Why Humans Will Likely Have Jobs in the Age of Human-Level AI
What is Comparative Advantage?
You can check out my other articles or say Hi👋🏾 on Twitter/X too!
On to today’s piece 🤝
Why Humans Will Likely Have Jobs in the Age of Human-Level AI
With headlines like “Klarna CEO says AI can do the job of 700 workers”, it can be easy to feel that automating customer service reps is just a skip, hop and a jump away from automating everyone else’s jobs too.
But despite the fact that it wasn’t as revolutionary as it seemed, it’s undeniable AI will only improve. How will it be 5 years from now? 10 years? 20? What happens once AI hits a human-level functionality? Will we still have jobs once AI reaches human-level capabilities? Or are we destined for a world of mass unemployment and economic irrelevance?
Surprisingly, the answer is that humans are likely to remain indispensable even in an age of unimaginably advanced AI - and the key to understanding why lies in a little-known economic principle called comparative advantage.
What is Comparative Advantage?
Typically when we think of an advantage, we usually mean we’re in a more favorable or superior situation compared to someone else.
But comparative advantage is none of that; in fact, it’s the reason why today you have a job. For example, how much better are you at physics or mathematics than Princeton Professor Edward Witten, creator of M-theory? Or how talented of a programmer are you compared to Linus Torvalds, creator of Linux? How great of a businessman are you compared to Mark Zuckerberg, king of a $1 trillion empire? I’d be willing to be bet, if I searched, I could find someone who is more talented than you at every role you could conceivably do.
And, yet you’re still employed.
Comparative advantage exists even if someone or something is better than you at everything.
Why?
5 cent cans.
Ever walked around San Francisco and observed how many used cans and bottles are lying around? Each can is worth 5 cents and with enough collected, you could make a living. Given your competitors, you could probably workout a solution where you could collect a lot more cans, way more efficiently. So why don’t you? Don’t you like making money?
Well you don’t because it’s not worth your time. You’d rather work a job or run a buisness or do other things because there’s a finite limit to your time & resources. If you were walking around picking up 5-cent bottles, you wouldn’t be doing what you’re doing now and the delta between the two would be the opportunity cost.
Therefore, even though you can make money picking up 5-cent cans, you don’t, because relative to your abilities, it’d be equivalent to taking a massive pay cut.
Comparative advantage operates on the basis that a entity is limited by certain constraints (opportunity cost) and therefore will likely pursue its relative advantages. Just as you don’t look to picking up 5 cent cans, AI is not comparing itself to you rather it’s comparing itself to every possible task it can do.
But surely I must be forgetting something; can AI just scale and do everything?
Well what is scaling? If I say my neighborhood lemonade stand is scalable, what I really mean is I can standardize my lemonade-creation process, productize the lemonade and sell it widely. But scaling my lemonade stand has constraints; the amount of lemons farms can seasonally produce, the transport trucks, factories to create lemon concentrate, store shelf space to stock my lemonade, and the capital to run the buisness.
AI has constraints too: energy and compute. While energy is important, sufficient compute will remain the constraining factor of AI and no matter how much of it there is, at any given point of time, there will be a finite amount of compute in the world.
Therefore, AI is not infinite.1
But as energy and compute become cheaper, and the cost of AI drops, won’t massive abundance kill any comparative advantage humans possess?
No, it won’t. In fact, the opposite is true - the more abundant and capable AI becomes, the more powerful the logic of comparative advantage will shape the AI-human economy.
As AI gets better and more plentiful, it dramatically increases the total amount of value produced in the economy. But all that new wealth has to get spent on something. And basic economics tells us that spending will flow to where there’s highest marginal returns - the greatest bang-for-your-buck. So the most valuable and productive applications of AI will see skyrocketing demand.
This means that no matter how cheap and abundant AI gets in an absolute sense, using it for anything other than its most valuable possible uses will get increasingly expensive in relative terms.
This is the same reason that although while smartphone usage has reached over 68% of humanity, the vast majority use smartphones for its most valuable properties instead of all of its properties. Texting their friends, watching funny cat videos or posting cringe duck kisses on Instagram are all the highest (relative) valuable use cases; people don’t buy smartphones to use them as paperweights. Few are crazy enough to consider dropping $500 on a <1lb device to hold some papers down; the opportunity cost is too high.
So, AI will inevitably get allocated away from any task where its comparative advantage over humans isn't that large (even if its absolute advantage is huge) and toward the tasks where its comparative advantage is greatest.
This is admittedly very counterintuitive and the zero-sum, competitive advantage frame is pretty deeply ingrained. But while there are still many challenges around inequality, adjustment, and AI agency, the employability of Tom, Dick and Harry looks set to keep the high ground.
- Kiran
Thanks to Claude for editing.
If you liked this piece, make sure to subscribe by adding your email below and check out my other posts!
Kiran’s Gems 💎
Highly recommend checking out
’s Plentiful, High Paying Jobs in the Age of AI which inspired this piece.“For years, some of the world’s sharpest minds have been quietly turning your life into a series of games. Not merely to amuse you, but because they realized that the easiest way to make you do what they want is to make it fun. To escape their control, you must understand the creeping phenomenon of gamification, and how it makes you act against your own interests. ”
’s piece on Why Everything is becoming a Game is exploding on Substack and I’d be remiss not to share it with you."Ra is a specific kind of glitch in intuition, which can roughly be summarized as the drive to idealize vagueness and despise clarity."
’s Ra is one of the best essays I’ve read recently.
This is important to remember as many times people often point to scalability as a way to get away with considering real world constraints.
Great summary.
I listened to the original podcast where Noah Smith talked about this and then read a bunch of later analysis.
It makes sense on a theoretical level, but it also feels really hard to estimate what the cost of compute will be in the future. On the one hand, there is almost unlimited need for more compute if the AI is smart enough.
On the other hand, we know that AI right now is incredibly inefficient compared to the human brain. And that all the AI companies have been making AI more efficient at a drastic pace.
What frameworks do you use to think about the price of compute in the future?
Nobody serious is saying we won’t. The question is how many, and how economies and governments will have to change so a huge spike in unemployment won’t lead to mass poverty.