As someone who frequently publishes in journals (my country's bean-counting system makes it almost mandatory) I think this frequent take is a case of grass is always greener. Journal publishing has its own issues and I'd say it leads to even more arbitrariness and randomness and worse quality.

It's also highly questionable to apply sticks based on subjective criteria (and reviewing quality is inherently subjective). Carrots are much less problematic in this respect.

Also worth noting that while I don't closely follow results on poetry, at least on narrative GPT-3.5 has been shown to be considerably worse than GPT-4.

Got 10/10. Not sure to what extent I'm good at this and to what extent lucky.

And the social event was among the best I remember in *ACL conferences, if not the best. Masterfully organized. Great food, places to sit, open bar, DJ, a full museum to explore including planetarium, and great views. The only possible improvement would have been to make it last longer 😊

The few of us who care would care a lot, though, and for a longer time than if they release some LLM improvement :)