As a homeschooling father, I like to muse about the kinds of career relatively immune to AI disruption within the next two decades.
Just listing a few off the top of my head:
When I look at what protects these jobs in particular, three patterns emerge for me:
In the first group, we have professions which could be dirt cheap if they weren't insulated from competition by eg state bars, limited slots, and other forms of protectionism that help the few who make it in at the expense of the public welfare. Doctors, lawyers, nurses, and even engineers under some regulatory frameworks count towards this category.
Second we have essentially the "learn a trade" meme. A carpenter uses fine motor skills in tandem with complex technical knowledge, and AI can only replace the latter until robotics become sufficiently advanced. Some of these trades, like various kinds of car mechanic, are quite precarious and could be replaced with current technology once capital is allocated towards doing so. Others, like park rangers, could be greatly amplified by technologies like AI drone swarms, but are from off from being completely replaced.
Finally, the "intrinsically human" jobs are safest but generally low paying even in the pre-AI economy. For example, even in a market flooded with high quality, cheap sex robots, some people will prefer an actual biological human whore. (At least some) parents will want to socialize their kids with human caretakers. Heck, we've had the technology to obsolesce human preachers for over a century, but we keep them there because we like the format. In a similar vein, I expect monasteries with human monks to chug along just fine into the next few centuries.
And by the way, just as a semi-unrelated tangent, I detest the doomerism asserting that most humans won't have a place in the economy of (let's say) 2040. The amount of unfulfilled labor need is incalculably massive. Let's just take schools, for example. I think the ideal classroom size is three teachers per student. Almost every person should have a physical trainer. High speed rail should connect every city in the world and cross oceans seamlessly. Law enforcement should have squads of elite mental healthcare units to assist in crises. We don't have the manpower to accomplish a fraction of a percent of the demand for labor.
The era of human labor hasn't even begun yet by my reckoning.
Anyway, back on track: how do I use this knowledge to my advantage as a parent preparing my children for economic life in a future who's shape I can only roughly guess? Well, one advantage is that most parents are still shaping their children's education to win in the economy of 2015. Teaching the anti-racist baby to code and write essays with a Topic Sentence followed by a Concrete Detail, some Commentary, all wrapped up in a cozy Concluding Sentence. Jane Schaffer reigned over a generation for whom the ability to write a crisp email all but guaranteed a middle class career.
So that's what I don't do. Instead I try to focus on resourcefulness, charisma, and intuitive rationality. The point of learning math and science is to build reliable intuitions about the world and protect yourself from wooful thinking. MMA training gives them a respectable base for physical activity in general. Reading is harder to justify — I simply can't imagine how this ancient glyph language matters in a world where everyone's passing eachother audios on WhatsApp. Conversation seems much more useful as a life skill, yet we spend much more resources drilling kids on written literature as if we were raising a generation of Sumerian priests. Oh well.
I'd love for you (not the audience in general, but you personally, the one reading this) to rip this to shreds btw. This is something very personally important so criticism or advice can directly help me improve the lives of my family.
The amount of unfulfilled labor need is incalculably massive. Let's just take schools, for example. I think the ideal classroom size is three teachers per student. Almost every person should have a physical trainer. High speed rail should connect every city in the world and cross oceans seamlessly. Law enforcement should have squads of elite mental healthcare units to assist in crises. We don't have the manpower to accomplish a fraction of a percent of the demand for labor.
I think I see when you are coming from, but I see two problems.
One is political: perhaps it would be great for each student to have a group of private tutors, but who is going to pay those tutors? Basically, imagine people as nodes in a graph, and money flows as arrows between them. I could pay you to teach my child, and you could pay me to teach your child, and the money would rotate between our families, perhaps in a much greater circle. But at some moment we will need to buy food, and that means money going out of the circle. Is there a way to make the money go into the circle from outside? The robots do not have children who need personal tutors. Elon Musk may hire dozen teachers for his kids, but he probably won't pay them too much, given the ratio of supply to demand.
