But engineers have these same problems. The key is that the content creator (engineers for codegen, doctors for medicine) is still responsible for the output of the AI, as if they wrote it themselves. If they make a mistake with an AI (eg, include false data - hallucinations), they should be held accountable in the same way they would if they made a mistake without it.

Some schools run a "breakfast club" that everybody's welcome to attend, where they provide things like toast or cereal to kids that don't get breakfast at home, and it's couched in shame-softening language, though most kids know that if you go to breakfast club it's probably because you can't actually afford breakfast.

Schools will often have some bread and spreads available in the office for kids who are sent to school without lunch. I'm not sure how widespread it is, but I know that in some schools this is just funded voluntarily by some of the staff who will pick up more bread or whatever when it's required, because they don't want to see kids go hungry.

I think the idea of having lunch provided as part of your school fees is actually a good one. No kid should go hungry, or be subject to humiliation and shame, because their parents can't afford or can't be bothered to provide them lunch.

* Not provide background information and let people figure it out for themselves. This will not help me achieve my goals.

* Link them to Google's SRE book and hope they read it. Still not achieving my goals, because they won't.

* Spend 3 hours writing the relevant background information out for them to read as part of my proposal. This will achieve my goals, but take an extra 3 hours.

* Tell the LLM what I'm looking for and why, then let it write it for me in 2 minutes, instead of 3 hours. I can check it over, make sure it's got everything, refine it a little, and I've still saved 2.5 hours.

So for me, I think the author has missed a primary reason people use LLMs. It saves a bunch of time.

I don't think so. I've been doing this for nearly 35 years now, and there's always been a lot to learn. Each layer of abstraction developed makes it easier to quickly iterate towards a new outcome faster or with more confidence, but hides away complexity that you might eventually need to know. In a lot of ways it's easier these days, because there's so much information available at your fingertips when you need it, presented in a multitude of different formats. I learned my first programming language by reading a QBasic textbook trying to debug a text-based adventure game that crashed at a critical moment. I had no Internet, no BBS, nobody to help, except my Dad who was a solo RPG programmer who had learned on the job after being promoted from sweeping floors in a warehouse.

The real change though comes from parent's perceptions. Right now there's age limits of 14-years-old on most social media platforms, however most parents just see this as a ToS thing, and nobody cares about actually violating it. Once it becomes law, the parents are suddenly responsible (and liable) for ensuring their children are not breaking the law by accessing social media. It's not going to stop everybody, but it'll certainly move the needle on a lot of people who are currently apathetic to the ToS of social media platforms.

It's a bit different for requests though. FastAPI will allow you to define your request schema (application/json or application/x-www-form-urlencoded) and validate it using pydantic, but starlette doesn't do that OOtB. It's trivial to implement though, and if it were me I would probably choose to do that rather than deal with FastAPI's inflexibility.