There’s a question related to what Gil asks that I have wanted to look at for a while.
I would put her suggestion in the category of “better democracies provide their citizens with better information for decision-making”, which I also wanted to look into. It also brings to the forefront another dimension of what one wants a democracy to do, if one wants to argue in favor of democracy as a framework. But the thing for this post was a bit different (though related). Redo the basis for pricing and economic decision theory based on real-goods input-output, and whole lifecycle material and energy constraints. So the background was: we know that marginalist pricing in economics (what we all use right now) commits two big distortions: A: it doesn’t clear markets very well, and the problem is not just the incompleteness but the skew. Rich people can bid up prices for luxury consumption to a level where many more ordinary or poor people can’t afford necessities or near-necessities for getting-by (let me not use the word “survival”; it’s too melodramatic). What form the skew takes and how bad it is depends on the way marginalist pricing interacts with wealth inequality in the population who are using the economy. B: Not pricing in aspects of one thing can distort the prices of everything. The not-pricing part is what economists call “externalities”; what one thinks of this term, and its suggestion of inevitability, can be discussed as a separate point. But the main use case here is energy sources. The main things not-priced in are climate and other ecological impacts now, and scarcity consequences later. Oil is the big one, but coal and natural gas are in there too; and equally well we could look at ores that are needed for electrification under current technologies, and much else. The goal in the re-do, then, is to make real input-output models of the economy that look more like chemical-reaction schemata, in the sense that we know a production event can’t happen unless all its required inputs are available, and all the unavoidable outputs have somewhere to go. von Neumann started this in 1937 and a few years after, and people worked on it for a while. He did it the way economists do (and also for the sake of ensuring a problem he could solve), with the requirement of all-positive prices and the assumption of “free disposal”, meaning if there is a pollutant you produce but don’t want, you just walk away from it and it doesn’t cost [you] anything. Within the last year or two, we redid the same framework, but allowing prices to become negative (you have to pay somebody to take things that neither of you want), and no free disposal (everything has to go concretely somewhere, and the economy should do a thorough accounting of those somewheres). It’s a trivial construction, of course (a human could do it), but I don’t think anybody has bothered to pursue it before. von Neumann’s point is that you can put a decision theory on this system and derive prices from that, the same way as you can for marginalist pricing. But the underlying market-clearing model is very different. Same goes for the version with both-sign prices and no free disposal. In the era of “smart” contingent contracts that are popular in the blockchain and proof-of-stake world, there is lots of other already-in-use method that could be built with the information that this kind of data-foundation supplies, beyond just dollar-prices. Then the whole-lifecycle thing, which can be done independently of, or as part of, a von Neumann representation, is to take account of inescapable baseline costs in such quantities as free energy, so that grades of ores have physical relations to the energy sources used to extract them and the alternatives such as more or less complicated and costly recycling. This too has some precedent, but by hand it is too costly and labor-consuming, so it was never going to scale that way. Both of these remind me of things like insurance, where decreased costs of integrative computation and communication can take something that we know perfectly-well how to do, and transition it from being something nobody wants to pursue, to something that there is a valid model for somebody to develop. Used to be that if you wanted to ensure a ship, you had to go to Lloyds of London and get a custom contract written, with big cost, delay, etc. Now insurance for very big things is a commonplace with standard models. It was just increased scale and decreased cost for finance, accounting, computation, and communication, that made that tip possible. The law around it could then follow. The combination of comprehensive and also fine-grained data collection that the big-compute companies already do, with the integrative capabilities and custom-rendering that the LLMs do, and the ability to generate large consistent models (as with the Lean applications to big complicated theorems and conjectures) make it seem to me like all the capabilities are already in place to try to prototype something like this. Incentives, not so much, but that identifies a concrete task. Eric > On Sep 5, 2026, at 8:58, Gillian Densmore <[email protected]> wrote: > > Ai Comp Sci Question: How come there is a need for arms race to get insanely > dense model..so as I am to understand it the big guys have massive amounts of > parameters (30bn from Ali labsqwen) just as layman it seems like we can make > the Neral Nets, unets, and GANS, more...compressed. Can anyone kindly explain > to me what the heck we're doing here, where models can speek idiot human like > me. (suposed to speek ) where the need for rediculous computes goes up, but > how good the hardware (GPUs and memory etc) have to go up x amount more to > run them. Where is the part where the software is made much more...correct > and elegant. > > Or is this just the nature of an exloding new field where we don't know what > the hell we're doing half the time. > > > Speculative question: Where's the generative ai that can do more than make > pervy pictures and cringe adds for coca cola I was just thinking about this: > it'd be so...handy: if generative ai could be used to create something where > its much much lighter for one thing. And also: It'd be so handy for maybe > videos showing over time what a importat meadow, or some other...analog thing > for want of a better phrase fromthe real world fead real GIS data and then > extrapolate and create videos, semi-interactive reality based simulations for > real big picture: Litium, and silicon aren't infinite (on 5 year scale) at > least so I understand: showing a very acruate simulation of: gen what comes > after Z? oO well them having to using pretty old tech: day to day life (what > ever that looks like) so...the normies with a very americana tight read get > shown in a extreme visceral way: no actuall not just dry but important nerdy > pre-rendered simulations. But: you could actually ask a ---hyper real person > who has simulated 5 year world: is it like? etc and get...dead on right > answers because very sophisticated agents. that's the world I'd love to have > for future gens...if it'd be a improvement. > What do folks think? > .- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... > --- -- . / .- .-. . / ..- ... . ..-. ..- .-.. > FRIAM Applied Complexity Group listserv > Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom > https://linkprotect.cudasvc.com/url?a=https%3a%2f%2fbit.ly%2fvirtualfriam&c=E,1,_m5pbtswSR-HXlli9a7K9EQSnH-OS3Pk81toCrNrcyinfNfl73nCo8Rt8yv3mcGs_xh-_a2K0CyNAx_W4dMOhozl4KiTNc5XATU5Bx8olCz6v3258mC4aKnrYcQm&typo=1 > to (un)subscribe > https://linkprotect.cudasvc.com/url?a=http%3a%2f%2fredfish.com%2fmailman%2flistinfo%2ffriam_redfish.com&c=E,1,-7YL5BNZS8S7sRlPhTYbViQW--uXP1AaGrwx0B6FGRKl-IJAuqsXRXk93aXT-dk0q8EK50wiZ20z9MLIOpGytH_3LHCnDd6oPnH-Jr5oTrShkfXE&typo=1 > FRIAM-COMIC > https://linkprotect.cudasvc.com/url?a=http%3a%2f%2ffriam-comic.blogspot.com%2f&c=E,1,wH9obAsHhOP3_9sIEsbAZO75ELskNLCCWjLKpCwBIb_1v_oXP0NvIvVnsrSNc-nkpRlOecffqodLvkfk7iL1mRsHeWnbAJSIgjJFG4iZw8jof-X6I2hCV80,&typo=1 > archives: 5/2017 thru present > https://linkprotect.cudasvc.com/url?a=https%3a%2f%2fredfish.com%2fpipermail%2ffriam_redfish.com%2f&c=E,1,us_qzYrFZfqD6UaObMAlmCt73BOcxD3Alcjq6nEvrzSCCTtrtwbPV9wwD6SHhZJSTmIsWvPTHptDMNfixQ52lmyk6SK0MCbMVWJ8KFKDnw,,&typo=1 > 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/ .- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-.. FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com FRIAM-COMIC http://friam-comic.blogspot.com/ archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