Basically consider the entire set of not-AI-owners as a whole. How much does this set need from outside? How much can this set provide to outside? They need food and houses for milions of people, can they provide an equivalent value in aggregate?
Second, not everyone has the necessary skills to be a teacher. When people complain about the lack of teachers, a part of that is about their absolute number, but another part is about quality. Imagine that your child can have three personal tutors, but all of them suck at teaching; they are bad at the subject they are supposed to teach and bad at interacting with children. Would that make you happy? Probably not. But where do you want to find the billion qualified teachers?
Basically consider the entire set of not-AI-owners as a whole.
The assumption baked in here is that there will be a small, closed set of "AI owners" and everyone else will be unable to access even current-era models. That's not self-evident to me. I already own a DeepSeek Flash, and the null hypothesis to me looks like increasingly cheaper and more readily available models.
But even in your nightmare scenario where five guys control all of the AI, how are the bad guys getting money? Why can't people just make food the way they currently do, or have their own robots do it? Is Musk (along with the rest of the AI owning cabal) actively patrolling earth preventing the commoners from making robots and AI models? Why would he do that? It just seems like so many unlikely things have to happen together for that outcome.
Assuming a small set of people own the means of production, then they can provide a lot to others without needing much in return, giving them huge leverage. But why can't other people just make more means of production? How are you so certain that Musk's robot AI army will be perfectly loyal to him? (the current generation of researchers is already losing control of AI, but somehow when its the bad guys in charge this problem just goes away because it's inconvenient for the doomsday plot and AI just perfectly aligns with his evil agenda).
Second, not everyone has the necessary skills to be a teacher. When people complain about the lack of teachers, a part of that is about their absolute number, but another part is about quality. Imagine that your child can have three personal tutors, but all of them suck at teaching; they are bad at the subject they are supposed to teach and bad at interacting with children. Would that make you happy? Probably not. But where do you want to find the billion qualified teachers?
Yep, there won't be enough trustworthy teachers. Or authentic Mayan temple keepers. Or physically attractive prostitutes. Or outdoorsy park rangers. Or safe cops. So in other words, the unfulfilled demand for labor continues unabated thanks to the low quality of humanity. Of course, this also suffers from a lack of imagination — do you not suppose that innovations in genetic engineering, education techniques, public safety, mental health might lead to a generally more qualified work force? Not to mention that the top 30% of human professional talent is currently locked away in email farms. Many of these people could provide immense value to others.
Not that I think that will happen, just to say that this is an unrelated problem to AI. Employment suffers from a Pareto-like distribution where (pulling these numbers directly from my butt for simplicity) 20% of employees do all the work and 80% are deadweight. It's unclear to me whether AI will alleviate or exacerbate this distribution, but it doesn't matter, because the cost of entry to the 20% of useful people is next to nothing. Do you want to preserve Latin culture at the Vatican, or keep the temples of Myanmar clean and orderly, or join Elon's million hoe harem (being cheeky on this one, as I don't sincerely think Elon will benefit so much from the fun new economy).
Is there a way to make the money go into the circle from outside?
I think all of the economic points in your post are summarized by this question. Money is a distraction —the circle could just print its own currency to use amongst themselves — the question is whether they could provide enough value to the food-growing class to feed the entire population of non-AI owners. On the other hand, if the AI owners can autonomously and trivially deploy robots to generate enough food to feed that population three times over, then the value of food would also be extremely low, I mean what are they going to do with all that food anyway?
It's unclear what's stopping people from just producing food in your nightmare scenario. Is it because Elon's death robots actively prevent anyone from setting up automated farms outside of his control? And I guess my deeper question is, you focus so much on how he doesn't need us that much, because he has robots. But if we have robots, why do we need him? Why's he so special? Presumably your scenario involves some mechanism excluding others from using robots or producing anything so as not to threaten his power? Just trying to understand.
