Half-automated solarpunk lesbian earth socialism - our UBI future

Why we can give EVERYONE $40k - $90k UBI's

So everyone thinks AI is a bubble right now, and that’s fine - I think we still get to worry about AI counterfeiting jobs, and so think about UBI and whether it can work. Because after all, what’s the surest way for it to suddenly *not* be a bubble? If some company puts together the right scaffolding, prompts, persistence, and ongoing learning to actually counterfeit some jobs!

In the limit, this can lead to the infamous “fully automated luxury gay space communism,” that progressive utopian vision of yore. I don’t think we can quite get there, but I think we can get maybe halfway there (hence the post title), because I’m going to make the argument that we can give basically everyone a $40k - $90k UBI.

Sounds crazy? I agree!

Let’s at least go over the argument, and you tell me where I’m wrong.

There are already fairly compelling signs that AI is automating some entry level programmer and customers service jobs (as of 2022 and 2023, at that!):

And if you use the things yourself, they’re obviously already smarter and more useful than, oh, ~2/3 of entry level direct reports on specific tasks even from an expensive and highly skilled finance or modeling / data science baseline, while being essentially free.

And white collar jobs are ~40% of all jobs! Moreover, they’re almost certainly more than 40% of the economy, because they’re generally more highly compensated (and so by simple math, must be contributing more to overall GDP than non-white-collar jobs, because you can’t be paid more than the value you contribute long-term without your employer going out of business).

In fact, the formula for thinking about “how much of a company’s revenue is reflected in GDP” is essentially “value added,” or a decent heuristic would be “Revenue - direct COGS,” so you’re not double counting the company buying wheat from an ag company, then producing bread in GDP.

But the nice thing about white collar workers is that they’re not considered direct COGS - they’re after that. So if you consider a consulting firm or law firm or software company, minus some negligible expenses like software and rent amortized per worker, nearly every dollar of revenue from those workers is pretty much directly reflected in GDP, and this generalizes.

This greatly simplifies our math!

Alright, we’ve got 40% of jobs that are white collar.

If we consider them to be contributing to ~60% of GDP that ends at an average of ~$270k in GDP-contributed-value per white collar job at today’s ~$30T GDP - sounds directionally fine. After all, in my recent “America has already differentiated into castes” post, the average HHI for PMC men was ~$190k if you include the only partially PMC “Above Median” segment, and ~$290k for the mid PMC and highers. The $190k baseline would be a ~40% margin of value created on top of salary.

Now remember, they *must* be being paid less than the value they contribute to their companies / GDP on average, or their companies would go broke. Maybe it’s not a 40% margin, which seems kind of fat, but I don’t really know. It’s certainly higher than that for software, where if you exclude Amazon because they deal with physical stuff, the remaining FANMGs average about $2M in revenue per employee, and ~$300k in comp across all employees.

Because we can just consider their GDP value directly due to the nature of white collar work, now we can directly map the number of white collar jobs to their total GDP impact.

Because the real question has always been - how can UBI ever work, given it would require both A) confiscatory taxes on the people still working, and B) result in really low incomes for everyone.

My answer is basically “because cheaper white collar jobs will lead to massive economic growth.”

The UBI proposal landscape

Scott himself surveys ~9 UBI plans, sees mostly nonsense, but also sees a road to a ~$12k per person UBI.

And he likens that level of UBI to college - back in the halcyon days he was writing that post, $12k was the average cost of room and board at college.

Which just as an aside: HAH! This reminds me of all the Boomers and Silent generation people that were able to buy their Tier 1 metro single family homes for two pennywhistles and a moon pie, who are totally baffled on why no young people buy houses now. Old people voice: “Did you try offering them a moon pie along with your bid of two pennywhistles? That worked great for me and my friends! What do you mean, ‘100k over list price??’”

But tangents aside, this is looking at it the wrong way! I think we can offer at LEAST $40k - $90k in real 2024 dollars UBI to everyone with some realistic assumptions.

All the other UBI plans weren’t considering realistic economic growth

Let’s make some baseline assumptions:

• The jobs that can be automated are all white collar

• I’m gonna assume the AI future scenario value to GDP of white collar jobs is salary + 30% (so haircutting that 40% estimate by a third)

• I’m going to assume 50% of white collar jobs counterfeited, with 50% still trucking along

• Jevon’s paradox will result in more white collar jobs existing than before, because when something gets cheaper, you use more of it, especially if it directly provides economic value to your company!

• I’m going to put things in terms of “real 2024 dollars,” because that way it will basically abstract away the division between consumer surplus / cheaper goods and “taxable income” on the companies’ part, because I’m largely interested in “what is the total surplus value created,” then dividing that among Americans without getting hung up on details

• I’m only going to look at endstates here, to abstract away all the potentially complicated interaction effects and halfway steps that are likely to happen on the way to that end state

So say we reach a world where you can automate half the white collar jobs for ~33% of their current salaries.

This is A) a huge advantage to the companies with white collar jobs, and B) still ends with up to ~$13T in net income annually captured by the frontier AI companies counterfeiting jobs, so everyone is happy.

