[Aloha]

I'm not certain that cybernetics or AI or whatever can solve the issues inherent in central planning.

As soon as a measure becomes a metric is ceases to become a good measurement of success.

The Soviet system failed both because the planners couldn't deal with modeling a complex economy, but also because they had no idea how much a given good cost to make, produce, or deliver to its point of sale.

It failed because all decision making was centralized, but no services were, so every enterprise had its own mix kindergartens, schools, recreation facilities, clinics, commissaries and dining facilities.

Any system that measured productivity of railways for example simply in ton-miles, with a quota of ton-miles to be met, I'd say is basically doomed to failure, particularly one with an inflexible plan governing it.

When the system worked, it was from local managers calling in favors and doing trades to keep production rolling.

A neat statistic, the planned economy of the SU, used as much or more oil per capita (I think twice as much) than as west, and produced 30% or more less in GDP, per capita. This is true of other consumable raw materials, all while spewing comparatively massive amounts of pollution out.

[Barrin92]

The notion that Cybersyn was a central planning system is common but untrue. That's not what cybernetics, which drew primarily from systems theory approaches, was ever about.

The so called 'cybernetic factory' at the heart of the project was about making real time information exchange between several levels of hierarchy possible, with much delegated to the factories themselves. Centralized decision-making was reserved for high-level planning akin to Auftragstaktik in the German army (I believe it's called mission command in the UK and US).

Friedrich Hayek of all people, who met Beer at a conference in 1960 in Illinois was actually very sympathetic to Cybersyn. The frequent comparisons between Cybersyn and Soviet Planning are pretty much the result of a political campaign against Allende, it doesn't have much to do with the scientific ideas behind the project.

If you have heard people talk about 'smart factories', 'industry 4.0', IoT devices gathering data in real time and relaying them to big computers where they can be analysed and quickly sent back to the workers you have in reality stumbled upon people who are trying to build CyberSyn 2.0, probably without knowing it and less socialist branding.

Cybernetic Revolutionaries: Technology and Politics in Allende's Chile by Eden Medina is a good, in depth book on the topic.

[Aloha]

I've expensively studied Soviet central planning, and I'm hard pressed to tell a difference here. But I'll go read that book you've liked.

Ironically, it's the tankies (neostalinists) who seem to be the loudest defenders of it. If it's propaganda it seemingly worked on the wrong audience.

[sudosysgen]

I don't think your comparison to Soviet central planning really works. As far as I understand it, Soviet central planning worked in long term plans and quotas, with orders given from GOSPLAN all the way down, and if factory owners disagreed they'd have to go send the information all the way up and back down the ladder, while the actual information on production would be manipulated to look good and sent manually by bureaucrats.

On the other hand, cybersyn had no quotas, and no long term rigid plans. Informational exchange did not have to go through bureaucrats, and could be done directly from the factory, eventually automatically by the equipment, and update a digital model that would do a large part of what the bureaucrats at GOSPLAN would do, but automatically and verifiably.

So it seems very, very different to me.

[Barrin92]

> If it's propaganda it seemingly worked on the wrong audience

Not uncommon though, right? The tankies have also adopted the narrative of China as a Communist threat, or in Latin America you end up with the paradoxical situation of most opposition parties in Venezuela belonging to the Socialist International, yet you find feverish and uncritical Maduro support among young radicalized left-wing Americans.

We have the same phenomenon in Germany where ironically the formerly West-German members of Die Linke have a much more cartoonish, ideological view compared to their actually ex-socialist former East-German counterparts.

[Aloha]

Honestly, the fact that tankies exist in 2022, still bewilders me.

[Aloha]

Err, extensively even.

[vidarh]

I've argued for some time that "solving" central planning is equivalent to solving the halting problem, because people can react to everything the planners do in ways that intentionally or not counteracts the plans in various ways.

Cybernetics can improve on that somewhat by changing static plans into dynamic sets of rules for how to react, but the need to respond dynamically is pretty much inescapable and then it would seem better to lean into that.

[sudosysgen]

>the need to respond dynamically is pretty much inescapable and then it would seem better to lean into that.

I don't quite understand your point. Could you elaborate further? What do you mean by leaning into the dynamism?

[vidarh]

What I mean is that instead of trying to plan and control details that requires all kinds of complex contingencies to even try to keep things under control, look at whether or not you need to control them or can avoid the need to exercise control wherever possible.

E.g. I get the desire to control prices - it's a way of feeling that one has control over important social goals. If you genuinely want to address poverty, for example, it may feel very reassuring to be able to say that food for one person should be possible to buy for X amount. If you dictate that and you're able to enforce it, then ensuring everyone earns or get at least X for food plus whatever they need for other things means you've solved their access. In theory. As long as production actually meets sufficient levels. The problem is that you're providing no effective signal into the system of how important sub-goals of this is. On the contrary, if you set prices of products across the board, you're suppressing such signals. E.g. if the agricultural sector is in dire need of new machinery and fertiliser to meet production goals for food at a certain price but can't actually bid more to get it, they depend on convincing a whole lot of people that they actually need it more. And in practice we've seen that fail time and time again.

