It feels like it's gotten harder to publish low-quality papers in the AI era.
1 week ago
Recently I came up with a fun idea, and after thinking it through a bit, it seemed like it could work. I sent it to GPT to help me work out some numbers and make a few plots. But GPT went straight to some lemma buried in a dozens-of-pages-long paper from over a decade ago...
Before AI, an idea like this would probably have been enough to get a low-quality paper published. Now I figure the referee just feeds it to GPT and rejects it.
If the lemma in that older paper is only close but not exactly the same setup, I'd suggest comparing the assumptions of the two line by line first, then deciding whether to work around it or cite it with an extension. The premise of this approach is that you can genuinely articulate where your contribution and novelty lie; otherwise, it's more practical to think ahead about how you'll respond to reviewers.