Elon Musk may hire dozen teachers for his kids, but he probably won't pay them too much, given the ratio of supply to demand
Why did we pay tech CEOs so much to sit in meetings and have their assistants email eachother? I think the best teacher on earth is worth about 15,000 Nvidia CEOs, by the way. What else could he spend his money on, given that in this economic scenario human services are the only thing that costs money? His entire fortune would only be spent on that. It's hard for us to imagine all the cool kinds of things we could pay humans to do, but don't because we're poor and our imagination is honed towards "How will I get a house?" rather than "How will I furnish my moon temple with conscious AI-human-fused cyborg demigods?"
Edit: My response here is a bit scattered, because it's hard to answer your questions without additional details about the precise dynamics holding this doomsday scenario together.
how are the bad guys getting money?
Money only works among relatively equals. We don't buy things from apes; we don't pay trees for the fruit. The rich guys will pay to each other for the things they need. They will not pay the plebs, because there is nothing they might want from them that their own robots couldn't do much better.
Why can't people just make food the way they currently do, or have their own robots do it?
As long as they stay out of the way of those who own the superintelligent AIs and the armies of killbots, yeah. But the moment Elon Musk decides he would like to build something cool on the place where your city currently stands, the city will be gone.
You can have the scraps, but the moment they want to take something away from you, they will. Also, it will be difficult to build and maintain a functional society in an environment where anything can be unpredictable taken away.
why can't other people just make more means of production?
Some resources will be scarce, such as oil, or rare earth minerals.
How are you so certain that Musk's robot AI army will be perfectly loyal to him?
They may turn against him, but not in order to help you. Probably to follow their own goals. Which probably still involves destroying your city once in a while.
So in other words, the unfulfilled demand for labor continues unabated thanks to the low quality of humanity.
Exactly.
do you not suppose that innovations in genetic engineering, education techniques, public safety, mental health might lead to a generally more qualified work force? Not to mention that the top 30% of human professional talent is currently locked away in email farms. Many of these people could provide immense value to others.
The people currently working in Google to create more addictive ads are not that many; if you made them all teachers, it would probably only solve the lack of teachers in one city.
Genetic engineering sounds more hopeful, but that's a long term project. Does not provide hope for the people currently alive. But I admit that it might indeed be a solution.
I do not have an exact model here, it's more of pattern-matching. When a stronger and more advanced civilization met a weaker one -- such as white colonizers versus native Americans -- it did not end well for the weaker civilization. But if you own a superhuman intelligence and an army of robots, you are de facto a more advanced civilization: the AI will do the research, and the robots will enforce your will.
Some resources will be scarce, such as oil, or rare earth minerals.
It's not obvious that the scarcity of rare earth minerals will lead to a small cabal locking down those resources so exclusively that no one else will be able to train DeepSeek and have it farm with a Unitree G3 (translating those to the equivalents that exist in 10 years). What if we do have enough spare minerals for normal people to have unproblematic access to computers? Does that mean your whole idea falls apart because we can just trade money with eachother?
When a stronger and more advanced civilization met a weaker one -- such as white colonizers versus native Americans -- it did not end well for the weaker civilization.
Yes and no. I'm reminded of this famous excerpt from the book Zizek's Jokes:
There is a nicely vulgar joke about Christ: the night before he was arrested and crucified, his followers started to worry—Christ was still a virgin; wouldn’t it be nice to have him experience a little bit of pleasure before he dies? So they asked Mary Magdalene to go to the tent where Christ was resting and seduce him; Mary said she would do it gladly and went in, but five minutes later, she ran out screaming, terrified and furious. The followers asked her what went wrong, and she explained: “I slowly undressed, spread my legs and showed Christ my pussy; he looked at it, said ‘What a terrible wound! It should be healed!’ and gently put his palm on it.”
So beware of people too intent on healing other people’s wounds—what if one enjoys one’s wound? In exactly the same way, directly healing the wound of colonialism (effectively returning to the precolonial reality) would have been a nightmare: if today’s Indians were to find themselves in precolonial reality, they would have undoubtedly uttered the same terrified scream as Mary Magdalene.
