This seems to be the author's main point:
"Recall that the advent of pre-AI computing made the expert judgment of professional decision- makers more consequential and more valuable by speeding the task of acquiring and organizing information. Simultaneously, computerization devalued and displaced the procedural expertise that was the stock-in-trade of many middle-skill workers. But imagine a technology that could invert this process: what would it look like? It would support and supplement judgment, thus enabling a larger set of non-elite workers to engage in high-stakes decision-making. It would simultaneously temper the monopoly power that doctors hold over medical care, lawyers over document production, software engineers over computer code, professors over undergraduate education, etc."
"Artificial Intelligence is this inversion technology. By providing decision support in the form of real-time guidance and guardrails, AI could enable a larger set of workers possessing complementary knowledge to perform some of the higher-stakes decision-making tasks currently arrogated to elite experts like doctors, lawyers, coders and educators. This would improve the quality of jobs for workers without college degrees, moderate earnings inequality, and — akin to what the Industrial Revolution did for consumer goods — lower the cost of key services such as healthcare, education and legal expertise."
"Moderate earnings inequality" means "fewer high-paying jobs", as someone pointed out in another comment. From where does the pressure come to raise incomes across the board? That's what unions were for. In the US, the unions were crushed. The whole idea of paying people more than they are worth as an economic unit, a key goal of the union movement, is almost forgotten. Yet that's what this paper assumes will happen. Somehow.
This is "trickle-down" economics with an "AI" label pasted on it.
The author has better papers. His "Why are there still so many jobs" (2014) [1] is worth a read. He makes predictions one can now check.
[1] https://economics.mit.edu/sites/default/files/publications/w...
The problem with the concept of a middle class job (one where you can comfortably afford living, fun, saving, etc I think is most people's interpretations), is that it represents an unstable equilibrium state. Incentives of most businesses are the opposite of someone in the middle class. Businesses don't think in terms how they can take so little such as to perpetuate some idyllic lifestyle unimpinged, they think in terms of how much can they possibly take off the table for our own gain without bleeding the beast dry. This is why over time, middle class lifestyles get clawed away into more desperate paycheck to paycheck ones. So even if AI used well restores some middle class jobs tomorrow, what about next week lets say when companies realize they can charge more for goods and services, and people are back to paycheck to paycheck? Short of a planned economy with price controls I don't know how you prevent these incentives from emerging.
This person assumes AI is going to stay in its current form: a technology that requires an expert operator to help accomplish tasks.
Most people are afraid of “next-level” AI, where no operator is required. If trucks can be moved without anyone at the wheel, truck driving is a job that won’t exist anymore. As simple as that.
Whether trucks can be 100% autonomous remains to see.
Productivity has been gaining for years while the number of working hours per week have remained the same and wages have stagnated or even declined. Why should I believe more productivity on the lower end will restore the middle class?
> Because of AI’s capacity to weave information and rules with acquired experience to support decision-making, it can be applied to enable a larger set of workers possessing complementary knowledge to perform some of the higher-stakes decision-making tasks that are currently arrogated to elite experts, e.g., medical care to doctors, document production to lawyers, software coding to computer engineers, and undergraduate education to professors.
Note that in this analysis, if you are a software engineer, you are not the "middle class" as most signs point to. Rather you are a member of an elite that the middle class must wrest power from.
The identity of the middle class is another question. You might assume that it's made of the people you know, but more likely it's underpaid workers from developing countries.
If AI truly does take over most jobs of today (low to mid tier level SWEs replaced with a couple of “prompt engineers”). I would be okay, but only if there were proper support pillars in place.
Healthcare should be free for all. UBI. And of course continued education provided to train workforce into something else.
It's an interesting point, but in our society and current economic system I have a hard time imagining it will be anything so different as to what we have now to "restore the middle class".
We have people and companies with so much power that it rivals governments, and in the US we also have a large segment of the population who has been conditioned to believe that the government is bad and having less of it will make it better, but it seems to me like better government that actually supports the needs/desires of the populace at large is required to make our society better for everyone.
Autor was just on the Odd Lots podcast, so if you don't feel like reading the paper you can listen to the interview where he talks about it.
