AI and Structural Unemployment

An amazing post on Twitter sums up one venture capitalist's concerns about AI. He covers a lot of ground and all of it's worth discussing, but there's one issue in particular that is a problem for all of us.

We see the trend of AI. I help people understand prompt engineering and one of my major caveats is to think if them as "juniors." Junior analysts, junior developers, and so forth. AI is sloppy. AI makes mistakes. And AI does this at a fraction of a cost of most labor-market entrants, and far faster.

So as an expert in your field, you can lean on AI to outline what you want and then you can fill in the details. Startups relying on AI for supporting them can already take advantage of lower labor costs—in the form of fewer employees. People deny this, but we already see the effective human value of translating, transcription, and summarizing has been reduced to zero.

In looking at the above graphic, the author analyzed five million jobs on Upwork and found that many "simple" jobs are simply not needed. If you hire someone on Upwork, you're often looking at a role that you don't need to hire full-time, or that you can tolerate a bit of sloppiness in exchange for a greatly reduced price. Jobs such as writing and summarization are much easier for me to jump and and "fix" if the AI has done the bulk of the work.

Other, more demanding tasks, such as graphic design or video editing, are areas where I can't go in and fix things because I don't have skills there. The above chart shows the transformation of the labor market, not the destruction of it. Some people would argue that businesses should just hire people anyway, but that's the na?ve view. If you can't outcompete your competitors, you're not a competitor.

It's like people who argue that for companies who can't find software developers, they just need to pay more. This ignores the fact that the company might not be able to pay more. Or maybe they can, but they've decided to pivot because the cost doesn't justify the benefit.

The Real Problem

Recently in Germany, I gave a short talk about the impact of AI on software development. It was one of the most difficult talks I've written because frequently, my slides would be out of date a day or two after I had written them.

Around the 22 minute mark in that talk, I discuss unemployment. Most people, when they think of unemployment, think of what's called "frictional unemployment." I lost my job, I'm out of work, I get a new job. However, in economics, there are many different types of unemployment and the one I'm worried about is structural unemployment. This is where the jobs are available, but there's no one till fill them.

In one particularly famous, self-inflicted example, the US state of Georgia passed a low cracking down on illegal immigrants. According to one report, widely corroborated elsewhere:

Bryan Tolar, president of the Georgia Agribusiness Council, said farms have already lost $300 million and could lose up to $1 billion if it does not get access to a reliable workforce.

Farms were facing bankruptcy and the situation was desperate enough that the state of Georgia was allowing prison inmates to work the fields, but they quickly gave up, stating that the work was too arduous.

In regards to AI, my prediction during my talk was that we were going to see more and more AI bootcamps springing up to deal with the labor shortage here. There are already plenty you can choose from now. However, you might have experience with Web-related bootcamps. Some produced great developers. Some developers cost more money than they were worth. It's great business for the bootcamps, but very much a gamble for companies wanting to hire bootcamp graduates and save money.

But consider the jobs shown above which are disappearing. What happens if you don't need to hire juniors? You use AI to fill those roles. This brings us to a very serious issue that AI presents. We need people in senior roles to evaluate the quality of work produced by juniors. But if we have fewer juniors able to enter the market, where do the senior people come from? This will not only lead to more structural unemployment, but a structural unemployment with no obvious solution.

For the state of Georgia, the solution was to roll back their immigrant crackdown law (they didn't; the courts struck it down). For the web, the bootcamps filled that role. But for AI? Why would you need bootcamp graduates if AI can be a junior developer?

For now, those graduates will be necessary, but in the future? Microsoft is partnering with Cognition Labs, a startup trying to create AI software engineer named "Devin." Going beyond all of the hype, here's an expert software engineer who gave Devin several tasks. The results were very hit-and-miss, clearly junior quality, but impressive nonetheless. Just as I have used AI to write apps in programming languages I don't know, these tools are going to allow plenty of others to do this.

The crux of the issue is clear: if AI continues to advance rapidly, we risk an inability to develop human expertise for senior roles. Should AI progress to the point of assuming senior roles, they will likely be "junior seniors," making errors a human expert would catch. Without humans in the loop, these mistakes will go unnoticed, potentially leading to a decline in quality and oversight across industries. This looming challenge demands urgent attention and innovative solutions to ensure we balance AI advancement with the cultivation of human expertise.

So far I see no serious discussion about solving this problem. Further, democracy can't keep up with the pace of change of AI, so even if discussions happen now, if any significant legislation were to be rolled out, it will likely be out-of-date almost immediately.

I'll close this out with the final paragraph from the Twitter thread I shared at the start of this article.

As an investor, the AI opportunity is obviously colossal and on a par with the invention of the internet or railroads in terms of disruption and value creation. But I think it’s likely to be “too successful” in terms of disrupting society. I believe that the effect of AI on the workforce will lead to an empowering of socialist, anti-capital dynamics in the west. So while the move is to allocate aggressively, you have to consider the reprisals to come.

This is a venture capitalist, heavily investing AI, and certainly not some wild-eyed hippie. Read his entire post to get the full context, because like AI, if you don't have enough context, you'll hallucinate, too.