How many Google TPUs do you need to generate database schemas from a natural language description of some state machines? Any competent programmer can solve that problem. Let me know when the folks at Google Brain figure out how to do that.
Why would they do that in the first place? To calculate the next decimal of π does not require too much TPU either, but many engineers stay aside from this problem too.
Why did you pick the database schemas as an example? Database schemas are not a must, even more, database schemas are so easy that even Rails can do. Why anybody would care about redoing this already done stuff? I am not aware of any complicated project that indeed involves the database. DB, CRUD, all that crap is perfectly doable with Wordpress, even Rails is an overkill here.
Database schemas are like elusive Joe.
Because database schemas require semantic understanding. That's why I chose database schemas.
We're a place where coders share, stay up-to-date and grow their careers.
We strive for transparency and don't collect excess data.