RE: https://neuromatch.social/@jonny/117237800074111038
A recurring thought through all this insanity is "what if we just spent a fraction of that money on [the obvious way] instead" and it always seems like the economics would be way less shitty, so
![Edit of XKCD #1404, two people walking and discussing over three panels:
1) A: Hear about that "impossible" [LLM feat]?
2) B: Yeah. Let me get this straight—they pumped [$2 trillion] into [the pursuit of knowledge]… and it only twitched a *little*? A: Yeah.
3) B: *That's* surprising. If you pumped [$2 trillion] into *me*, I'd twitch a *lot*. A: But you're not pushing on the [bounds of human knowledge]. A: I *might* be. I do a lot of things. B: I guess we can't be sure.](https://media.hachyderm.io/media_attachments/files/117/239/280/970/627/851/original/7ee5454f786debff.png)
anyway this is probably worth a complete stall in the academic pipeline across disciplines, right?
that's probably fine.
just imagine what a single lump of $100M would do to fuzzing research
and then reflect on that being a literal rounding error in a speculative frenzy measured $100B at a time
and you can just keep going like this
$100M is *six years* of NSF funding for *six* materials science research centers
https://www.nsf.gov/news/building-future-atom-atom-nsf-deploys-108m-materials-science
but yea I'm sure google doing a fucking denial-of-service attack on materials science is good too
the entire world's radioisotope market is worth $10B, or about 10% of a single 100B deal, of which we have now seen so many that the number barely even phases people
"if only there were some other way" says only* economy that fails to fund obvious public multipliers
*if you reply to this joke to explain that there is more than one, you will be ZOTted
the cost of constructing and operating the LHC for years is less than 5B CHF
anyway, yea, markets are tools for efficiently allocating resources, and *this* kind of market is fundamentally better at that than every other kind
we have spent O(Afghanistan War) money on this fucking shit
in a quarter of the time
and what we have to show for it is mostly uhhhhh child pornography and more effective vandalism
@SnoopJ Don't forget the environmental laws violations!
@FunkyBob the amount of money spent on genAI (and the war in Afghanistan) is basically the same as complete compliance in the Paris Accords would have cost
@SnoopJ ::cries::
@SnoopJ "the truth is, these are not very bright guys, and things got out of hand"
@glyph I've never seen it, but *gestures* I sure do know it
Sep 9, 2026, 6:49 AManyway, yea, markets are tools for efficiently allocating resources
Markets are tools for efficiently allocating resources, but with some very important caveats:
I wish more free-market fanboys in politics understood how markets actually work. They can be amazing tools but few machines work well if you remove their regulators.
@david_chisnall @SnoopJ Isn't there a bunch of research out there that shows constraints make you more creative?
Seems to work at the macro level, too.
Intuitively, this is why the unlimited freedom espoused by libertarians is very unappealing.
@GeoffWozniak @david_chisnall @SnoopJ also in a global world you'll need even more regulation.
In a more localized world, autocrats sort of understood guillotines and pitchforks could make a mob of peasants quite a danger to look out for.
Hence private islands and bunkers... We need a substitute on global level because there is no local binding (and "safeguards") any more...
@david_chisnall @SnoopJ I'm reminded of a point someone (was it @HalvarFlake ?) made about specialized vs. generalized hardware, where specialized hardware will always be more efficient to do the task, but the added overhead of getting there vs. just using whatever is general purpose and on the shelf just doesn't make it worth it.
Having a giant "do the thing inefficiently with zero preparatory overhead" can be useful if whatever it is that's required for preparation is scarce. And apparently it is, or we'd be doing the other thing
Sep 9, 2026, 12:25 PM@david_chisnall @SnoopJ Whet I was in highschool they taught us that the benefits of the free market relied on having "fully informed and rational consumers" -- otherwise they can't/won't spend their money on the best solution.
Ever since I've wondered if such a thing has ever existed.
I'm pretty sure that shit is a myth.
Sep 9, 2026, 1:11 PMIt's not quite that binary. Markets decay fairly gradually.
Markets are a multi-round game. In each round, participants allocate capital. In the subsequent round, the people who allocated it more efficiently are given more capital to allocate (note all of the caveats from the previous post in this).
If there's a small amount of information asymmetry, but many participants, the market itself acts as an information-discovery channel. Imagine a simple commodities market where you're buying bags of grain but can't see how full they are before you buy. If the variation is only 5%, no one will lose particularly badly but the people who picked the sellers who sold more-full bags will get more capital next round because they're able to do more of whatever they do with grain (make flour for bread, for example). And that provides feedback in the other direction: other people in the next round will favour those suppliers because the people who used them did better last time. Over multiple rounds, the optimisation works and the information about which sellers give short measures spreads and they have to lower their prices to be able to sell anything.
If the information asymmetry is more pronounced, this doesn't work as well. If the amount of grain you get is between 0 and double what you expect, then some buyers will make no money and possibly drop out of the market altogether (can't sell bread, no money, no business). Others, through luck rather than judgement, will double their investment and will now have a lot more capital to invest next round.
The market in this example failed badly in the first round (capital wasn't effectively distributed) but now can't easily correct either. Some people who are no better at allocating capital than others now have a lot more capital. They may decide to spread it between two sellers. On average, they'll break even, but any round where they make more increases their share of the total resources.
There's a nice game that models this. Imagine you a load of participants who each start with $10. Every round, they're betting on the outcome of a coin toss. Each participant can bet 10% of their capital. At the end of the first round, you'll have a load of people with $9 and a load of people with $11. But the second round, the people with $9 will be allowed to bet 90¢, the people with $11 can bet $1.1. If you lose the first round and then win the second, you have $9.90. The same applies if you won in the other order if you bet the maximum amount. But consider someone deciding in their second round to bet only $1. They now have $12 if they win, $10 if they lose. They can structure their outcomes so that, whatever happens in the second round, they're better off than someone who lost in the first round.
Play this game for a few dozen rounds and you'll see wealth concentration.
And this is why markets work best when you have very high upper tax benefits and antitrust regulations.
@david_chisnall granted
@david_chisnall @SnoopJ "properly regulated" is doing quite a lot of heavy lifting here. Creating a framework for which markets operate is critical. Unfortunately regulatory capture feeds corporate entities and ends up having a worse effect: the destruction of the petit bourgeois
Can I add
- There are many areas where reliability is much, much more important than efficiency.
Sep 9, 2026, 6:43 PM@david_chisnall @SnoopJ Be also aware that efficiency is often not that what we want. All concepts of well-being and luxury contain the idea of not needing to do things in the most efficient way. Efficiency is good when it concerns things (which are produced in the most efficient way, or with minimal resource use), but it is bad if it is about people. The people whose work is most efficiently organised tend to have the least amount of freedom and the lowest pay. (The low pay is also an effect of efficiency.)
@SnoopJ It's hard to express just how maddening this is to explain to people who fall over themselves to fawn over the latest, oh, I don't know, math result.
@SnoopJ 25+ years ago as an undergrad I realized that if you had the compute power you could do quite a bit.
My mistake, apparently, for thinking of being efficient in my resource usage.
@SnoopJ Reminds me of the time “We got an AI to write a compiler in a few weeks” and every unemployed compiler engineer was like “If you gave me $50,000 I could also write a shitty compiler in a few weeks”