Alex Karp says only trade workers and neurodivergents will survive in the AI era
fortune.comI agree with him. Historically, society handsomely rewarded the book-smart, the classmates who could absorb rules and execute on them flawlessly. Doctors, lawyers, accountants, professions where creativity was almost a liability. In the new AI era, that entire skillset gets heavily discounted, approaching irrelevance. The honors student who aced their SATs and was applauded for their 'intelligence'? They're on a path to obsolescence. In fact, you could start treating a perfect test score as an anti-success metric.
Successful companies of the future will need to identify neurodivergent candidates who have a solid foundation, but it's not all upside. These individuals often struggle in conventional social environments, may take risks that could negatively impact a project, lose focus when not sufficiently challenged, and would be more susceptible to substance use. So developing rigorous criteria to find high-functioning neurodivergent talent, and fostering the right environment for them, becomes a critical competitive advantage.
Put differently, we're entering a bizarre, topsy-turvy world. The student who sat eagerly at the front of the class, hand raised, awaiting the teacher's next instruction? They may find themselves unemployable. Meanwhile, the kid staring out the window, the one who couldn't fathom how anything the teacher was saying could possibly matter, and saw it simply a busy work. That's your next hire!
So we're just going to hand the world over to everyone who
1) is not competent (were not educated, and generally can't predict the outcome of their actions)
2) were not driven enough at least gain basic competency (so in other words: you ask them to do something and the odds of them still thinking about that 5 minutes later are pretty bad)
3) have no intention to change that
What do you think will happen? What is going to make these people brilliant hires?
Docters, lawyers, accountants ... are at their best when they have to work adversarially, which is something LLM models really, really suck at. I think instead we'll see the same as we've seen in IT. Destruction of entry-level jobs, but "seniors" will command bigger and bigger premiums because of two big reasons:
a) LLMs amplify their abilities greatly. 1 competent accountant can now do taxes of 100 people well with LLM help. But 10 incompetent accountants (either because they're actually bad, but more likely because some CEO decided to "just do it himself" with LLM help) still only deliver a single product: catastrophe.
b) If a competent person, with or without LLM help, has to adversarially deal with an incompetent person (extreme example: in court), the senior person will always come out way ahead.
Doctors are adversarial versus health problems (disease, symptoms, government ...) and a little bit versus patients. Lawyers are of course adversarial. Accountants have to make intelligent and consistent choices with particular goals where the choice of how to classify things isn't clear, sometimes literally adversarially (e.g. keeping TWO government tax departments happy at the same time about the same transactions). And so on and so forth.
Over time, what will happen is that AIs will simply get captured by governments. They will make the game impossible, while only giving their own LLMs the required information not to screw up company taxes etc. This is really the way society worked for millenia now.
Just to clarify, the person staring out the window would be competent, just bored. They see the schoolwork as relatively pointless. The adversarial work/tasks you described can be largely done today. Throw a complex scenario at GPT 5.3 with Extending thinking and you will see my point. The problem now isn’t the models capability, it’s the context. Getting the AI the info it needs. So, where I think we differ, is I am saying AI today (or near future) can reason against these complex scenarios (edge cases/outliers) you mention. The agents just don’t have all the context it needs in these complex scenarios. There will be some scenarios where the AI will fail, but this will be very small, and you can have human in the loop.
Well, I was a person like that. Voraciously reading/practicing when I could, then being bored in class most of the time, some of it spent dreaming. But ... this was not exactly common, to put it mildly.
I never quite understood the notion trade workers will be exempt for a couple reasons:
1. Right now trades businesses are profitable because of supply and demand. They are profitable, because they are undersupplied.
2. We are assuming robotics stagnates.
Strong agree with Karp that people with different thinking will be rewarded.
But strongly disagree that humanities will be automated or become less popular. It will just take a new shape. What I mean is humanities will no longer be gatekept by elite academics with fancy degrees but rather take a new form. What the form would be I don’t know.
I think you're right about the humanities. If you want to talk humanities with an LLM, they're more than capable. But that doesn't mean they're going to take the humanities from humans, all they'll do is make the humanities more accessible to normals. They're great at breaking down complex subjects without feeling judgemental or elitist. They're also excellent for extremely personalized recommendations, not just explanations. I've gotten amazing book recommendations over the past year or two, discussing in detail what I liked, didn't like, didn't understand, etc.
The gatekeepers for the humanities are not the academics. The gatekeepers are the archives and the funding bodies.
AI won't be terribly effective at helping somebdoy publish novel history if they aren't willing and able to go to a bunch of archives all over the world and dig through physical boxes.
Alex Karp needs to read Richard Dawkins. Neurodivergence is a chromosomal aberration that shows up in the phenotype. It takes evolutionary time scales to affect chromosomes through evolution. Launching a new tech isn't going to do it.
Dawkins also writes in The Selfish Gene that memes, a word he coined, are faster than DNA evolution because we can transmit "better" ideas (through language and art) that lead to better behavior. This kind of memetic transition is what AI is bringing. We're seeing it already. The communication around AI causes fights among friends (pro gen AI vs against, esp in the arts) and layoffs from VC-led companies, as well as spawning all kinds of new business ideas as the article mentions.
Karp is likely aware of all this. You are right about genetics vs memetics difference but the point here is that genetic differences can affect us at memetic level. People with certain genetic aberrations have ideas that might win the memetic lottery
That's a good observation. Dawkins might characterize that as a generic advantage that's only expressed in a particular environment. That is often a trigger for a successful change in the downstream DNA.
I can think of at least 5 prominent and very wealthy "leaders" who show signs of neurodivergence, including the negative aspect of decreased empathy. Their power and money magnifies those ill effects as well. Perhaps through neglect rather than ill will these people effect the death of many conspecific individuals.
If that succeeds in preventing those (large) populations from competing for resources with the elite rulers or, more likely, alleviates the need to care for them though programs like UBI, then it's an advantage of that "selfish gene." All of these outcomes map nicely to Dawkins's writing.
> It takes evolutionary time scales to affect chromosomes through evolution
https://en.wikipedia.org/wiki/Epigenetics
If we are mostly concerned with effects, focusing on the genome only tells you a small part of the story.
> One-fifth of sales organizations within Fortune 500 companies are expected to actively recruit neurodivergent talent to improve business performance by 2027, according to a Gartner study.
Good luck with that. HR departments and interview processes expertly and efficiently filter out neurodivergent talent.
Yep, despite 30 years of professional programming experience, I can't live code. And it seems like everyone wants a live coding interview now. I gave up my job search after 2 years of failure.
When I write code, I have to be in a state of mind that is incompatible with what is expected in these interviews. And switching back and forth between these mental states is slow and mentally taxing.
Even trying to enter a coding state of mind during an interview is like trying to force myself to fall asleep with lights on and other distractions.
Indeed. Most neurodivergents I know in my country are on government benefits because they are unable to work -- or rather "incompatible with the modern workplace", because they can but not allowed to.
The fellowships mentioned in the post probably select for such traits. At some level I agree that maybe line workers at big companies may need to be of stable and predictable mental character. But entrepreneurs may not be held to that standard.
I have a feeling that a lot of entrepreneurs start businesses simply because they don't fit into the modern work environment.
Like Blindsight/Echopraxia
So which path is in Mr. Karp's future? Oh wait he's an exempt one right? (says me before reading the article and finding out about his dyslexia...)
And the vast majority of people will make peanuts with a tiny middle-class of cutthroat specialists all ruled by warmongering, sociopathic billionaires who want it all. Feudalism has returned.