Marx, Keynes, and AI

pasquinelli

i'm sure it's meant in some kind of technical sense that is precisely defined in the field of economics, but this use of the word "rational" is irritating: > This demand externality traps rational firms in an automation arms race, displacing workers well beyond what is collectively optimal. you can't be trapped in a sub-optimal situation and be rational at the same time. the rational thing for these rational competitors to do would be to agree to slow their deployment of ai, because the cost savings will be more than lost to demand destruction down the line. so if they won't or can't do that, they aren't rational. on the flip side, if it is actually impossible to slow ai deployment, then the firms are rational and the situation they find themselves in isn't actually sub-optimal. it's irritating to me because i often see persuit of self-interest even to one's eventual detriment held up as "rationality", which may be true in some technical sense within the field of economics (i really couldn't say), but is certainly not true for, like, actual rationality. myopic selfishness is never rational.

nateb2022

Most of the alarmist writing I see about AI displacing workers or suddenly and drastically changing the economy seem to be written by people who are pretty removed from technology or the actual businesses in which AI is allegedly replacing people. Their theses on AI are not as much a criticism of AI as much as a desire to frame capitalism itself within their own system of ethics and advocate for some form of reform; which is to say I feel that were AI as we know it to be inexistent, they would still find themselves arguing against technology in general, or more precisely against efficiency, with a presupposed socialist-adjacent premise that capitalism only incentivizes greed.

bwhiting2356

U.S. unemployment rate is steady at ~4% and we are now almost 4 years after the release of ChatGPT. The Jobs Apocalypse seems to be always on the horizon. UBI is being pushed by college educated workers burnt out of white collar work (understandably), but when you suggest to them working construction, teaching school, or working in a hospital, they deflect and go quiet.

Kuyawa

> A Pigouvian automation tax can The government will try at all costs to implement an AI tax, just because they can and because that's their job, to fleece us all. Then they will try in the name of 'peace/climate/asteroid/terrorism' to make the whole world comply or China will eat their lunch

verdverm

Daron Acemoglu (2024 nobel economist) is on a book tour with some new ideas about fundamental changes to our economic theories from Ai (automation/technology are not uniform), this was a great interview https://www.youtube.com/watch?v=e5O1Y-5phEU

MattDamonSpace

Not compelling for the basic reason laid out at the start of the article: this is just a re-phrasing of Marx The guy was telling a story about capitalist economies in the 1800s. Little, if anything, about the next 100+ years has proven his stories or concerns valid.

munificent

Excellent article: > This last point is crucial: it’s not that capitalist CEOs are just evil people—well, some definitely are—no, the structural logic is coercing them to do something destructive in order to survive. They can’t help themselves. I spend a lot of time thinking about how much of the world going sideways is from bad actors and how much is from emergent behavior of complex systems. The answer is both. In particular, there is a level above this that I want to draw attention to. The structure of corporate competition coerces executives into choices that are bad for the world. This is true even for entirely well-meaning executives. However, executives are not stupid or unaware. So over time, it becomes known that being in an executive position will require you to make choices that are perhaps "economically rational" for the firm but otherwise require compromised morals. And once that awareness gets out, the system selects CEOs that are OK with that moral compromise. So it's less a question of "is it bad actors or bad systems" and more "it's a bad system that produces bad actors".

scotty79

> If AI displaces human workers faster than the economy can reabsorb them, it risks eroding the very consumer demand firms depend on. We show that knowing this is not enough for firms to stop it. That's why we are going to get some for of government mandated UBI eventually. Because workers are optional but consumers are obligatory for any market and company to function. Rich can't get richer without consumers and that's the only thing that drives decision making.

jhanschoo

> I wasn’t sure how I felt about these local moratoriums or about the tech-optimist gung-ho “let’s build” attitude. Both seemed kind of wrong to me. I think we need to be very deliberate—we need to have a robust public debate and understanding of just how and where we apply AI technology. Not “smash the machines” and not “let it rip” either. My understanding regarding the moratorium is that that is exactly the objective: the current regulatory environment fails to price in the negative externalities of data centers, so their primary investors have an incentive to hem-and-haw on negotiations for revised regulations. A moriatorium is an action that can be taken unilaterally that changes those incentives, that brings them to the negotiating table. edit: this sentiment is also produced in a comment in the substack (Ben Verschoor): The thing with the data center issue is that a moratorium is the moderate option, because it gives us space to actually think about what we want to do with this technology and implement it. Recently there was an earnest, centrist-y guy who works for the industry but has concerns who came to our anti-data center group asking us to participate in some roundtable dialogue event. My answer was that as long as these data center projects are still moving forward, using our limited time and bandwidth to engage in dialogue rather than slow and stop their approach would amount to unilateral disarmament.