An alternative is to focus on attempting to control the final outputs you actually care about by bidding for the outcomes you want in an open market, and leaving your control problem to one of controlling your bids to get the best outcomes possible. You can "bid" either directly - by buying contracts for e.g. delivery of food at certain prices if that is what you want - or indirectly by e.g. adjusting welfare payments to compensate for prices outside of your desired levels, or a whole lot of similar mechanisms.

In that context cybernetic thinking can still have a place in terms of control loops to try to keep competing outputs within reasonable parameters based on as up-to-date information of factors that may affect prices as possible.

The point is that you get the best out of dynamic systems when you focus on controlling only the bits you actually care about well rather than imagining you know how it all hangs together and trying to control all the intermediate steps in detail because the latter requires you to keep up to date with every aspect of the most cost effective ways of meeting your targets. Not least, by focusing on controlling outputs you allow for redundancy in the methods used to achieve them that has a good shot at mitigating failures, as well as to avoid catastrophic results of conflicting priorities.

(EDIT: I'm involved in another discussion about typing in programming languages, and the paragraphs above actually fit into much of my thinking of static vs. dynamic typing as well as in much of life in general: A lot of problems are tempting to solve by exerting as much control as possible, but in dynamic, fast changing system expending your effort on exerting the control only over the things you actually care about allows for a more flexible overall system)

Even for a socialist system there's absolutely no ideological reasons for requiring central planning. The notion of the need to do away with market mechanisms as planning signals in favour of a central, state controlled plan, was largely a Bolshevik conceit that a lot of other socialist ideologies strongly disagreed with (not least because many of them want to dismantle the state, leaving no organ with power to enforce central planning). But at the time there was also a severely lacking understanding of the inherent complexity of large scale planning mixed with a rather unjustified optimism about the ability of humanity to overcome them.

EDIT: Allende represented a mid-point - nowhere near going as far as the Bolsheviks, and Cybersyn addressed a lot of problems in introducing feedback loops rather than strict central planning, to be clear.

[sudosysgen]

I don't think we can actually attribute the refusal of markets as a Bolshevik position. Lenin was very favourable to the use of markets, while Stalin later was opposed it to it in no small part because he felt it wasn't developing heavy industry rapidly enough, which turned out to be crucial to winning WW2.

The post-Stalin dogma as well as the calcification of the Soviet economy was an impediment to the trials of such planning, but some Soviet economists like Kantorovich in fact proposed essentially the same system. It was refused for two reasons - large computational cost at a time where computers were quite weak, and cynical Soviet bureaucrats not wanting to lose their power, hence many opposed it. But quite a few were indeed supportive.

Later in the 70s, the Soviets did adopt some market economic implements in the form of cooperatives, so I don't think they were unconditionally opposed to it.

Later, there would be the issue that any planning system would be opposed not only by pro-planning bureaucrats, but also by the rising anti-communist faction of the party.

Interestingly, I think it is Lenin that most acutely understood the difficulty of central planning, hence why he proposed the NEP that Stalin later dismantled.

On the rest however, I think we are both in agreement :)

[vidarh]

I halfway agree with this. NEP certainly shows that Lenin at least in theory was open to it, though NEP was a climb-down. It's of course impossible to see what Lenin would have done if he'd stayed alive, but I guess the closest parallel would be Deng's reforms in China and Đổi Mới in Vietnam, both of which certainly either improved growth or coincided with it (the one caution is that you can see Chinese growth starting before the reforms - China's GDP growth was punctuated through Mao's reign and into Deng's by every political upheaval Mao caused, and as such the period of stability Deng brought clearly is at least part of the reason).

I don't think I'd agree Lenin was all that favourable to markets initially though. It seemed to me to be more of a pragmatic attitude and acceptance that things couldn't continue the way they did prior to NEP, and it being short enough after the civil war created a very convenient way of drawing a line under the old policies as driven by the war.

Stalins role in dismantling it is certainly critical.

[sudosysgen]

The big idea of cybernetic centralized planning is exactly to not have to compute how much to charge for an item, but instead to set a lower margin and let supply and demand be processed with consumer markets. It's not explicitly written that way, but it's the actual effect.

So instead of having a quota of ton-miles, you will simply have a bunch of goods on sale. If a lot of people are buying them, you can increase their prices and at the same time increase production.

Once you compute how best to minimize the difference between demand and supply, you compute what goods have to be produced, and work backwards from there until you can calculate what inputs they need.

There are a few proposed metric. Essentially the idea is to completely do away with any quotas outside of exceptional circumstances, and respond to demand dynamically in a way to reduce the amounts of hours worked and increase the consumer satisfaction, as measured by their comparative subjective value for various goods.

