It's interesting to look at things from a controls perspective.
An economy (any kind of economy!) is just a feedback control system.
Generally, when you're choosing a feedback controller, you have two options: PID, where you have a set of 3 "dumb" rules, and all you have to do is tune the gains on those dumb rules. Or MPC, where you need to actually build a model of the system you are controlling, and then you can optimally plan to control it the future.
In general, MPC is quite good, much better than PID, if you have a great model of your system. It fails when the model is bad, or when you can't determine your state with accuracy.
PID can be very good -- for systems with very fast dynamics and response times, it is better than MPC, because you don't spend a lot of time in the feedback loop evaluating the model, and you don't need to worry about modeling at all ("model-free")
I think of an economy (very broadly) the same way. The mythical "free market" that everyone talks about is like a PID controller over the whole economy. It's a simple rule, supply and demand, and you can keep prices under control quite well with just a PID controller. But PID can fail spectacularly, too, and cause instability.
In general, you know when instability is happening when you look at a graph of your state variables and it looks like an exponential with a positive coefficient [1]. I'm sure you can see where I'm going with this: every single chart we can draw about our economy right now looks like an exponential. Energy, probably the most base measurement, looks like an exponential. This is bad [2] and everyone knows it, which I think is the source of a lot of the modern ennui.
Our old MPC models failed spectacularly, because we didn't have a good way of modeling the whole economy, we only had very granular and incomplete information about the markets, and we didn't have a way of actually solving the optimization problem. The obvious missing link here is the computer.
I agree wholeheartedly with the author, and I think it's really valuable to bring up that we already have central planning going on in the form of the corporation. There are thousands of people whose work is completely decoupled from profit and who engage in long term planning at a scale larger than the market they operate in.
What he describes is like a cascade controller, where the inner loops occur in the linear supply/demand region and the outer loops are a smarter model-informed controller. Rather than pretend like the outer loops are just like the inner loops (a pie-in-the-sky libertarian view) or pretend like the inner loops are just like the outer loops (the anti-capitalist view) we should be thinking about the strengths and weaknesses of both. Right now we're very good at the inner loop but we're basically randomly flailing about with the outer loop.
An economy (any kind of economy!) is just a feedback control system.
Generally, when you're choosing a feedback controller, you have two options: PID, where you have a set of 3 "dumb" rules, and all you have to do is tune the gains on those dumb rules. Or MPC, where you need to actually build a model of the system you are controlling, and then you can optimally plan to control it the future.
In general, MPC is quite good, much better than PID, if you have a great model of your system. It fails when the model is bad, or when you can't determine your state with accuracy.
PID can be very good -- for systems with very fast dynamics and response times, it is better than MPC, because you don't spend a lot of time in the feedback loop evaluating the model, and you don't need to worry about modeling at all ("model-free")
I think of an economy (very broadly) the same way. The mythical "free market" that everyone talks about is like a PID controller over the whole economy. It's a simple rule, supply and demand, and you can keep prices under control quite well with just a PID controller. But PID can fail spectacularly, too, and cause instability.
In general, you know when instability is happening when you look at a graph of your state variables and it looks like an exponential with a positive coefficient [1]. I'm sure you can see where I'm going with this: every single chart we can draw about our economy right now looks like an exponential. Energy, probably the most base measurement, looks like an exponential. This is bad [2] and everyone knows it, which I think is the source of a lot of the modern ennui.
Our old MPC models failed spectacularly, because we didn't have a good way of modeling the whole economy, we only had very granular and incomplete information about the markets, and we didn't have a way of actually solving the optimization problem. The obvious missing link here is the computer.
I agree wholeheartedly with the author, and I think it's really valuable to bring up that we already have central planning going on in the form of the corporation. There are thousands of people whose work is completely decoupled from profit and who engage in long term planning at a scale larger than the market they operate in.
What he describes is like a cascade controller, where the inner loops occur in the linear supply/demand region and the outer loops are a smarter model-informed controller. Rather than pretend like the outer loops are just like the inner loops (a pie-in-the-sky libertarian view) or pretend like the inner loops are just like the outer loops (the anti-capitalist view) we should be thinking about the strengths and weaknesses of both. Right now we're very good at the inner loop but we're basically randomly flailing about with the outer loop.
[1] https://en.wikipedia.org/wiki/Exponential_function
[2] https://dothemath.ucsd.edu/2011/07/galactic-scale-energy/