AI and the Lump of Labor Fallacy
In the history of humanity, there has never been a single technology capable of replacing human minds and bodies in all types of work.
Literally never.
Until now.
Sure, there have been numerous advances that have replaced some humans in some types of work, often while simultaneously enabling new types of work, but not a single thing capable of replacing all humans in all work.
For example, it is commonly cited that ~100 years ago most people in the US worked, directly or indirectly, in agriculture. Now that number is under 2%. That’s a big shift…and yet there are far more people living and working in the US today than there were back then.
The lump of labor fallacy is the (historically) flawed belief that the amount of work available in an economy is fixed, so if one group takes jobs, another must lose them. In reality, innovation, demand, and productivity continuously expand the job pool, creating new opportunities. While some jobs go away, and some people have to retrain, and some people won’t retrain and just go without work, on the whole it tends to shake out fine.
People who cite this fallacy like to point and laugh at the “this time it’s different” crowd, because historically it hasn’t been different, just more of the same.
Well, this time really, truly IS different, and here’s why.
As I’ve already stated, no technological invention in history has been capable of replacing all humans in all work, and in fact most technological advancements tend to unlock entirely new skill trees for people to pursue.
What’s different this time around, with AI and robotics specifically, is threefold:
No technology, in terms of speed of adoption or speed of improvement in capabilities, has EVER moved as fast as AI is moving now. And robotics, largely thanks to AI, is picking up the pace as well.
In just over 2 years, the capabilities of AI broadly speaking have gone from a child to a PhD. This is FAST. And, despite many naysayers, there are still no signs of this slowing down. If anything, it is still accelerating. There is zero reason to believe, if the pace continues, that AI won’t be vastly more intelligent than any human on earth, and maybe every human on earth, within another year or two. Five tops.
As AI capabilities advance, they simultaneously speed up the advancement of technology and knowledge acquisition in almost all other fields. But since AI will be smarter than any living human, and possibly smarter than all humans combined, why on earth would we want dumb humans doing things an AI can do better, faster, safer, and cheaper? And assuming robotics, and very likely nanotechnology, accelerates just as quickly thanks to these advances, there will very soon be *nothing* mental OR physical that a computer can’t do better, faster, safer, and cheaper than a human.
This time IS different, because we’ve created the first ever technology genuinely capable of replacing humans in all forms of work* (and by *work I mean “for money” work, what most people think of as work).
This tech WILL create more work to be done…it just won’t be humans doing it.
AND THIS IS A GOOD THING!
Can you, just for a moment, set aside any fear or anxiety, and think about a world that meets the following parameters:
Everyone has access to nearly limitless intelligence to help them solve any problem.
Energy is too cheap to meter.
Any disease can be cured. Death is optional. Human intelligence can be augmented to amazing levels.
You can have almost anything you want at basically no cost, because it’s possible, using a small amount of energy and compute, and a tiny amount of physical matter, to make you virtually anything.
By eliminating all the misaligned incentive stacks, virtually everything wrong with the world is sorted out.
You are free to spend your time on the things you love and enjoy. No more grinding for survival, doing work you hate, spending time you can’t ever get back. No more settings aside hopes, dreams, or hobbies because you need to earn money to eat.
Any experience you can think of can be had, without every harming anyone else.
And while I realize this sounds, and for many probably *feels*, farfetched, I assure you it is not. This is why:
The barriers to the things above fall in what we call the Known Unknown bucket. We know that, from the perspective of physics, all of the above is technically possible. No laws of the universe forbid any of this.
We’ve already partially solved many of these things, and there’s a good chance the remaining solutions are already within reach, but currently lost in the noise/data.
Many of these problems are likely already solvable, if we could get the right minds and resources working on them.
We have many times many examples of new technology not only making things that seemed impossible, possible, but unlocking new tech trees that make even more things possible. Now picture that, but accelerated, massively.
A sufficiently capable AI should be able to make rapid advances in our ability to run computer simulations of all sorts, enabling rapid experimentation and iteration in a virtual environment that could never be accomplished in a physical environment.
A sufficiently capable AI will be able to iterate on and improve any other technology we have, up to the limits of physical reality (which we still don’t even know).
Some of this, even if you can accept it’s technically possible, may also feel much farther away than I’m making it seem. This is also wrong.
The human brain is TERRIBLE at exponential thinking. We evolved in a linear world, and our tendency is to predict forward based on looking backwards at how things used to work. That doesn’t work for exponentials, not at all.
A famous story, to illustrate:
A wise man did a great service for a king, and the king offered the man a reward of his choosing. The wise man asks the king for grains of wheat, one grain on the first square, two on the second, four on the third, and so on, doubling each time. The king thinks this is very reasonable, and agrees, but fails to realize that by the 64th square, the total exceeds 18 quintillion grains, far more than was in his kingdom.
Most humans just fucking suck at exponential thinking ¯\_(ツ)_/¯
This time around is different in scope, scale, AND speed. There are no historical comps to point to. This is not like books, or agriculture, the internet, or even computers.
We are entering a period where literally every problem you can possibly think of becomes solvable, and soon.
AI is already being used for:
Designing new, better, more efficient computer chips and architecture
New and better algorithms
Novel proteins, drugs, and treatment protocols
Making nuclear fusion viable
Coming up with entirely new metamaterials
Generating synthetic training data to create better, cheaper AI models
Training robots in virtual environments orders of magnitude faster than could be done in the physical world
Improving diagnostics and treatment in healthcare
Self-driving cars and trucks
Supply chain optimization
And many, many more things besides…
All of this, improving wildly in the last few years, and STILL ACCELERATING.
So, please, tell me precisely how, in a world where AI and robots can do literally anything humans can do, better, faster, cheaper, and safer, do humans continue to have jobs?
I’ll wait…
And while you’re formulating your counterargument, I will say that this doesn’t mean humans will be useless, or that they will do nothing.
To even begin to think that without work humans are somehow incapable of finding joy or meaning shows a truly spectacular lack of imagination.
I think this technological path will lead to more human flourishing, and less human suffering, than has ever been possible in the history of this planet. We are really, truly on the cusp of achieving something utterly remarkable.
Will it be a bumpy transition? Probably.
Is the world prepared for it? Nope.
Does that mean we should slow roll it? ABSOLUTELY NOT.
I’d bet my bottom dollar, due to the way incentives are stacked up, that if we don’t speedrun this transition, yanking the band-aid off so to speak, we’ll end up royally fucked.
And as far as I can tell, we are going to just speedrun this. No government on earth can outpace this tech, it moves too fast, and there are too many open source options, and too many incentives to speed it up and make it work.
The genie is well and truly out of the bottle.
So please, for the love of fuck, stop fighting it. Stop resisting. Stop with the doomer bullshit, the anon schitzo shitposting, and the politicized twaddle. The vast, VAST majority of humanity will be better off on the other side of this. Less suffering, more joy, more peace.
The super rich might not be to happy about some of it, but…meh, fuck ‘em.
As for how exactly this transition plays out, I’ve addressed that in more depth here, and plan to write an even more comprehensive post on what day-to-day life might look like 2 years, 5 years, and 10 years out. Coming soon :)
—
If you’d prefer, I recorded a video with some high-level thoughts: