103 Comments
Commenting has been turned off for this post

I've kinda wondered if AI will turn out to be a bit of a bust. I like the self-driving cars example. Even if the perfect self-driving car came out tomorrow, it would still be worse than old-school mass transit. More emissions, less walkable cities, can't walk around and go to the bathroom like I can on a train and some larger busses.

I could see the same happening in other areas of our life. Maybe we have our "automated" purchases, but it's a clunky system that gets a lot wrong. Then we want to get the thing we actually want, we have to spend 3 hours on an "automated" phone line to get like a pound of potatoes or whatever.

Expand full comment

We consider the current system a capitalist one because companies more or less try to maximize profits, individuals make fairly free decisions about purchases and employment, and the government to whatever degree stays out of all of those company and individual decisions. It's actually a mixed economy, but we can leave that aside for now. I don't see those fundamental characteristics changing much as a result of delegating more of the specific tasks, so I'm not sure why we wouldn't still call the overall system capitalist. We could consider employees as an analogy. When the owner of a firm delegates a decision to an employee, the intent remains more or less the same. Sure, there are some agency problems, but that's a side effect. I would overall expect decisions to remain fundamentally capitalistic in their goals and criteria even if they're made by an AI rather than an employee.

Perhaps we would see some emergent changes in the system once most decisions are made by AIs (e.g., more implicit collusion), but that seems purely speculative at this point.

Expand full comment

I am not much of an economic thinker, but "AI" stuff drives me crazy. I think we should strike out "artificial intelligence" from our usage completely -- it's too hyped and imprecise (and in my opinion speaks of something as a given that I believe is impossible). And even "algorithm" has been corrupted. We just mean computer programs, right? Some of which are machine learning programs, which behave a little differently than traditional ones, but yeah still computer programs, which are written and maintained by people and are just electrical signals performing binary math to represent information that is intelligible to people.

I know it quickly gets to a point where the complexity of the operations being done and the speed at which they're done can make it difficult and time-consuming for people to untangle what a computer is doing, but the computer itself actually does not know what it is doing. It needs human assistance for everything it does. Behind the magic feeling of Amazon or Apple or automated trading or whatever is thousands of people working on clunky machines all day.

Whenever I am welcomed to imagine a magical mechanized future by a hyped up PR statement that just sounds like industry press to make investors excited masquerading as academic lecture at Harvard from Goldman Sachs, where they brag that ...

" 'In 2000, we had 600 humans making markets in U.S. stocks,' Chavez told the crowd in Harvard Science Center. 'Today, we have two people and a lot of software.' ...

Now, one in three Goldman Sachs employees are engineers, Chavez said.

'The future of the financial industry lies in virtual machines and strong API contracts,” Chavez said. “We are redesigning our businesses around those principles.' "

... I have mental sirens blaring. What the fuck does the Chief Financial Officer at Goldman Sachs know about "strong API contracts" ???

Ok, yes, the nature of their business has changed, and so also has the nature of the economy. But computers are just code and code is written by people. And code, though it feels magic when it works, is actually very weak, messy, dumb, vulnerable, and constantly changing, constantly in need of human help.

Anyway this is tangential to your point, which is that people like it when computers do shit for them, which I think is very true. And I think you're right that the spread of more "automated" experiences in everyday life will be gradual, like with anti-lock brakes or cruise control.

But man I get angry when I read "AI" hype !! Computers are not "intelligent"! There is no such thing as a "driverless" car and never will be! People like it when computers do shit for them, but they shouldn't. In fact it is not the computers doing the shit, but the code, which was written by other people, and I guess in the case of your linked article, Goldman Sachs people. Anyway sometimes I think about developing this general anger into more detailed arguments, but I'll stop now.

Expand full comment

An AI-managed economy is absolutely in our future, but it is a huge jump to assume that it will be centrally managed.

It's counter-intuitive to most, but the more the economy depends on AI (really math and programming), the lower the barrier to entry becomes. Renaissance Technologies, by far the most successful investors of all time, is a couple dozen Math and physics PhDs. Jane Street, a massive player in trading, has 1% the headcount of major banks. There are countless examples of a few programmers with VC backing upending entire industries in a few years. A single paper with a few authors can shift the direction of the entire field of Machine Learning.

The future I see is that of a few big players doing what they always have: undercutting or just buying out the competition. But also an endless stream of new competing AIs that are better

and keep the system decentralized.

Expand full comment

Though I've always been a techie and early adopter of just about everything I can reasonably get my hands on, I'm also the one that won't give Apple or anyone else my fingerprints, I cover up cameras on all my devices, until needed, and I never ever use or buy anything Alexa, Siri, Google, or anything else that purports to "smart" my life by spying on me and collecting all data about me and my household as is possible. I puke all over the smart fridge that tells you when you need to replenish eggs, and the Amazon tokens that auto reorder essentials. I know tech is still collecting all my data it can get its hands on, but I'm not willingly handing any of it over. There is massive resistance to the conveniences companies are trying to sell us, because they are creepy as hell. I don't see people jumping en masse on the grocery-reordering-fridge. I don't know how long that product will last before it ends in the failed-products dump heap. All the tech I gladly adopted throughout my life made my life more pleasurable and efficient in some way, brought something new I didn't have before, did something in a better, faster way. The current spy-tech is designed to wring value out of me, under the guise of giving value to me, value I reject for being undesirable.