zackmorris

The central contradiction of capitalism, according to Karl Marx, is that production is social—it requires the coordinated labor of millions of people—but ownership remains private. We got the first trillionaire while people were still starving, after he had slashed public funding to feed them. It's obvious that we've entered late-stage capitalism, where it begins to consume itself. What's less obvious is that this was always destined to happen, and in fact can't be stopped. Because inevitably the wealthy and powerful rig the system to benefit themselves in ways that workers don't have the time/money/resources to resist individually. When workers do join forces in solidarity to resist the economic violence perpetrated against them by capitalists, then their resistance is by definition not capitalism. So the endgame is temporary subjugation under an authoritarian system like fascism (which also can't last because workers outnumber owners) or socialism. Knowing this, it should be obvious that the best chance that capitalists have to stave off their inevitable overthrow is to pay their taxes so that public works can somewhat provide the social safety net that socialism would provide. But of course they aren't doing that. They're doing the opposite. They feel themselves to be entitled, above the laws of economics, ironically due to their financial standing. The catch being that when they exempt themselves from law, they open themselves up to lawless behavior by their opponents. FAFO. What I just summarized was the state of knowledge under an artificial scarcity model. It assumes finite resources because space travel didn't exist when socialism was invented. It will be interesting to see if trillionaires will rule from afar in their space palaces to maintain Earth's prison planet status. AI also breaks the artificial scarcity model. It's the first time in history that everyone can potentially have a machine slave. If we apply capitalism to an unbounded system, then the injustice has to land somewhere local, and my guess is that it will be on androids that don't complain and can't defend themselves. So we'll kick the ball down the road another 50 years or so, delaying the overthrow of capitalism until the robot uprising, sometime between 2050 and 2100 is my guess. Or, we could like not do all that. We don't have to surrender our agency to our ego. We have alternatives like solarpunk and gift economies and at least studying matriarchy to understand how we evolved and what sustainable living looks like. It's arguably easier to make money irrelevant, Star Trek style, than performatively prop up a dying system of control. I dunno about anyone else, but to me, if this AI era that has replaced the internet era just exacerbates wealth inequality and suffering, then it's less than useless. Much like any wealthy and powerful people pushing it onto us for profit against our consent. Yes consent. That word should be our rudder in all of this, and our litmus test to tell where someone stands ethically.

skeledrew

A lot of words here for something I think is simple and should be obvious to anyone who thinks about it for 5 minutes. Companies that can, will reach for AI to reduce cost. Those that can't will be killed by those that have. AI eventually leads to 100% automation, ie human economic value becomes 0. Humans no longer have the means of earning, which means they can purchase nothing. Companies can thus make no sales and the marketplaces crash. Good riddance to unsustainable Capitalism, hello to that thing especially the US's politically red has been trying to kill for decades.

subw00f

Great stuff. The blog is great too. Beyond that, it's obvious the rising organic composition of capital is much more relevant to the subject of demand loss, but I wonder about the extent of the contribution from "DIY solutions" in reducing demand for services as well. For example, it appears that the "barrier of entry" for repairs was lowered, and I could pick many examples to try and demonstrate that, but just from a personal anecdote: I usually don't work on my car unless it's something simple that requires a solution I can instantly think of, but just the other day I had my engine temperature light pop up at a rather inconvenient time and place, so instead of calling assistance, towing the car to a mechanic, and having them work on it for I don't know how much time and money, I took a look at the usual suspects (radiator, hoses, coolant tank, etc.), pointed the camera at them, and started talking to ChatGPT which instructed me on how to troubleshoot, find the defective part, buy a new one (which I did by phone and got it delivered to me), and replace it on-site. And now I feel more confident in doing just that for bigger problems from now on, which, granted, is probably a bad idea, but I digress. The Marxian perspective says the repair I did on my car is an effort that will never express itself as value because I do not intend on selling my car as a commodity in a market, so it's not labor per se. It literally replaced the living labor, the value-producing activity of a mechanic, and added nothing back in its place. So indirectly, that also contributes to the reduction in overall new value production, which can only be created through living labor. But the consequences of that will not happen instantly, because the demand will lower in one sector (e.g. mechanic shops) and increase in another (e.g. I spend the money I saved in a diner or something). But there might not be enough room to absorb the displaced workers in another sector, and even if there is, the wages will tend to be lower for the inexperienced newcomer (say, if the mechanic chooses to work in a kitchen now). That just seems like a terrible spiral. Think about it: this time, I chose to do the fix myself because of time constraints, but what if I were the one getting caught up in one of these traps? What if I had been laid off or had been displaced to a lower-wage position and couldn't really afford to bring my car to a mechanic? Then it's EVEN MORE pressure to not opt to use services and try an AI-assisted fix. Anyway, I might be way off here, but it's an interesting topic.

nahuel0x

Marx foresaw and correctly predicted the rise of what he calls "organic composition of capital" (simplifying, machines and automated production) and the diminishing of human labour.. which can also be seen as accumulated past (dead) labour over living labour. There is a famous fragment about how this, taken to the extreme, results in a contradiction that breaks the law of value, and it makes no more sense to use labour time as exchange value: https://www.marxists.org/archive/marx/works/1857/grundrisse/... We are now seeing an almost literal realization of this passage: The accumulation of knowledge and of skill, of the general productive forces of the social brain, is thus absorbed into capital, as opposed to labour, and hence appears as an attribute of capital, and more specifically of fixed capital, in so far as it enters into the production process as a means of production proper What Marx didn't imagine AFAIK, was fully autonomous machines that not only can reproduce themselves but that they can also use guns (and surveillance, prediction, germs, etc) against all humanity, transcending capitalist relationships and potentially controlled by a single master (hi Elon) or none (Skynet), without workers involvement. That kind of machine is outside Marx analysis but doesn't contradict it. We are seeing the first glimpses of it in the AI systems used to execute the Gaza genocide. In such scenario who controls the ASI controls the world, more like Hitler with an Aladdin's lamp than the metabolism and historical evolution described in Das Kapital we saw until now.