This assumes a level of institutional control that is nearly impossible (even now). Even if hardware is prohibitively expensive now, I can't imagine training compute will remain that way for long.
I wonder if the JIT model prevents predictable benchmarking/optimization, making long-term robustness difficult? It might also make using MPI difficult. But this is mostly speculation on my part.
In a bit of fairness, I know a lot of very competent engineers who do not code for fun at all. Programming is purely a job for them, and then they go home and hack on nothing.
There's honestly nothing wrong with that. It's perfectly valid to do something purely for the pay (so long as what you're doing is ethical, obviously). I suspect there are plenty of young people who see the bloated salaries that software engineers get and decide to learn programming for purely economic reasons.
This probably contrasts somewhat to the 80's and 90's, where it was substantially more difficult to "get into" computers, and as a result there was a strong selection bias towards really passionate people.
I struggle a lot with my hobbies because I do almost no coding or learning about code in my free time, at least not directly. I think I could be a great(ish) programmer if I did programming as a hobby, but there's just so many other interesting and worthwhile things to do. I get a lot of FOMO from people who code as a hobby, and I think it doesn't help that they don't seem to get FOMO about my hobbies. But I know I will always feel something is missing if I don't spend time marveling at the natural world and creating art, so I guess I'm stuck.
As someone who studied programming in the 90's, I can tell you that I was surprised about how many students were not passionated like myself. And this was already before the wave of scientist coming from other fields (physics, chemistry, biology) to computing in early 2000s.
There are the circles in which the nerds cluster who are really in the tunnel about whatever subject it is, here it's programming or CS in general. They do it for fun, they come home and sit at the computer all evening and half the night. Occasionally they do the homework assignments, and if those are interesting, they nerd out about them endlessly. And they are really good at it. They don't really go to parties and are a tad (or a lot) anti-social. But they don't care. Typical minor subjects are math or physics. It's beautiful.
Then there are the circles of the in-it-for-the-moneys. They do homework assignments because they have to, they go to classes because they have to, they read business news, go to parties, feel important and have dollar signs in their eyes every time they think about graduating and finally making big bucks. If lucky then they get help by some of the nerds (see above). But they actually detest the nerds, just as the nerds detest them.
There are more circles and I don't want to deep-dive here, but I actually this also reflects at the workplace 10-20 years later. It's easy to see from which of those two groups the high-performing ICs (some of them a bit socially awkward) come, and from which their managers, and why there is tension between those.
I interview a lot of students for co-op roles. Almost none seem to program outside of assignments -- or even show an interest in programming outside of assignments. We use that as a weeding-out criteria. Some terms we don't hire any students because we need programmers, not warm bodies looking for money in a hot field.