Everyone seems to be asking how AI is going to fit in the current paradigm - creating or destroying jobs and so forth - but I don't see a ton of folks talking about something I feel is patently obvious; that the most relevant applications of this stuff are going to exist outside the "company hire people to make product / give service and sell service" cycle.
Yes, AI will lower the barrier for entry level positions (IMO a good thing; more chances for smart creative people to get in the game who may have been failed by more traditional systems). But it will also allow the creation of projects by much smaller teams, who no longer need to stress the details and have a focus on making a thing just to try it out.
Individuals will be able to get much more experimental and iterate way faster; if the amount of investment to build an MVP is less than the investment to research and assess all that crap that usually has to go into a commercial product (market fit, funding, marketing, sales, etc.) then you can just make the damn thing and see if the pasta sticks to the wall. This could be a really great environment for testing out weird ideas, some of which might be completely novel and break the mold in ways that are simply not feasible if you have to manage all that other stuff.
I'm also excited to see what new stuff pops into the OSS space as a result of devs-with-jobs being able to hack out a proof of concept in a couple hours and smacking some stuff together over weekends.
Yes, let's hope that AI won't remove the need for drudgery!
I knew a very comfortably well-off lady who made a fine point of how she never used the automatic machines in the supermarket, because she wanted to ensure that those poor cashiers didn't lose their jobs.
How noble!
One new job which I believe will be high paying and lot of people will be able to do, in the post-AI world => Content verification and moderation. With so many bots and fake content, I don't want to waste my time on low quality stuff. I want other humans to verify it, test it, validate it etc. and then only it should see my eyes. Those platforms which will employ this approach will have better content coming to timelines and will attract more users.
Can, doesn't mean it will.
Sorry, I've become very pessimistic these days. If something can be used to permanantly reduce FTEs, it will be done.
Progressive social policies, better voter representation, and strong unions can restore middle class jobs.
It won't, and that's a good thing. Workplace deaths will plummet, people will have more free time, and the environment will improve from less people commuting. I don't see any reason why replacing certain jobs with AI is bad, so long as this is paired with UBI.
The most charitable thing I can say about this is that it’s likely wrong.
> My thesis is not a forecast but an argument about what is possible: AI, if used well, can assist with restoring the middle-skill, middle-class heart of the US labor market that has been hollowed out by automation and globalization.
> The unique opportunity that AI offers humanity is to push back against the process started by computerization
Yes. More automation and computers will clearly solve the problems created by automation and computers.
I can't find where he probably says something like "outsourcing of these jobs will also improve middle class outcomes" but I assume it's in there through some kind of weasel-wording.
Ted Chiang put it best - "most fears about A.I. are best understood as fears about capitalism."
https://kottke.org/21/04/ted-chiang-fears-of-technology-are-...
So long as the dominant theology in this country is extractive capitalism, nothing that tilts the cost balance towards capital and away from labor will be good for the middle class or the average person.
Editorialised title.
Actual title of PDF: "Applying AI to Rebuild Middle Class Jobs"
All of this analysis is based on inaccurate assumptions. Primarily that some rules of supply and demand will dictate how the tech is deployed.
Let's look at history. The industrial revolution was predicted to create copious leasure time for the working class. How did that turn out?
The steam engine, the cotton gin, the automobile, the computer...
Every single technological advance that was predicted to improve the lives of the working class due to massive increases in productivity, was used instead to almost exclusivly bennefit ownership.
This is exactly what will happen with current LLM tech (please don't call the bullshit generators "AI")
If the tech cannot be deployed so as to exclusively bennefit ownership, it won't be deployed at all...
"air traffic controllers were paid a median annual salary of $132,250 in 2022, or nearly four times the $33,380 median annual pay of crossing guards"
It's not quite "expertise" that justifies the wage gap, it's supply and demand. Almost anyone can be a crossing guard--not everyone can be an air traffic controller. If AI theoretically could make anyone an air traffic controller, one would expect the salary to collapse as well.
Additionally, the notion of a middle class relies on a wage differential. If AI levels the playing field so dramatically, the notion of middle class will entirely evaporate since everyone's purchasing power equalizes.