This article by Alex Danco has been very thought-provoking. It touches on many things I have been thinking about as I observe this new AI paradigm unfold over the past year.

The core argument is that AI technology will make societies that adopt it much more productive, elevating the prices of tasks that will not use AI, such as dog walking. However, I find this argument to hinge on the assumption that this paradigm will create new, high-paying jobs for humans whose tasks are displaced by AI. If these new jobs do not materialize over time, many will gravitate towards jobs that are not affected, making them cheaper as they become one of the few non-automatable jobs left for a massive pool of human labor.

The article does offer one other path: the "Turbo-Baumol's" effect, where regulation creates high-paying "bottleneck" jobs for humans who just sign off on AI's work. This will likely be a way that governments fight job displacement, but I am skeptical that enough jobs will be regulated this way. Furthermore, society will push back against this regulation, especially if these services are considered cost-prohibitive.

All of this is very theoretical. I believe that full automation of knowledge work will happen in my lifetime, but it will take over 10 years, not only because the technology is not there yet, but also because there will be a lot of friction in the adoption phase of these technologies. In the meantime, the gradual adoption of AI will continue, and as compute gets cheaper, people are going to find novel ways to adapt AI into their workflows in ways we can't even imagine right now.

https://a16z.substack.com/p/why-ac-is-cheap-but-ac-repair-is