I don't think improvement for the sake of competition (and nothing else) is worth discussing, because it presupposes that competition is something worth doing. That's why I'd exclude that type of improvement.
I'm willing to exclude this type of competition, but I'll add the following caveat: don't assume that just because it exists as a motivation, there are no other motivations.
With the AI and such, as with all your examples, competition isn't necessary to discover the issues. Testing it, putting it in different situations, getting advice/help/testing from others, and all kinds of other methods could flush it out. There's nothing going on in a competition that's going to be different, at least in that situation.
What say they'd never think to subject it to said examples, and that a competition, seminar, etc. is the only way they would ever discover the imperfection? Even if it
is possible, there are obvious cases where it's easier, more efficient, more rewarding, and more beneficial through competition. Science fairs are a superb example.
Nor have you. Hooray for defending generalizations!
That's the thing: I'm not making a general claim.
I'm skeptical of your claim, and trying to convince you that you should be too.