What is an (AI) planned economy? | Ideology explained

Transcript:
Welcome back to ideology explained. Today I will tell you another variant of the centrally planned economy. This time it is planned by an AI rather than by humans. As I defined it last time A planned economy can be summed up as an economic system in which the distribution of resources is taken over by a centralized institution. In this case an AI.
That means artificial intelligence for my non tech savvy audience. It’s basically the same thing as the centrally planned economy. Stuff get’s produced in some way, the planning institution has no influence on how and then the AI manages the distribution.
This is very much the modern approach to a soviet style planned economy, except better for reasons I will go into in a moment. It allows central management which is very good if you like to avoid waste. Remember the free market overproduces so much that up to 40% of everything that is produced get’s thrown away before being bought or used.
A centrally planned economy avoids that. And an AI planned economy would be even better at that. Remember what the downsides of the centrally planned economy where?
“This is one of the main problems with the planned economy. It gives a few people the power to plan the entire economy” “Another problem with a planned economy can be human error. If the planning institution makes a mistake there could be a shortage of resources.”
Well, with a good AI both of those would be solved! Humans can be very bad at planning ahead and predicting changes in demand which is why the soviet union had a 8 year waiting period for cars at one point. But theoretically an AI would not have that problem.
And an AI wouldn’t become corrupt either so that’s not an issue. But of course there are problems, as always. One is that an AI like that has yet to be written or even simulated. I happened to go to an IT school so I know a bit about that. Basically it would be a self teaching AI.
The idea is you show the AI a lot of data and it analyses it and searches for patterns. If you give it enough data at some point even the people who programmed it don’t know how it works anymore. And of course that means it can make good decisions which humans wouldn’t come up with.
Though my semi-professional opinion on that is that it could be very dangerous to trust the entire global, or just national, economy on a black box which nobody understands. And nobody can understand. AI always does SOMETHING different from how you want it to act. If you want an example look at the Youtube demonetization bot. It does something and nobody can explain what.
Of course this may sound like I am scared of technology but I am really not. I just think we should be very careful about how much we want to trust a single system. What if it makes wrong decisions? What if it develops biases? Yea that happens, because of the way they are programmed many AIs take on the basis of the people who programmed them. The reason for that is out of the scope of this video.
And what if there is a power outage? Will our entire global trading system grind to a halt if our central supercomputer goes offline? What even is to say that technology will allow such a powerful AI? Mores Law is basically made up so don’t cite that. An AI which could plan the entire economy would need an incredibly detailed understanding of the world.
It would have to know about weather patterns in order to predict droughts and about future demand of products. That takes huge amounts of computing power. All of those questions can be answered, but we need to ask them before we can apply a system like that.
In conclusion. An AI planned economy is basically a planned economy without the downsides but with more uncertainty about how it would work. There are many questions which have yet to be answered, though I have to admit that none of them seem like impossible challenges. Unless you are an anarcho-primitivist in which case this system is not for you.
A tanks to my patrons for motivating me to make videos and an extra special thanks to Aaron J Peton for your generous donation.

[joeys]

Decentralized AI. Each commune gets their own giant computer and then they all link up

[thomasnewman]

The black box is a really important point to understand, with self-learning algorithms you have NO IDEA how the AI reaches a certain conclusion. You give it example scenarios and feedback on results (right here, wrong here etc.). But self-learning algorithms could potentially come up with any way to distribute goods, for as long as it meets the criteria it thinks you want.

[azertyQ]

a computer planned economy wouldn't even necessarily have to be an AI (neural network) system, though it is frequently suggested because they benefit from large data sets and don't have to program anything beyond constraints and update procedures.

The "calculation problem" is ultimately a HUGE set of equations that must be simultaneously solved; focusing on the methods of calculating those solutions obscures the problems of collecting data that is accurate both to the system as designed and to the people it ultimately serves.

[afroslacker]

So basically the Sybil system, but for managing the economy.

[missk]

Paul Cockshott made a prototype of such AI

[jurgenruegenberg]

I would love to read this in an article or magazine or book can you please recommend one?

[dhirom]

Hey, have you ever thought of making a video on open source software?

[Samgurney]

This is a little bit fuzzy. There's no need in computerised planning for demand forecasts being the basis for production (besides this would risks being stultifying, because demand changes as innovation increases options in the product mix). It's not something I would rule out but it's unncessary and none of the details have been spelled out other than handwaving about neural nets. E.g. are there prices in this system? If so, how does the algorithm predict demand if price or quantity changes? And if this is the basis on which planning is to be done, why not just create a program that does regression analyses to estimate demand elasticities - why leave it to an incomprehensible black-box?

Computerised planning is an excellent idea, but I think you are perhaps misunderstanding what it's all about. It's mostly a response to the equilibration aspect of the calculation debates: with modern computers we don't need Lange's unwieldy tatonnement process, and the equations for equilibrium (given a set of convex production functions and either a target vector of final goods or an objective function to optimise) are now computationally tractable in a way they were not prior to computers. That is the relevance of computerised planning, rather than the ability to forecast demand. This is compatible with a having a system of prices, such that extensive and fine-grained demand forecasts are irrelevant.

[gradedwash]

The only problem with AI is what it starts with, if that is human error of design then that human error may not necessarily be eradicated by the AI since it cannot make decisions of that nature but the human error is likely to be magnified as desirable within that Algo. Imagine that an AI used punishment for unemployment upon a populace where the available work was already allocated to AI but the algo used by the AI was told to punish such people as were not earning income by reducing their income even more! as is current Government (both left and right ) thinking. Imagine also that a fresh recorded death being in the order of 3 persons outlived others to reach 101 and consequently millions of people having their pension ages rebooted as twenty years away from where it is now meaning that hardly any of them would live to see pensions and would continue to be punished as out of work. Perhaps in future they will not have pensions in any case having a much better strategy in place. (Lets hope not Solent Green) It all depends on what ideology starts such a system. Tell an AI that all must work and also tell it that any paid activity can mean work and you might have an AI designing how to poke in to peoples private worlds to determine if children are receiving money for shopping trips etc, they may even be given to make the assumption and deduct in any case without any proofs down to extra purchases being made at the supermarket in a family of four, two of which are children. Life could get very tricky the more intrusive government becomes. That is why free speech and liberty are so important right now, allowing person to gather facts and make up their own minds. Protecting that right instead of dictating as is currently underway.

[Isomoar]

Basically any large corporation such as Walmart or Amazon is run this way.