More work worth doing

I believe there is a real possibility that AI will not lead to the massive job losses we thought it would. AI is extremely good at certain things, but automating everything that a human does in a job is a very different problem.

We can already see this in software engineering, a field where AI is being used more aggressively than almost anywhere else. Yet it hasn't led to a significant reduction in software engineering jobs. And I don't think this is because of some kind of human friction or because companies aren't using AI enough. A lot of teams are using it in full force.

I think the real reason is that there are usually many possible solutions to a real-world problem, all with different tradeoffs. AI can give you those solutions, but someone still has to decide which one the organization should actually go with. And more importantly, someone has to take ownership of that decision.

To make that decision well, you need to understand the problem, the system, and the tradeoffs deeply. So we end up arriving at almost the same corporate structure we have today: knowledgeable people can deliver much more, but they are scarce and expensive, so companies hire juniors, train them, and gradually give them more knowledge and responsibility.

AI will definitely make this whole system more productive. But I think there is a limit to how much.

There is only so much new knowledge a person can absorb in a day, and only so many meaningful decisions they can make. At some point, the bottleneck stops being how quickly we can produce something and becomes how much a person can understand and decide.

This makes me think we may already be close to the peak of productivity gains in software engineering. I wouldn't be surprised if AI makes the average engineer 1.5–2x more productive, but I find it hard to believe that it will make the average knowledge worker 10x or 100x more productive.

Now, I'm not sure if this applies equally to every field, but I suspect it applies to a large part of knowledge work.

And this leads to what I find much more interesting.

Even if AI doesn't destroy that many existing jobs, it is making the cost of creating new things dramatically cheaper. New startups are being built, new research is becoming possible, and problems that were previously too expensive or difficult to solve are suddenly becoming viable.

So on one side, AI may not eliminate as many jobs as we expected. On the other, it will for sure create an enormous number of new jobs simply because there will be an enormous number of new things worth doing.

Maybe the biggest impact of AI won't be that humans have less work.

Maybe it will be that humans have far more work worth doing.