Meditations on Jobs

The market will squeeze, and not just gently but rather hard.

First we will see an automation of most things that can be done at a PC by a human. There will be some tasks where we want a human in the loop, but this will be mainly because these new systems need to build trust first before we can completely defer to them. This will at first only lead to small increases in productivity because from a hiring perspective for most of these jobs, you will still need a human that can oversee the tasks and is somewhat value-aligned with the tasks the institution wants them to finish.

The flywheel will start at the point where management realizes that the employees have not corrected the AI in weeks and have become just glorified proofreaders. I have seen this firsthand in the translation of articles inside of one of the largest papers in the world. There, the transformation desk, which is just a bunch of student workers, does the grunt work for the editors that then rerelease the English articles in their names in their respective territories.

To give you some insight into how this is currently being done. I was given the free rein of choosing between a list of texts that should be translated. I then had to create a copy of the text, press the WordPress ChatGPT Translate button tool, and then had to double-check the AI’s translation (which was perfect every time) and then had to edit some of the corporate language into the AI text and had to check if the text made sense in German. Just for me to hand in the text to an editor who read the text again and, if needed, handed the text back to me with annotations, and I had to reedit the text just to hand it in for review again. I bet my boss could have saved herself a lot of hours if she just did the correction and pushed the text. The company could have saved a lot of hours by not having me choose texts, apply translation, and cooperate on edits. And the whole organization probably would have benefited from firing me and my other student buddies and just switching us out for one headmaster who would be doing what the editor previously did, but just directly publishing the piece.

The profitability squeeze is there, but the market and industries aren’t adapting at the pace that technology is advancing. In a sense this is a good thing, because this decreases the adoption speed, which decreases the speed of the AI data flywheel. Where we get a lot of adoption, there is a lot of user feedback, which makes the product better and leads to even more users. I just know if the first newspaper said that they downsized their team of 8 people to an AI and a person, this will lead to further adoption, and it will be relentless.

The speed of automation, thus, is not only a function of feedback you are getting on a task and the corresponding amount but also the rate at which this application of the technology will find adoption. Price is a nice signal, but you undercutting the competition would only be relevant in a 100% rational market where price would be one of, if not the most, relevant signals.

So a lot of yapping without any implication.

If you are currently doing something that people like because it’s done by a human or you assume this task will be done by a human in the future, don’t be fooled. The sensibilities of people might change quicker than you think, especially if it’s for their perceived benefit. Are people really faster at the self-checkout? Are we really better at building the furniture that IKEA hands to us, or are we really learning more on the internet than in classical universities? Yes, for some people, and no, for most people. All of this changed rather recently, and we had an entire culture of doing X rather than Y. I just want to warn you and be careful: thinking just because something has been done by humans, this doesn’t mean that other humans would not be open to eliminating the human part in the current process at all.

My recommendation would be to find something where you as a person hold some type of network power. That means what you do is intimately tied to you as a person. So there are things just a specific person can do. For example, you can run a sandwich shop like Subway, which is somewhat faceless and corporate and could be done by basically everybody. It’s a safe brand with a safe recipe, but the point of a Subway is that it is, in fact, exchangeable. You might get away with being the last human in the Subway at a point when you are mainly orchestrating machines, but running that specific deli in the neighborhood with people specifically enjoying your food, the quick chat with you, or something else about the environment is different. This shop, in my opinion, has value, because the value is very hard to capture, i.e., it’s a vibe, a feeling, a mood, something ghostly, and not really touchable.

It is also very hard for all the people that once liked the vibe to lose the vibe at the same time. If the not-so-good times come, you will feel and might be able to adjust or pivot from what people liked about your place. The same applies to the business of coaching, motivating, therapy, and other services that rely on you as a conductor. A lot of money-making advice will tell you not to pursue these activities because they don’t scale or you are trading money for your time. How will you become a billionaire like this? The answer is you most likely won’t become a billionaire by doing this. But be honest, what was the probability that you would become one anyway?

Everybody is telling people that we need to become more unique and acquire skills that are hard to reproduce for AI. If you can train on the task and measure some type of “loss,” you will be able to train an AI on this, and especially if the task is rare and valuable, AI companies will flock to this cash cow. So what you would rather want is to find a stable niche in the job market that can be exploited. A stable niche cannot be disrupted easily; that means you have some type of idea of how a disruption would look and where it might be coming from, and you judge that this is far enough away that you might as well pick this up. Another thing for the niche that would make it good would be little attention. I.e., it has some natural barriers that make it less attractive for other people to look into this topic. Then it also has a rather small or old community of people running this issue. This usually means there is some free energy to be had by making the process/topic more efficient.