I wonder what portion of the overall economy is devoted to moving around non-essentials. Utilities are a standard and largely unembellished, one-size-fits-all essential (we don't shop for the best electricity or water). You can say health care is an essential that doesn't need to be all that customized. Housing is an essential, but it's far from standard in its particulars. Just about everything else that I can think of falls into the highly customizable purchase category, in terms of quantity, quality, and other utilitarian or stylistic features. It's one thing to delegate tiresome chores (transportation) and decisions about essentials (including how to produce and deliver it) to AI. It's quite another to hand over decision making over non-essentials, a big chunk of the economy, and our entitlement to pleasures and escaping the hell that's sameness.

Expand full comment
founding

I can see our consumption habits becoming more automatic and controlled by AI. That’s certainly Amazon’s goal – they advertise products I want, and they really push subscriptions. Amazon once tried to get me to subscribe to a hairbrush (like, the default selection was that I’d get a new hairbrush every three months).

But I’m not sure how the change will impact workers. I assume the purpose of sending me products is to bill me for them – and that the products will stop coming if I can’t pay. I will still need a job. We could use all this technology to free people from the 40+ hour work week, and to provide everyone with what they need regardless of their ability to pay -- but it’s sort of hard to imagine that we will.

Expand full comment

While I generally concede the logic, I always roll my eyes at most "AI takes over the world" predictions and don't actually expect them to happen, and this post has crystallized a big part of why that is for me. I feel like there's a general idea that AI is a single entity, and there's one paperclip maximizer that will rule everything, but...why? Netflix has AI, Amazon has AI, Apple has AI, and as they get bigger and more sophisticated, there's no reason they should be able to cooperate any better than humans do. Collecting action dilemmas don't exist because people are stupid, they exist because oftentimes incentives genuinely line up that way, and it won't be any less true for AI.

Expand full comment

Relatively free markets, the rule of law and goverment by the consent of the governed has brought us the richest society in history. Why do socialists pine for a return for centralized fudalism to the point where they imagine a perfect AI to centrally control everything without humans even noticing? First off, it's never been tried before, and it sounds pseudo scientific, so Socialism's history of failure is made irrelevant by this fantasy. Part of the dream is that the people won't notice thy're no longer in control, because Socialism's universal failure makes it unacceptable under it's own name.

The funny part of this is the misunderstanding of how AIs work. They codify human knowledge, then apply it to large numbers of cases. They "learn" based on rules for learning programmed into them by humans. They don't run whole companies, let alone whole economies. They can improve on successful human efforts to run things better. They can't invent new ways to run things that improve on failure.

I spent a 45 year career in IT. Computers ain't magic. Socialist Garbage In, Socialist Garbage Out.

Expand full comment

This strikes me as a bit utopian. The economy isn't capitalist because the capitalists are in charge of making day-to-day investment and purchasing decisions, it's capitalist because the benefits of those decisions accrue to the capitalists (I am aware of the seeming tautology thanks to the terminology, but I trust you can look past that), and I don't know why algorithms would change that. I mean thanks to managerial capitalism, for a lot of capitalists it probably already largely resembles your vision anyway. Who cares who is making those decisions, whether a computer or an MBA; the money still flows upward in ever-greater proportions. The fundamental change will still have to be political. Cybersyn wouldn't have been socialist without the decision to spread its benefits among the population.

Expand full comment

Such an AI cannot exist in the real world, any more than your pitcher who always gets exactly one run per inning.

Butterfly effect. Quantum chaos.

Expand full comment

The AI described here, rather than seeming alien, reminds me of a familiar figure in human history: the trusted servant. In this future, humans are Bertie and AIs are Jeeves, managing our households and making financial decisions on our behalf. Were Edwardian era aristocrats living in post-capitalism? I'm not sure I would agree.

Expand full comment

The thing that stops your car is a brake, not a break

Expand full comment

The parts of capitalism that won't go away because of AI are the ideas and the risk. How can AI replace those?

Expand full comment

The Scottish science-fiction writer Charles Stross painted an exaggerated version of this picture in his novel “Accelerando.” He called it Economy 2.0.

Expand full comment

Glad I won't be around to see it. Though I would have liked driverless cars.

Expand full comment

I'm pretty sure Amazon already had a pilot program where they just shipped you things based on what it thought you'd want. The way it worked is you could keep the stuff you wanted but then just seal the box back up and ship back what you didn't want.

Expand full comment