I was recently listening to an episode of the All-In podcast (from 10/18) where they were discussing the impact that AI will have on jobs.

Unsurprisingly, they argued that AI will bring abundance, and that any job loss will be temporary and offset by economic gains. It's an argument you hear a lot in Silicon Valley. It goes something like this: “Don’t worry, we’ve been here before, every big technology shift destroys some jobs but creates more in the long run.”

So surely the AI revolution will be the same, right?

Maybe. But I can’t help feeling like this time, it’s different.

The Case for Optimism

In the podcast, they laid out three main supporting arguments:

1. There’s a recruiting cycle for new jobs to build the new “thing” before demand for the old thing goes away.
2. AI will create new industries we can’t predict today and, as a result, create new jobs.
3. People whose jobs are impacted will have the opportunity to do more interesting work and get paid more.

And historically speaking, this has been mostly true:

• Industrial Revolution — automation in textiles displaced artisans, but industrialization overall exploded total employment as new sectors emerged (finance, retail, manufacturing, etc.).
• Electrification — factories scaled up, new manufacturing and service jobs appeared.
• Computing — computers automated manual data entry but created software, semiconductors, telecom, and e-commerce industries.

I’m not going to bore you with graphs showing how GDP and life expectancy have skyrocketed, that’s all true.

But like I said, this time feels different. And the best way to test that is to see whether the arguments from the podcast still hold up.

The Horse and Buggy Fallacy

The podcast uses the example that horse-and-buggy drivers were displaced by the Model T, but new demand for assembly-line workers created opportunities that didn’t exist before. Eventually, the economic gains from the automobile industry far outweighed the losses.

I think this is a bad comparison to modern AI-driven job loss.

Horse-and-buggy drivers weren’t disrupted by the engine itself, the animal vs. combustion engine was just an implementation detail. You still needed a human to operate it. A buggy driver could become a taxi driver.

But an Uber driver can’t become a Waymo driver. There’s no lateral skill set the Uber driver can apply in a new domain. Autonomous driving doesn’t replace the transportation engine, it replaces the operator. And when you replace the reasoning engine of the work, you usually remove the human from the loop entirely.

It’s not hard to imagine that models will soon be good enough that keeping a human around will just slow things down or make things more dangerous. We’re already starting to see that, as Waymo is significantly safer than human drivers.

The “More Interesting Work” Argument

Another claim they make is that people will simply move on to “more interesting” or “safer” jobs. They talk about how demand for AI is driving hiring in industries like mining rare earth metals for AI infrastructure and robotics. While that might be true, I’d argue that working on an assembly line wasn’t more interesting than driving a buggy, and mining for rare earth metals is probably not safer than being an Uber driver.

The main problem with this argument is that it assumes those who are displaced will have access to the newly created jobs. That was true in the past, when the new jobs required similar skills. Like I said above, a buggy operator could become a taxi driver. But in this new world, it’s not clear whether the new jobs will be available to those displaced and if they aren’t, will the people who do take those jobs have any need for the skills of those who were displaced?

The reality is that Uber drivers aren’t behind the wheel because they can’t find financial modeling jobs, they’re there because they can’t do financial modeling jobs. Taking that job away doesn’t suddenly turn them into analysts.

Do I think AI will create new industries we can’t predict today? Almost certainly.

Do I think those industries will also be heavily automated using AI? Absolutely. Who’s going to build a new warehouse full of analog labor when they can fill it with AI-driven robots instead?

Where Do Displaced Workers Go?

A recent New York Times report covered how Amazon plans to replace over 500,000 warehouse workers with AI and robotics.

So where should those workers go? If the All-In hosts’ logic holds, there should be a recruiting cycle for new jobs to build the new “thing”. But we're already seeing job displacement, entry-level software engineering and customer service employement has fallen 13%. yet there are no signs of a new "thing" that should be driving demand that will result in new jobs. Maybe it's data centers?

Except warehouse workers aren’t going to be building data centers. Building data centers is a highly complex and technical job. It seems wildly unlikely to me that someone is going to go from packing boxes in a data center to wiring up GPU clusters.

They’ll likely just move to another warehouse, one that hasn’t been automated yet, until that one eventually follows the same path. This migration will continue until most of these jobs are automated, leaving only low-wage warehouses that can’t afford to upgrade.

Realism ≠ Doom

It might sound like I’m a doomer here, but I’m not. I think AI will change our world for the better, probably faster and more profoundly than we expect, especially in medicine, science, and technology.

But it’s irresponsible not to soberly evaluate the economic consequences.

Evis