AI Planned-Economies

[Azalfia]

Planned Economy is a very scary term to some people, and really, I can't blame you. It's associated with the Soviet-Bloc Central Planning, which is generally not viewed greatly considering the mess the system could be, although it had it's achievements. Before we enter this conversation, let's define "Planned Economy"; a planned economy is a economic system where the investment, the production of goods, and allocation of resources is taken up through economy-wide planning. For most of it's history, the 2 major options was through a central government agency taking holding (eg the USSR and much of the Eastern Bloc), or through a decentralized approach (eg Rojava, The Zapatistas).

However, recently, the idea of an AI planned economy; a planned economy that isn't based off of a central government agency but through a computer system; has become more popular. If it works, it could quite possibly solve the Economic Calculation Debate (eg markets are the best at the allocation of resources) thanks to advanced technology and AI that is being made today.

So NS: how do you feel about the idea of this?

[Postauthoritarian America]

Siri, create the socialist workers' paradise.

[Azalfia]

Siri, create the socialist workers' paradise.

I really think an AI planned economy, mixed with decentral planning, could create a economy based on human need instead of profit.

[Rusozak]

Siri, create the socialist workers' paradise.

I really think an AI planned economy, mixed with decentral planning, could create a economy based on human need instead of profit.

So.. socialism.

[Azalfia]

I really think an AI planned economy, mixed with decentral planning, could create a economy based on human need instead of profit.

So.. socialism.

Yes indeed.

[Rusozak]

So.. socialism.

Yes indeed.

Until said AI realizes humans are inefficient and self-destructive and having no humans would achieve its desired goal state.

[Resilient Acceleration]

Can AI innovate?

[Merconitonitopia]

No markets, no price system, no rational cost accounting, no rational utility valuation. A more technologically sophisticated means of planning does not address the basic problem.

[Resilient Acceleration]

Yes indeed.

Until said AI realizes humans are inefficient and self-destructive and having no humans would achieve its desired goal state.

Tbh, I do think the present AI model of machine learning will hit a wall of diminishing return. The next stage that will be apparent likely involve neuroscience, quantum computer, or both (then again, they're also being heavily researched).

[Waldoven]

Hmm, I think it would be a good idea. Though I'd wait till we get closer to the singularity first. Like, waaaaaaaaaaay closer. Wouldn't want Siri making the price of bread 1 million dollars one night

[Cavirfi]

I had very recently watched the "Humans Need Not Apply" video by CGP Grey which doesn't talk about AI planned economies but the principles of such a video can still apply to the theory we are currently talking about. Overall, as you said; there is an Economic Calculation Debate where a lot of people believe that private enterprise is the way to go because overall private entrepreneurs are better at allocating the resources and finances needed to sustain society. But to quote CGP:

Robots don't have to be perfect, they just have to make less mistakes than humans.

I believe that an AI planned economy could work though it may take some time, mainly because the current driver of the economy is consumer spending which is controlled by us, humans and of course; an AI planned economy is obviously going to be ran by AI or computers. So a transition may be required and as you see with some mixed economies around the world (for example: The Nordic System), such planned economic system is impossible until all countries have adopted a mixed economy, which some countries are not willing to give up a totally free market (like the US) or a highly regulated market (like China)

[Merconitonitopia]

Siri, create the socialist workers' paradise.

I really think an AI planned economy, mixed with decentral planning, could create a economy based on human need instead of profit.

There can exist no such thing as a decentralised system of nonmarket economic planning. All production within an economic community takes place through a series of interconnected systems of production, thus there must exist some common system of coordination. In the absense of some form of market exchange, there must exist some centralised body that exercises total command over all production within the economic community for production to take place.

[Postauthoritarian America]

No markets, no price system, no rational cost accounting, no rational utility valuation. A more technologically sophisticated means of planning does not address the basic problem.

Markets, price systems, cost accounting and utility valuation based on the worth of things to society as opposed to their ability to transfer ever-increasing amounts of wealth to The One Percent. fify

[Resilient Acceleration]

I really think an AI planned economy, mixed with decentral planning, could create a economy based on human need instead of profit.

There can exist no such thing as a decentralised system of nonmarket economic planning. All production within an economic community takes place through a series of interconnected systems of production, thus there must exist some common system of coordination. In the absense of some form of market exchange, there must exist some centralised body that exercises total command over all production within the economic community for production to take place.

Arguably this is achievable through a hivemind system where everyone will have brain implants and instant access to all information, with far greater cognitive computing power through "external cloud brains" to discern it. Then again, in this case the word "economy", and "humans" or "psychology" or "utility" or "happiness" for that matter, would become irrelevant.

[Rusozak]

I also wonder about implementation. Would everyone just do what the AI says without question? Star Trek did something like that, but with managing patients at a hospital. Spoiler alert, it wasn't pleasant.

