Clojurians Log v2

This part is wild… do we think the author is making this as a universal truth? I've never thought of myself in this way.. “since human wants are unlimited, abundance is unreachable”

It just feels so pessimistic.. I want for us to reach the Gene Roddenberry ideal 😢

I don't know if we should say human wants, I think many people can find content at some level. Look at all the FIRE people wanting some baseline and then happily not wanting more then that. But I do think humans have a drive to push limits.

The social contract has been on shaky ground for a while. Modern economics has a belief that you don't need trust if you have sufficient trade, when it is trust that underpins our ability to trade. With respect to the unlimited wants quote that is a response to socialism, and as someone who is pretty left myself I do think it's a legitimate critique of socialist policy. People in general are happy to pay for what they want, not so happy to pay for what they need.

I like the article but one thing it misses is I still don't see a moat for these models. If you own the compute sure, but everyone else is just trying to make hay while the sun shines. Even the tech-elites.

Ya, fair enough, most people want more than we produce at this time

What do you mean about the moat? People pay them to use the models. Is that not the moat?

If someone comes along with a better model all the money jumps ship. And, as almost all the training data is already openly accessible, the barriers to building a better model are really not that high (compared to the current valuations).

I feel in practice that seems wrong. Europe for example can't even find the cash to train a model. Only the two most capital rich countries have been able to even enter the stage to compete, US and China. Chinese models have never surpassed US frontier, likely because they are derived from them.

I find this incredulous and possibly only conceived by someone who has not had to suffer the consequences of being or growing up underpriveleged. The few “possible futures conceived” are 1. elite entrenchment 2. class dissolution, which is just another way of saying elite entrenchment since they also say “The homoploutic household in this case reverts to a purely capitalist household that lives entirely off the earnings from its investment portfolio; in other words, it reverts to Mexico’s classical pattern where the rich don’t work.” What sort of fantasy have they lived in their whole lives to believe that the world was anywhere close to a meritocracy? I am reminded by Yitang Zhang, who survived off working in Subways because he studied his PhD under someone who was working on now-disproven conjecture, and his case gets polished into a “golden case for meritocracy” https://x.com/AlexKontorovich/status/2111913572972351946 when it is clearly a case of survivorship bias.

@danieltanfh95 Doesn't the article address your exact point? > The homoploutic class has a sociological face and an ideological one as well. In “The Meritocracy Trap(2019),” the law professorDaniel Markovits described the class’s machinery: The elite is hard-working and credentialed and is therefore convinced that its position is earned — even as the value of parental investment separating an elite child as compared to a middle-class child by the end of graduate school is equivalent to a traditional inheritance in the neighborhood of $10 million per child. He calls merit “an ideological conceit, constructed to launder a fundamentally unjust allocation of advantage.” One can put it as a modification of Marx’s M–C–M’ (money-commodity-money) circuit: here the circuit becomes M–E–M’, where E is the education of one’s children. But if E gets out of reach for all but the children of homoplouts, it means that class mobility has effectively closed.

> moat ... [around the businesses of models] > If someone comes along with a better model all the money jumps ship I agree in principle. However, the idea of a moat only temporarily protecting (or not protecting) a company from competitors is rooted in Adam Smith-style capitalism where there are many buyers and many sellers. For frontier models, that's not quite what we have. Instead we have something more like an oligopoly where there are a few firms selling, with high barriers to entry. The companies can act almost like a monopoly and give poor service with high prices. This condition can last for a very long time. Aside: Even an oligopoly/monopoly moat may not last permanently. Witness Microsoft on PCs in 2000, IBM in the 1970s, ATT, etc... In the latter two, the US government investigated or broke them up. With Microsoft, Apple invented a rival parallel technology (phones) while many new customers paid a price to use to a rival OS for servers - Linux. The question is - is it different this time? If it is, and the moats hold long enough for the companies to gain/retain political power, something like the thought experiment in the article may come to pass. It's still a prediction though, and as they say, "Predictions are hard. Especially about the future."

Wall of text - TL;DR: the frontier model companies might well be an oligopoly, that might be bad for business and our well-being, but who knows if it could lead to the scenario in the article.