But if a given white collar job costs ~33% as much, companies are going to create more of them! Even with an extremely conservative Jevon’s Paradox multiplier of only 2, this leads to an additional ~$40T of GDP, between the value created by those additional jobs, and the value being captured by the AI companies.

That’s more than doubling the economy! Moreover, most of that doubling is “value created,” ie would go to net in the absence of competition, because the cost of the AI white collar workers is so much lower.

Now remember, I’m baselining in “real 2024 dollars” to abstract away the “taxable corporate income vs cheaper goods” tradeoff and make the math simpler.

Even with current corporate tax rates, that lets you hit a ~$35k per person UBI, with zero filtering.

If you bump us back up to the ~35% corporate tax rates we had before the 2018 Tax Cuts and Jobs Act, you’re at $45k.

That’s for *everyone,* including people who still have jobs and people who have capital.

Capital? Oh right

A ~2x multiplier on the economy is a ~2x+ multiplier on capital’s value, too. Arguably, it would be even higher, because capital can now directly turn into the highest value labor, but for the sake of simplicity, we’ll assume it just expands at the current capital-to-GDP ratios. Right now capital owners average ~$88k in passive income between dividends, owner disbursements, safe withdrawal rates, and other disbursements.

Assuming no change in capital ownership, those same capital owners will average $214k in our glorious AI-expanded-economy future.

So what does that mean? We don’t need to give those guys any UBI!

That brings the UBI available up to non-capital owners to $50k - $66k in real 2024 dollars, at current / 2018 corporate tax rates, respectively.

And would we keep current / 2018 corporate tax rates if a bunch of white collar jobs have been counterfeited? Proooobbably not?

If corporate tax rates go up to 50%, the UBI for non-capital owners goes up to $95k per person in real 2024 dollars.

That’s pretty comfortable!

Implications

• A lot of white collar job holders are already capital owners, so they’re actually the best segment to counterfeit jobs in, because they’ll get to take part in the outsized capital growth value as our entire economy doubles, and still end up about twice as affluent on average than the UBI’d rest of America, even with zero UBI

• If we UBI everyone, including the 60% of non-white collar jobs, it’s going to drive wages up for all the “still real world” jobs, and incentivize some formerly white collar people to do those jobs too, and will likely reach an equilibrium where they’re probably compensated about 1.5x higher, with better working conditions

• This means UBI’d people can live anywhere - it’s like the COVID dynamics, but much more, because you can actually count on the income to be there in the future and never have to go “back to office” to support decaying commercial real estate values

• Overseas arbitrage - all Americans are rich now in real world terms and can travel a lot and / or live overseas like kings in a large number of countries

• The numbers I outlined above don’t even include ANY income from the same AI companies counterfeiting white collar jobs all over the world, which would conservatively raise the US GDP by another ~$17T and bring the non-capital holding UBI to $60k - $120k at current / 50% rates.

• Citizenship is now literally about free money for life, so immigration necessarily gets a lot more restrictive, including marriage based immigration because…

• Median dudes only get married at 20-50% rates these days, so suddenly having ~$90k in passive income they can splash around anywhere in the world will get a lot of them overseas wives. But if we let that ~60% of the population add other UBI-receivers to our economy, that would obviously reduce the total UBI available by ~60%, so we’re probably going to get a lot more strict there, too

• But that also means we probably permanently solve the USA fertility crisis, alone among all countries in the world!

Mea culpa, call for challenge / correction

This is literally just some napkin math at the highest levels of abstraction that I put together in an afternoon - I could have missed something really important, or gotten something wrong.

So I urge everyone out there to look at my assumptions and math and double check everything - the Google sheet is here.

Objections after looking at the sheet

• You can’t just abstract away “taxable corporate nets vs cheaper consumer goods” because all the *important* goods won’t get cheaper! Real estate in Tier 1 metros / good school districts? Art? Positional goods?

  • But yeah, I think you basically can. For people that opt into UBI, they’re largely opting out of those status games - they can live anywhere, including cheap places in the US or overseas. People don’t actually need art and positional goods. They also don’t NEED to live in US Tier 1 metros for jobs, and a lot of them won’t. Finally, only the PMC cares about school quality (just look at “percent of parents who are involved / volunteer for PTA”).

• You seem to be dropping “value of a given white collar job” straight to taxable net in the AI future - wtf? That’s nowhere near what companies get today!

  • Yeah, look at it this way - let’s say current white collar job value created = salary + 30% , or ~$260k. But in the future, that will be ~$60k in costs per job instead of $200k in salary! The job *still drives* $260k in value creation, but now costs $60k, for $200k dropping straight to net, or 100% of current white collar salary values dropping to net.

  • Sure, that is making a lot of assumptions in terms of net income, competition, and how much goes into consumer surplus - but for a high level treatment, it’s close enough given the first assumption - “real 2024 dollars”. I’m mostly after “what are the high level amounts of surplus in that AI future, and what does it roughly equate to in 2024 buying power,” and I think this is good enough for that.