We both agree that the quota system and the inefficiency of the Soviet system was considerable. However, another interesting statistic you may not be aware of is that, outside of times of war and for the last few months, Soviet GDP per capita never decreased, while still providing fairly acceptable living standards, and good ones by global standards.

The fact that it was actually not really doomed to failure at all and it's comparative performance compared to the rest of the world, as well as that it lasted quite a while shows that there is enough value in the good sides of the historical centrally planned economies to partly make up for the atrocious and seemingly obviously debilitating issues.

The idea of cybernetic central planning is to reduce stress, offer more social safety, reduce economic inequalities, all while doing away with quotas, inflexible plans, precomputed prices, inflexibility to demand, and sensitivity to fraudulent information.

Basically, give up on trying to guess what to make and in what quantities, and let the consumer decide what they prefer, but then intelligently use that information to determine how to organize production as well as possible, and then measure the results of these organizations in real-time to provide feedback and adjust various processes.

An interesting way to look at it is that, in capitalism, Capital has a mind of it's own and basically acts as a control system. In essence, in capitalism, capital wants to grow, and it does so by estimating all of the returns it could be generating, and if it is credible to increase returns, re-allocate itself to the most profitable return. So, through a few dozens of layers of price signals, it can control the economy to increase it's own profit, and in doing itself it also increases productions and allows capital pledged to consumption to have goods and service. Cybernetic planning intends to directly measure the price signals as close to the source as possible, and when possible even underlying signals before that, and use those to optimize some function that will allow people to work as little as possible but have as many goods and services as possible. The idea is that, as with the control systems humans program, we deal with the lower complexity of the model (at least for the foreseeable future) with more direct access and dynamic access to the data and to the levers of action.

[vidarh]

Just saw this, after writing my lengthy reply to you elsewhere (plus a couple of edits), and I think we agree on quite a bit. Cybersyn certainly was a substantial improvement in thinking over strict central planning. At the same time I think it still it aimed for unnecessary levels of complexity vs. using market mechanisms as a large part of both the input and control mechanisms, and trying to assert too much control.

[pastacacioepepe]

And how is the global market not unnecessarily complex? It is optimized for the wrong metric (profit) at the cost of every other one, causing additional complexity and hurting overall resiliency.

The market is a constant failure at the only thing it should do, properly distributing resources. It's failing clamorously to distribute covid vaccines globally, for example. Additionally, it is easily exploitable everytime there is any unbalance, see bargainers.

The lack of control you seem to present as an advantage is exactly what makes it controllable by a very small minority of powerful entities.

Actually I start to see some parallels with the theoretical "lack of control" that is claimed to be the advantage of cryptocurrencies by enthusiasts.

[vidarh]

You're conflating market mechanisms as a general concept to the current global market. You can use market mechanisms as a resource allocation method in all kinds of different models. Market mechanisms are used in all kinds of settings to match prices of all kinds of things, and they work well when they are designed to be fit for purpose.

A market is ultimately nothing more than the iterative application of a function that takes a list of buy and sell orders and outputs matches. It can be entirely unregulated or however regulated you want. There can be no constraints on who accesses it, or lots. There can be synthetic manipulation on one or both sides to adjust the functioning of the market, or not.

I'd argue for all kinds of regulations - I don't think unregulated markets are a good thing at all. But a well regulated market is simply a consensus mechanism that reduces what is a fundamentally computationally intractable control problem if you try to handle it in a centralised manner by dividing it into a range of smaller problems and distribution the effort of figuring out the best way of meetings goals. It's never going to give an optimal outcome, but you can mitigate the risk that it will produce really bad outcomes, and get closer to decent results.

I'm also not arguing for one market, but for many, I favour using market mechanisms to price externalities whenever possible.

The current global market is not set up to "properly distributing resources" because that is not the goal of its participants, so you have a good argument against current global markets for products. If you were to create a market actually intended to fairly distribute scarce healthcare supplies and it would look very different, with pricing based on e.g. predicted outcomes and benefits for different bidders.

You can still plan in the face of well regulated markets, but the planning process would involve packaging up incentives and regulating the pricing of externalities without trying to guess what will work best in ways that forces the planners to become experts in everything.

> The lack of control you seem to present as an advantage is exactly what makes it controllable by a very small minority of powerful entities.

Even if we were to agree this was unavoidable (I'd agree it's unavoidable in an unregulated market with vastly different access to resources, but that is just one subset of possible markets), a planned system is potentially controlled by an even smaller entity of even more powerful entities. We've seen that fail time and time again, because of the hubris of people with power thinking they knew best. Decentralising power is if nothing else a risk mitigation, and the failure of current markets to do so well enough is not an argument for a mechanism with a distinctly worse track record.

[Aloha]

You're largely hitting the points I wanted to make, there fundamentally is no replacement for the market price setting mechanism.