[Cerbia]

There was a book written about this or something like this. If anyone has any knowledge about the book I would be interested to hear what they have to say.

https://en.wikipedia.org/wiki/Towards_a_New_Socialism

[Bear Stearns]

this sounds like some #newnormal #buildbackbetter bullshit

[Senkaku]

get off NSG, Grimes


(anyways it would be v cool but my understanding is it would require computational power that we don't seem likely to acquire anytime soon)

[Neanderthaland]

Can AI innovate?

Actually yes.

Or, at least, AI frequently comes up with solutions to problems that it's designers didn't anticipate.

[Exxosia]

I think the biggest problems are:

1. When you deprive people of agency, things go south real fast. This would result in distortions that the AI would use in its functions that would create negative effects.

2. What you program into the AI. Any AI is going to be beholden to its programming and its data. Even under the most ideal conditions, the AI is going to perceive the data within the confines of it programming. Also, considering that anyone capable of putting an AI planned economy into effect has everyone's worst interests at heart would lead towards negative outcomes.

3. Corruption and hacking. We've seen already that overly connected systems are a liability right now. We have entire industries brought to heel by ransomware attacks, entire supply chains devastated by bad actors and data, massive security holes placed in basic components by China and the US. If we had an AI controlling an entire economy, it could take one bad actor, one RAM error, one read-write error, one state-sponsored switch-flip to completely demolish an entire economy.

[Nilokeras]

get off NSG, Grimes


(anyways it would be v cool but my understanding is it would require computational power that we don't seem likely to acquire anytime soon)

Yeah AI is just repackaged cyber Christianity. If we really wanted to we can do perfectly good central planning with just networks and the limited automation we have now - no need to invent a hypothetical god to do it for us.

[Mercatus]

So basically Skynet, but no nukes, and in control of just about everything.

I’m gonna have to give that a big fat NO. Letting a computer decide what is good for humans is a bad idea in and of itself.

[An Alan Smithee Nation]

I think an AI could out perform humans rather easily at running an economy. Hell I bet a well tuned algorithm could outperform a human at running an economy.

The problem is that like automated trading, it could cause unforeseen problems on a massive scale. This would exacerbate the trust issues people would have with it already.

I suppose the reality is that governments will use AI as a helper for economic planning for a long rime before they actually put one in charge.

[Kilobugya]

It's an interesting idea.

First a bit of precision about USSR-style central planning. You've to realize how they did it. They didn't have computers at all at the beginning. They did all the computations manually. One factory producing rivets is hit by a natural disaster, they'll have to fetch from giant files all the plans for production depending on those rivets and manually recompute all the possible targets for accounting with lower amount of rivets, try to get other factories produce more rivets, or substitute the rivets with something else. It was a nightmare. And yet... they managed quite fine. Sure there was lots of problem of corruption/bureaucracy/authoratarianism but that's another issue, and they didn't handle so well the end-customer goods, but overall they managed to become the world's second superpower in record times and beat the USA in most steps of the space race despite WW1, civil war, WW2, cold war.

Now let's go back to "AI planning". There was one attempt to do it, at least partially. Cybersyn, in Allende's Chile, 1970-1973. The total computing power they had was less than any cell phone now packs. They had a very short time, they were facing sabotage and economical sanctions. And yet... they managed. Thanks to a partially-functional Cybersyn, the Allende government managed to withstand the truck drivers' "strike" (drivers were paid 2x to 3x their wage by the CIA to strike), and keep Santiago supplied despite operating with less than 10% of normal logistics capacity.

So, yes, I do believe with the modern computing capacity we have, all the progress done on multi-linear optimization and IA, the previous example of Cybersyn, ... we could have a cybernetic planned economy that would work quite efficiently at allocating resources, quickly finding harm-minimizing solutions even in case of unexpected contingency (such as hostile action or natural disaster), or at least a partially cybernetic planned economy (a few sectors could be kept as a market, like even in most capitalist countries a few sectors are still handled by the state outside of the market).

But the main problem is what the goals should be. AI, at least current-stage non-sentient AI, can't set goals nor priorities. So we would need ways to set those goals and priorities, and without them being decided by a nomenklatura or an elite team of programmers, but reflecting the desires and preferences of the people. That's a very hard problem, reflecting people's will into the programming of the AI. But we might be able to do that progressively, first by using cybernetic planning for stuff that's already mostly (or at least should) government-runned, like infrastructure, healthcare, public transport, ... and progressively extend it.

[CoraSpia]

I'm not a fan, not just because my background is richer than average in my country either. I wouldn't even agree with it if it was guaranteed to make me richer, because I'm fundamentally opposed to giving anyone or anything high levels of control within our daily lives and our personal choices with the law to back it up.