Horrible analogy. While the discriminator improves the generator, there are no resulting consequences for the real/fake samples. This is completely different.

Besides, you can do this without the obligation of reviewing (which I encourage you to do).

IMO absolutely nothing.

The truth is that without free reviewing from the community: research would simply not exist (at least peer-reviewing wouldn't). That is why we "must" do it.

We then do mental gymnastics to convince ourselves that there are benefits (many cited here in the thread), when the truth is time would be better spent (at the individual level) on other tasks.

That being said, please put time into reviewing so the community can continue to exist ;)