Comment by sudosysgen | Hacker News Reader

Comment by sudosysgen | Hacker News Reader

An interesting point here:

>would correspond to an optimization problem that would take a thousand years to solve on a modern desktop computer

If this really is so, then the problem is tractable nowadays - a thousand 2016 desktop computers years is something we could do multiple times a day with a modern supercomputer, so definitely feasible.

Same for data, modern networked devices could provide data of sufficient quality.

That being said, the big one is of course to find the function to optimize. I had an idea about this not long ago - the issue communists have with capitalism is that it has a class divide in production. So why not try this: all consumer goods must be sold on a pseudo-market. So, after production, a second algorithm sells the products at such a price that supply and demand is solved.

Using the prices determined by this market, one calculates the ratio of labour (time x complexity) and resources per dollar of price found when watching supply and demand, and the optimization system could be run so as to attempt to maximize the ratio of labour+resources/price.

This way, people will get what they want the most, and the system should optimize for the best way to make people work the least, which seems like the primary contradiction between an individual's work as producer and consumer.

Of course, there would still be quite a few inefficiencies, but given the fact that the Soviet economy was not that far from that of the US, it could be a net positive in pure economic output? Especially since this wouldn't suffer from any economic crises at all.

What would remain is a way to incentivize innovation, maybe somehow by including successful innovation into the labour price of a commodity?

Of course, outside of this, there would also be serious political issues with the necessary centralization for this, and you'd need to find a way to trust the economic calculation process.

All in all, this feels... much more tractable to accomplish than what I would've thought. Maybe even tractable enough that it could be, in theory, implemented today, given sufficient political will and a way to get the most competent people possible on it, which of course is far from a given.

The main problem I can think of is the politicking involved. eg. special interests leaning on politicians to get their industry prioritized some way in the system. Even without lobbying it's possible for this to go sideways, if well intentioned policy makers decide to use the system to promote certain agendas. For instance, the central government might place a huge bonus on steel production to promote industrialization[1]

[1] https://en.wikipedia.org/wiki/Backyard_furnace

Well, I think a big reason why this happened is that there really was no objective way of deciding which industries should be prioritized. It just so happened that steel production seemed to be the industry that prioritizing would allow for the most future growth.

In a way, this was true for the USSR - the Red Army in WW2 was incredibly well equipped in numbers, which would have been impossible if they had not focused on so hard in steel production.

But in theory, if one could get such a system going, and if it could be implemented in a verifiable way, then there would be no way for the government to justify prioritizing one industry over the other. Indeed, how could you argue that you make a better job than the algorithm? If the government decided to incentivize further productive growth, that should be done via a hyperparameter.

I think the major politicking issue would be getting the politicians to willingly give up their power to this system, which indeed would be a huge problem.

Correct, the politics and power dynamics are the primary barrier, not because the theories fail (they tend to go untested).

Look at Venezuela. A country that has been dependent on an oil-based extraction economy for decades and subject to brutal imperialism elects a socialist government that nationalizes... just that one industry, and the entirety of Western powers turn against them along with local national business owners, sabotaging their own points of sale as a form of economic warfare, instead selling across the border to support a black market.

As a reminder, despite the socialist label of the party in power, Venezuela has a significantly lower level of nationalization than France. But because France is not a target of both imperial and capital interests, but is instead firmly established as an imperial country itself, we don't see sanctions or (inaccurate) dithering over election integrity.

Venezuela is what you get when you propose to move some of the power and profits away from capital and don't have the footing to withstand their counterattack.

Your first statement assumes that linear programming solvers can be massively parallelized. Can they?

IIRC, yes they can be massively parallelized because they reduce into matrix algebra.

One could also go even further and make an ASIC for solving these problems à la Tensor Cote.

I'm being downvoted for some reason, so here is a paper on a massively parallel linear programming solver that exploits the fact that the problem can be reduced to matrix operations : https://www.tandfonline.com/doi/abs/10.1080/1063719940891547...

Seems like I remembered my optimization class correctly after all!

It is more tractable some argue [1] that companies like Walmart show that it is possible to centrally plan an economy.

[1] https://vancouversun.com/entertainment/books/book-review-att...

That's also the argument made in the recently reposted JBS Haldane "on being the right size":

http://www.phys.ufl.edu/courses/phy3221/spring10/HaldaneRigh...

A small country is already smaller than a gigantic corporation. Of course the larger corporation try to have an internal economogy (with department charging each other), but they're still mostly centrally planned.

So you could say the modern capitalism with gigantic corporation is already an exercise in central planning.

"""To the biologist the problem of socialism appears largely as a problem of size. The extreme socialists desire to run every nation as a single business 5 concern. I do not suppose that Henry Ford would find much difficulty in running Andorra or Luxembourg on a socialistic basis. He has already more men on his pay-roll than their population. It is conceivable that a syndicate of Fords, if we could find them, would make Belgium Ltd or Denmark Inc. pay their way. But while nationalization of certain industries is an obvious possibility in the largest of states, I find it no easier to picture a completely socialized British Empire or United States than an elephant turning somersaults or a hippopotamus jumping a hedge."""

The other thing to consider is failure modes. If a corporation misallocates resources, it's the shareholders that are footing the bill. Corporations make mistakes and go out of business all the time. If you apply this to the entire country, you end up with the the whole country footing the bill for policy mistakes. It already happens to an extent in capitalist systems (eg. lobbying by special interests), but centrally planned economy on a national scale will only exacerbate it.

Corporations incorrectly allocating resources leads to exactly the same issues as a centrally planned economy doing so.

The issue is, corporations invariably do so.

EDIT: Downstream comments are disputing the assertion that corporations misallocating resources causes issues for everyone. Here is why :

Not really. I'm going to use some vaguely marxian terms here because it's the only framework that applies both to centrally planned economies and market economies, but I won't be using any explicitly marxian theories, correct or not.

So, let's say that a member of the capitalist class (shareholder) sees their profit drop or even go negative as a consequence of the misallocation of the resources.

This reduction in profit without any benefit anywhere else means that the capitalist cannot increase their amount of capital. Assuming a more or less fixed rate of profit otherwise, this decrease in the amount of constant capital means that the amount of variable capital (labour) necessary to produce the same commodity is going to be higher, which means that it will cost more labour to buy and produce.

It will also mean that it is less likely that the capitalist class will increase variable capital, which would increase the monetary price of labour as demand increases, which means that wages are depressed for non-producible items such as land even at the same profit rate.

Finally, it might push the capitalist to reduce their profit rate in order to catch up to the competition, which might in turn push them to reduce their profit rate as they have more of a margin, which can lead to various crises down the line.

In the end, in capitalism, everyone pays for the errors of the shareholders, perhaps even more than the shareholder class themselves.

>leads to exactly the same issues as a centrally planned economy doing so.

how so? In a market economy the shareholders are footing the bill, in a command economy it's evenly distributed across everyone.

I edited my comment to contain a response to why this is actually distributed across everyone, almost to the same extent as in a planned economy.

however corporations are limited by the damage they can do to the extent that they are limited to voluntary interactions because they run out of money/resources to waste. If you allowed a corporation to commandeer more resources at gun point it would continue to waste them.

I'm that case, I always want the latest PC build every year that ends up costing at least $10k, regardless of the fact that I have literally no use for 1TB of RAM.

Reply on news.ycombinator.com