• Your AI companies only have 5% COGS?? Are you high!?

  • How much are they building right now? About $500B of stuff over the next 3 years? Let’s say that 10xes in the future to get to “counterfeiting white collar jobs” capability levels, so it costs $5T. Remember, once you have a smarter model at all, you can carve inference down to a 10x smaller footprint, while keeping most of the smarts. In this future, they are spending $2.25T on capex + inference every YEAR with that 5%! That’s nearly $7T over 3 years, handily outstripping the $5T we just estimated we’d need, and if they grew to the other 50% of jobs, they’re getting $14T to spend on capex / inference. Counterfeiting a bunch of jobs is so hugely profitable that it makes sense, in other words.

• Your tax rates are off! Sure the face values are ~21% (plus 5% state) today, but companies actually PAY some much smaller fraction of that via accounting tricks and tax minimization!

  • Yeah, and? Obviously in a future where they’re suddenly printing money because they laid off everyone, that’s the first thing that’s going to change, and we’ll be actually getting face rates, either due to blocking most tax minimization (ie charging VAT, which captures flow, or compute, or “AI worker seats,” or something else that’s harder to hide) or scaling up face rates so they’re actually paying taxes in amounts closer to the stated face rates.

• Lol, you think the way things are going to go is currently high status and politically relevant / connected white collar job people and capital owners are just going to shut up and go quietly into that good night while letting all the proles have $90k UBI’s without getting a cut?

  • Sure, this is probably the most unreasonable assumption. But that’s why I did a cut by “everybody” for each of these. Also, we are a democracy, sirrah! And the proles vastly outnumber both the white collar people and capital owners, who are largely an entirely overlapping venn diagram! So uh, good luck with actually keeping a lot of that surplus, especially when your capital disbursements also doubled?

Summa

The “AI counterfeiting all jobs” future is gonna be great!

If you assumed we were collectively reasonable and would act like adults societally, what better use could there be than spending our resources to create machines smart enough to automate ~40% of jobs away?

Arguably, we’re under-spending right now, and that’s without anyone actually believing we’re going to double GDP yet!

Actually, being at a point where we can only counterfeit white collar jobs at first is a huge free lunch, for several reasons! Physical jobs has much lower margins for companies, contribute notably less to GDP, and would require substantially more hardware / capital expenditures to counterfeit. This is entirely ignoring the fact that robotics is much harder than software for numerous other reasons.

Having ended at a place where we can counterfeit only the highest margin jobs, that generate huge amounts of value both for GDP and the companies with those jobs, and in which the current job-holders are all predominantly capital owners who will largely be covered by company value growth in a much-higher GDP future? It’s so much easier on every front!

Remember how everyone fantasized about 10 hour work weeks? How everyone talks about “bullshit jobs” and how most people hate their bosses and spending 10+ hours a day on stuff they don’t want to do, including commuting? All that, gone! People able to spend their days the way they want to! More time for kids, family, hobbies, those projects you always meant to get around to!

And the jobs left will be “real” jobs with tangible impacts and direct feedback when it comes to improving the world, which is a huge psychological bonus for many people.

I actually expect a pretty serious shift in workers, as the non-white-collar people who hate their jobs opt in to UBI, and the “need to work to feel relevant” former white collar workers flood into the remaining jobs, because they enjoy working and having a positive impact in the world. Collectively, that’s going to lead to a HUGE bump in execution quality, at the margins. It’s also going to lead to salaries and working conditions improving for all jobs, as employers have to compete harder for people with more optionality. Sounds like a win for everyone!

“Ah, but we’re *definitely* not reasonable collectively...just look at all the idiots around you! Then look at who they’re putting into office!”

But honestly, I think at the forty thousand foot view, we are. If we counterfeit even 10% of white collar jobs, people will be up in arms, and politicians will do UBI or something like it because for better or worse, we’re a democracy, and all the frontier AI companies that will be capable of counterfeiting jobs are US companies, so those worldwide revenues are harvestable, too.

It’s halfway to fully automated luxury gay space communism!

Half-automated solarpunk lesbian earth socialism? Close enough!

So with material needs taken care of, many will be freed of the tyranny of pointless jobs, and can spend time on all the stuff they’d rather be doing than jobs.

Will 80%+ of people use that extra time to eat junk food while staring at screens in particularly mindless and self destructive ways,¹ in ever increasing quantities? And how! And I expect that to only get worse as we go forward!

But you know, that’s their choice, who are we to argue with revealed preferences and free will? If you don’t want people to do dumb things, you’re going to have to send them to camps, and the solution is worse than the problem.

And it genuinely frees up non-screen-starers to do stuff you might consider more virtuous like raising kids, hobbies, etc.

I personally look forward to a future America where a lot more people travel, have interesting hobbies, have and raise kids, and don’t go to jobs that they hate!

1. With screen time currently at 10-11 hours a day in recreational time for every generation, with the only difference being the mix / size of the screens, with older having more tv and younger more phones: