If, as thinkers like Hayek suggest, the economy can be understood as a mechanism for dealing with scarcity of information, what new possibilities open up - or don’t - by virtue of the explosion of information we produce, analyze, correlate, &c.?
cybersyn on your phone
There’s some interesting discussion happening at places like CASPER forum and Thomas Hardin’s blog on the mathematics of the planning process. The bottom line, however, is that
planning with a very large number of goods is feasible on modern supercomputers
Moore’s law means that what modern supercomputers can do, phones will be able to do within a relatively short amount of time
What this means is that the average citizen should be able to engage with debates over an economic plan to a much greater degree than we often suppose they might. A Cockshottian “Scottish model” could be much more democratic than otherwise supposed. AI that describes mathematical results in plain English, Spanish, or Chinese and then back again could allow a greater degree of non-expert participation in the process. Further, we can expand the use cases of mass access to planning tools to include consultative as well as prescriptive uses of such models.
political and economic implications of pocket supercomputers
Even if (at the limit) there is a single central planning body with economic authority over the whole planet, being able to run planning simulations on your phone has a few important implications:
You can verify that the math checks out on central plans. Unless there’s a heavy information-destroying censorship regime, which would be a bad outcome for unrelated reasons anyway, it should be pretty difficult to lie in these plans. Public verifiability should also reduce incidental corruption, and so on. (Corruption might still exist in spheres where we want to deliberately add in privacy, so perhaps a certain degree of embezzlement of value would occur in the distribution of sexy underwear or something.)
You can propose alternatives. Groups with different agendas (degrowth, space colonization, whatever) could throw around very concrete proposals amongst each other, debate them, and then place them as agendas for public consideration. These agendas could be much more legible than most policy agendas are today. For instance, there’s currently a debate about how much degrowth would adversely impact ordinary people’s living standards; publishing plans with explicit assumptions and results would clarify the terms of these debates.
You can play around with different assumptions. For instance, we have a pretty clear consensus that carbon emissions lead to global temperature rises, but within those models there’s a great deal of uncertainty. We also have to guess about the effectiveness of later technology. Because planning models can be given different assumptions, and because these assumptions are legible as code, individuals and groups can play around with them to make suggestions - for instance, to make the case for their chancy “startup” to be a part of the economy.
You can keep a track record of the accuracy of assumptions, which then feeds into what assumptions are considered plausible by the public. (Newer questions might rely on prediction markets, as in a previous post.)
the transformation of quantity into quality (and back again)
Large language models are capable of describing things in prose when a large class of those things have been described in prose by humans; they’re also capable of doing the reverse. The implication is that, even without a great deal of technical expertise, the average person should be able to participate in suggesting and critiquing economic plans.
Of course, this does not mean a role for experts in the ecosystem. Teams of experts would mostly likely be involved in producing the human-written critiques and evaluations of models - at the very least, we could imagine that all major political parties or factions would be writing positive spin in their models and negative spin on those of their opponents; an additional range of evaluations and descriptions would come from academic institutions, think tanks, and hobbyists. These human-written (or at least heavily human-supervised) proposals and descriptions would serve as the basis for large language models which laypeople could spin up in their basement (including the ability to get a “Blue Team” and “Red Team” evaluation of these!)
Of course this has limitations. Biases of the expert community would show up in the descriptions, and insofar as people could “choose their experts” we might expect some degree of epistemic closure to occur. Prompting an AI system is its own skill that has to be developed, and at which some have more natural talent than others. The more novel the assumptions brought in, or more complex their interactions, the worse training on existing models would be. However, I do think the ability to do this would mitigate the power held by experts and lower the barriers to participation in the planning process.
non-prescriptive planning
The economy need not have an actual central planning mechanism for all this to be useful (though it helps!) Just as prediction markets can be used to generate models that other groups consider, so can widespread access to linear algebraic planning models. Supply chains could coordinate on their basis. Insofar as the planner models market behavior, firms in a market could forcast, and individuals debating public policy (even if not concrete plans) could discuss a range of virtual plans that correspond to market behavior in a world where those policies are enacted. Workers in a firm could use this to more actively participate in discussions in the decisions of a firm, especially when any one firm has fewer variables to juggle, and workers are more intimately familiar with it than the global economy as a whole (though language models would likely be shallower.)