In the long run, absent intervention, virtually all income flows to the owners of compute.

We need more than UBI. AGI is the culmination of all human activity up to that point and all humanity deserves ownership of it. It should not belong solely to those who put the cherry on top with the rest of us at their mercy. They don't deserve to control the humanity's destiny. AGI, at some point, has to be made into ... I don't know. Not nationalized - something more. A force of pure good for all humans unaffiliated with any corporation or state.

I wonder if "AGI" is going to end up like quantum computing. With expectations and predictions so unmoored from reality, that everyone just sort of pretends it's a thing without every actually genuinely thinking about what's going on.

Edit: words

I imagine we know we have reached AGI when a technocrat stops sharing their AI. It goes from being something they can sell to something that they don’t want to share. Instead they ask it how they can dominate the world and be the first trillionaire and how they can stop anyone else acquiring an AGI etc.

This even works at a smaller not so general level: imagine that one of today’s popular code models improved to the point it is better (narrowly at programming) than a human. Suddenly the owner shouldn’t sell it to everyone: instead they should pivot and make software that outcompetes anything a human can make. So it doesn’t just replace humans making software but also replaces the programs that people made…

For more economics of AGI, see also:

This tweet recapping this paper https://x.com/lugaricano/status/1969159707693891972

This tweet with recaps of various papers presented at "The Economics of Transformative AI" by NBER in Palo Alto a few weeks ago https://x.com/lugaricano/status/1968704695381156142

I remain amused at classic Sci-Fi being reimagined as scientific papers. Authors have warned and celebrated the potentials of AGI on human society for decades if not millennia (if we include deities, gods, and other super-human machinations): we know full well that a post-AGI economy will be hell or paradise solely depending on the underlying civilizational structures when it is birthed into existence (hell if Capital reigns supreme and Nations remain in conflict, paradise if we get our species' collective shit together and consciously shift to a planetary societal mindset for economics and governance).

As for what humans should do when labor can be accomplished via AI? Man, if you're asking yourself that question in 2025, you're kinda late to the party in a lot of ways - and desperately need to read more books. AGI may eliminate the need for human labor in Capital terms, but it does not suddenly eradicate all value of human labor - only the value Capital ascribes to it. It's why, in a post-AGI fictional setting, you see so many more adventurers, artisans, explorers, researchers, engineers, teachers, and other roles often undervalued by Capital but highly valued by humans.

I imagine most humans will simply turn their favorite hobby into their new "profession", as a means to meet other humans and continue growing social bonds. Maybe that means painting full-time, or striking out on a photography journey. Maybe it's opening a retro game "store" to connect with like-minded enthusiasts. Maybe it's running your own museum for local artists and creators, sharing your tastes with other visitors.

A lot of "valueless" work under Capitalism suddenly becomes highly prized in a post-AGI, post-scarcity society. Most humans will figure that out to varying degrees, and the rest will be able to benefit from a wider array of fee-less services. Provided, of course, we begin changing society today to enable that future tomorrow.

Assuming AGI or something close to it hits, it'll likely be a nightmare for the average person.

The model in the paper talks about freeing the labor bottlenecks, which will be great for the capitalists who own the compute. Assuming we don't have a social safety net in place, that will ultimately mean that a lot of stuff is cheaper for the wealthy with nobody else being able to afford it.

What would a middle class job be? Almost all knowledge work would be obliterated. Engineering, science, maybe even art. All evaporated.

The paper suggests that people will shift into markets like manual labor. But what do we do when those jobs have no real bottleneck as the paper describes? There's only so many people needed to for care work or picking berries. And right now, it seems we have enough as the current salaries are pitiful in all the sectors the article mentions. What pressure would actually make wages increase? Sure not the fact that those are the only jobs that exist for normal people that weren't born to a family that owns a datacenter.

And it isn't like datacenter jobs are going to replace the army of jobs AGI would displace. You need so few people to operate a compute center.

That's why it'll likely be hell for most people. If you don't actually own resources, you'll be left out in the cold. Even if you do have pretty good resources, you'll be in a world with AGI set to perfectly extract every single bit of resource from you imaginable in ways we currently don't imagine. For example, knowing everything about you and knowing that you are willing to spend $11 for a widget while someone else is more willing to spend $10.