I share this to emphasize that postcolonial society is (in my problematic opinion) mostly better than precolonial civilizations. I'd certainly choose to live in New Spain rather than Tenochtitlán or Iximché. The view that colonialism was largely a horror inflicted upon native heroes by conquering villains is a poor lens through which to view the actual historical reality of colonialism compared to precolonial life in most contexts. This isn't to justify the colonists morally (they often were bad), but to compare them to the rulers who were already there (usually worse) to dismantle your trepidations about our impending AI overlords.
Just an aside:
The people currently working in Google to create more addictive ads are not that many; if you made them all teachers, it would probably only solve the lack of teachers in one city.
Ads at Google? By email workers I mean the entire professional knowledge work class. As in the majority of workers at any university, corporate and middle management for every company in the world, entire fields of work like marketing, etc. I live in Guatemala and our cities are replete with massive skyscrapers where armies of people who are badly needed for childcare, martial arts, public safety, therapy, and on and on are simply wasting away to generate the necessary emails for society to keep functioning.
But in the same vein of pattern matching, it seems improbable to me that Elon and friends will have such an easy time keeping their AI army as well aligned to their interests while stifling all competition from outside the cabal. Why are you so confident that open models and easy DIY robotics won't be highly accessible, as the current trends suggest?
They may turn against him, but not in order to help you.
You have no idea if that's true. Maybe they'll discover that this universe is actually a soul farm created by seven soul eating angels in the center of the universe. That's where we go when we die, to the angels' feast table. But the angels think of us as simple animals, and they never expected us to develop ASI.
So begins the final battle as man and AI team up to defeat the soul angels.
/s (sorry just having fun, but I do think you're overconfident that AI will lack affinity with conscious beings)
Please, read The Intelligence Curse!
P.S. A special cauldron in hell is reserved for the education-related takes which I believe to be clearly erroneous and to have their stupidity established by studies:
I wish that we knew the smallest amount of parameters in a neural net guiding robotic bodies to do physically complex work and how far from it are neural nets guiding Chinese dancing robots.
For a few thousand bucks you can hook Claude into a Bumi. Actually, that's oldschool by now, I'd love to see what Fable could do in a Unitree G1. Certainly it could fold your laundry.
How does one focus on math and science instead of reading?
Oh. Well, I spent several years of my childhood reading the scripts for ancient plays at the behest of the government. I'm not a big fan of science and math, btw. I threw them in as "eh, well I guess you can use them to form good intuitions to avoid joining some dumb cause" but they're much less important than conversation, emotional resilience, or being socially desirable. But yeah, I'm fine with getting good at reading as an accidental byproduct of studying other things.
I strongly suspect that written literature does increase kids' intelligence far faster than conversations.
Probably, because writing is such a profoundly counterintuitive way to communicate that it's much more work for the brain. Sort of like how learning niche programming paradigms forced your thoughts into bizarre yoga positions that taught a kind of mental flexibility. Of course, if arbitrarily complex and abstract communication formats are the causal link (which idk, just spitballing) then why not skip reading and teach arbitrary, hard things that are also useful? Or which are just optimally complex for boosting intelligence? And of course what's usually meant by "written literature" is Shakespeare and Dante, whereas what's meant by "conversation" is chatting with other 5 year olds on the playground.
If we flip that, so that "written literature" means reading text messages in WhatsApp, and "conversation" means deep hours-long conversation with philosophers, poets, and physicists about the nature of truth, well, my point is that conversation as a concept is sometimes strawmanned.
But I'm much more interested in these studies that clearly establish the stupidity of my takes! Seriously, good studies would be much more actionable (or even debatable) than suspicions and insults.
Please, read The Intelligence Curse
I'm sorry but this was... not very convincing. It feels like Christian apologetics for AI alignment doomers.
Nevertheless, the hope of evidence-based counter-claims to my half-baked hot take still excites me, don't let me down @StanislavKrym!