It seems that the author prefers to ignore the fact that, throughout history, apprenticing has been reserved for the common people, while teaching was reserved for their masters, the rich and the powerful.
Having both is better, but at some point you need to learn the theory.
* By some accounts Toyota's electronic throttle control system (ETCS) had bugs that could cause sudden unintended acceleration
* The Boeing 787 Dreamliner experienced an integer overflow bug which could shut down all electrical generators if the aircraft was on for more than 248 days
* In early 2019, the transportation-rental firm Lime discovered a firmware bug with its electric scooters that can cause them to brake unexpectedly very hard, which may hurl and injure riders.[6
* A bug in the code controlling the Therac-25 radiation therapy machine was directly responsible for at least five patient deaths in the 1980s when it administered excessive quantities of beta radiation
- teachers : if you have let's say 30 students, you need one teacher, in you have 300 you probably need 10 teachers. I don't know how can you one person deal with 300 students giving them the required attention. Since you have the school program you can start paralyzing the tasks.
- doctors : If you have 10 sick peoples... 1 doctor, with 100 you probably need 10 doctors. A doctor need time with a single patient to study the case.
There is probably plenty of good examples, but not that ones.
About building softwares, even if you pick the right guys and split the task properly (which requires peoples, time and a high level of understanding) and let peoples work on their own, you lead to inconsistency. Splitting an existing task to add new developers and save time requires redesign and headaches to avoid incompatibilities, probably breaking new things or requiring some rework in the work done by the others.
This article is purely theoretical and I don't think they do such brainstorming about how to train the IA. It's proved that they just take a brunch of poor peoples poorly paid to deal with real messages and try to moderate them sometimes with a poor understanding of the context, and use it as an input to teach the IA how to do "proper" moderation.