This will ultimately force the question of "what do we do with the unemployed" and I worry the answer is already "well, they should have worked harder. Sucks to suck".

okay so what exactly is this "accessory" work that will be left to the humans lol

I think if AGI ever becomes real, we will all end up as Indiots of the Twenty-Fourth Travel of Ijon Tichy.

AGI is the new Marxism—a utopian dream unmoored from reality, which does not account for the nature of people, economics, states, or even nature itself. A fantasy society that will never come to pass but, if attempted by fanatics, will probably do great damage.

Yet another paper using the term “AGI” without rigorously defining it.

There is a oblique definition embedded in the context that AGI is literally any labor transfer technology, but again without a distinction between measurement (which is possible) then these conversations and concepts are going to stay unmoored from reality

I made this google form link and so far the answers are all over the place including

- “you can’t” - “It Can ask: I, why?” (Odd answer)

and my personal favorite: “It can do everything better than humans with no prior information” (impossible according to no free lunch theorem):

https://howdoyoudefineagi.com/

As labor ceases to be the primary driver of value, economic policy must confront a basic question: how can we share the income generated by compute? In a world where AGI performs all bottleneck work, income generated by production flows to those who own or control computational resources. One approach is to redistribute the gains from compute through universal dividends. An alternative is to reimagine compute as a public or semi-public resource, akin to land or natural capital, with returns broadly shared.

It might be clearer to say "how can we share goods and services provided by intelligent machines?"; the paper seems overly focused on compute as "the" bottleneck when natural resources are still needed. (Even AGI can't magic up new copper atoms from scratch, though it can exploit low grade ores that were previously economically useless.)

I think that referring to income instead of goods and services predisposes people to think of this in a currency-centered way, when currency-denominated market transactions may become much less important in a world of intelligent machines. If (v) the share of labor income in GDP converges to zero is actually true, then machines can do everything, including copying other machines. Co-ops, municipalities, provinces, and states can vertically integrate to provide goods and services outside the market if intelligent machines are actually doing the work. Compare the old "buy an OS, buy a database, buy a compiler" approach that one had to take circa 1990 with the "copy a free Linux distribution" approach circa 2000.

If wage income is obsolete, so is intellectual property rent. One common fear in these discussions is that billionaires with robots will let the masses die of starvation when their labor is no longer needed, but the billionaires who own IP don't have much leverage either.

For one, most of the world's population/governments aren't beholden to IP billionaires now. It might seem like that to those of us who live in the Anglosphere (a plurality of HN users), but globally speaking most billionaires didn't make their money from IP. They'll back national movements to ignore copyright and patents to Just Copy The Foreign Robots when the time comes. (South Africans who were wealthy from mining didn't have an incentive to side with foreigners who were wealthy from pharma patents back when South Africa was ignoring IP to fight HIV/AIDS, as a parallel example). For another, billionaires aren't even in charge everywhere. See the example of China with Jack Ma, or the fabulously wealthy oligarchs that have been brutally demoted in Putin's Russia. If a leader can accumulate power and rally popular support by giving free robot goods and services to the people, they will; the IP billionaires don't have anything to trump that offer.

For these reasons I don't worry about mass immiseration/starvation if smart robots actually take over all productive work. I'm sure there will be struggles, but I don't think that IP owners can win the fight any more than the MPAA actually ever eliminated movie piracy.

The thing that worries me more is mass empowerment of even the world's most inflexible and violent personalities. Nuclear weapons, ballistic missiles, and nerve gas are now old technologies. The main thing that prevents every angry separatist movement or cult from becoming armed like North Korea or Aum Shinrikyo is lack of material and technical resources. But in a world of smart robots, all you need to get your own enriched uranium and ballistic missiles is those smart robots, a region of several square kilometers that no outside force is policing, and a few years to bootstrap the precursor technologies that don't already have blueprints in public databases.

Mass material abundance probably means a decrease in "ordinary" crime driven by the stress of material deprivation but an increase in tail risks from unhinged individuals. The sort of person who kills 5 strangers with a gun in the name of their ideology could become a person who shoots down an airliner, killing 200, in the name of the same ideology.