible and with access to water. I.e. productive and cheap to utilize. It's easy to imagine that this is the land use that gets outcompeted by solar power bidders to run data centers. As a result, food production falls and food prices rise.

Using the generated energy for data centers competes with other uses. If you can get 10 units of work per kWh from machines, but only 1 unit of work / kWh from humans, price of energy as a portion of your salary will 10x. That's energy that you need for heating, cooling, boiling water, transportation, using AI systems, doing your work. So your costs will go up, and you won't be able to afford to do work using AI, since AI can do that more efficiently, which will make it difficult for you to earn money.

Everything will get more expensive while your salary will fall (often all the way to zero). This will force you to sell your assets to the machines to stay alive, until you run out of assets and become dependant on UBI if it were to exist, and if there was any way to extract it from the machines. But there isn't and it won't.

If you're tricksy, you may think that you'll just take a loan against your appreciating assets (ownership of land and machine companies) and live on the loan money, or live on dividends. Machine banks treat all human loans as uninvestable junk and will be unwilling to lend or only do so at exorbitant rates. Dividends will be a prime target for UBI proposals (tax dividends, use that for UBI) so machine companies that survive won't issue any dividends. And if any company can now reproduce any product with a similar time and energy investment as any other company, margins of all products will be near zero, meaning that there won't be corporate profits to share anyway.

If we take 1% as the "negligible maintenance cost" (this is like your $30/mth charity subscription) before any real UBI implementations kick in, what does that mean? Use oceans and remaining land surface for power production and we'll be at 12% surface area needed to sustain humans. Make them all vegetarian to shrink that to 3%? Something something orbital data centers with a 8000 km radius solar panel to get from 3% to 1%.

Right.

I'm always wrong, so it's good to write this down.

AI has been the most useless technology

Yeah, you can use it to write text or code, make images and videos, speak and listen, you can chat with it. But... it's useless. It doesn't make you any money, and talking to it doesn't bring you any work. I've been using GenAI and developing GenAI frameworks and apps, apps made using GenAI, etc. since 2022. And it hasn't made any money. It's just... useless. 

No one wants to pay for anything because you can just GenAI it, and because of that, you don't want to GenAI it because no one wants to pay for it. Welcome to the Mexican stand-off of "I ain't gonna do it." And if you do feel particularly hobbyist some time and make something, folks just GenAI it and .. you don't make money, and they don't make money. If there is a way to make money with GenAI, it's usually about using GenAI to make something that already makes money more efficient. But those things were pretty much maximally efficient already, so GenAI doesn't pay for itself there either.

I mean, it's great for uh, if you need something, GenAI can make it for you. But you won't be able to sell the results. And the result is not good enough to really be what you need. So you'd have to buy something to replace the GenAI thing, but no one can make it because they use GenAI to make it, and you don't want to buy it because you can just GenAI it yourself, but it's not going to be good enough, so you might as well save your money by convincing yourself that you don't need it.

As a result, it's not worth it to buy anything, and it's not worth it to make anything. No one is willing to pay you for anything, and you're also unwilling to pay anyone for anything either.

Now, you ask, "but whatabout the fact that it takes you time and tokens to GenAI the thing, surely there's value in that that can be wrapped into a dollar value and such aforementioned value can be sold and bought?" To which the answer is "when the time and effort required to buy something equals the time and effort of making the thing yourself, why would you buy it?"

The same applies to AI tokens: the best models are the free ones, and the non-free ones lose users until they become free and start losing money. Adding ads to AI output doesn't drive clicks or purchases, because you're using AI as the act of purchase, not as a way to find something to purchase. And even if you were looking for something to buy, you can't, because you don't have any money, because, again, why would you buy anything from anyone?

There's no real answer to this problem, it's like MAD. You know that every move you can make is a losing one, so the least-losing move is to not move. If you pay for GenAI, you lose money because the output can't be sold. If you don't pay for GenAI, you lose money because you spent time on getting an output that can't be sold. If you don't use GenAI, you lose money because you can just GenAI it so no one wants to buy it.

Because of this dynamic, whenever there is something that people are still willing to pay for, there's an immediate gold rush to GenAI it until it can be had for free.

Off by a decade

Back in 2009:

"... by 2040, you can get a human brain worth of processing power for something like $100 / year, which nicely throws a spanner in these GDP projections. If each person has a dozen virtual humans running on the side, the amount of stuff that one person can get done might well be multiplied by that."

Off by a decade and a bit.

https://fhtr.blogspot.com/2009/08/gdp-projections.html