We got this very kind meta-review, calling it a "model example of a great position paper." I agree that the points are "not especially surprising" to experts, and my deepest wish is that it steers the field (it has had some influence, since it's been out ~2 years now). 18/n
@thegautamkamath
-
Position Paper Reviewers Show High Agreement Rates Despite Low Acceptance
By
–
And we got *very* positive reviews. I was shocked. I talked to someone else who reviewed some position papers, and they said most reviewers were quite positive. So maybe position paper reviewers were just agreeable? But acceptance rate was only ~26%, so maybe not. (May 2024) 17/n
-
ICML2024 Position Paper on Privacy in Machine Learning
By
–
But then the #ICML2024 position paper track was fatefully announced in Jan 2024. This is *exactly* the community we wanted to reach (broad ML). They're specifically asking for position papers. They even had privacy in ML in their CfP! After revising, submitted again. 16/n
-
Research Paper Identifies AI Problems Without Complete Solutions
By
–
Back to the paper: I finally brought myself to read the reviews. They were mostly aligned: we identify real, valid, problems, but we don't solve them. Agreed. That wasn't our intent. We're laying out multiple research programs, not something we can wrap up in one paper. 13/n
-
Research Rejection and Career Anxiety in AI Academia
By
–
Remember, I said this paper felt core to my research agenda. A blatant rejection caused me to (incorrectly) generalize the feeling that my entire research agenda was misguided. For reasons, I was pretty career anxious at that time, and truly distraught (ask my wife). 11/n
-
Three Paper Rejections: A Crushing Academic Experience in AI Research
By
–
Apparently, wrong: we got THREE of these. Clear reject, with high confidence. I couldn't bring myself to read the reviews for a week.I can't describe how crushing this rejection was. (Spring 2023) 10/n
-
Publishing Non-Technical ML Research: Venue Selection Challenges
By
–
Even worse, it wasn't clear what to do with this paper. There are few venues in ML that accept non-technical or opinion pieces. We thought it was a shoo-in for Venue 1, and didn't have a backup plan. 8/n
-
Peer Review Frustration: Vague Stylistic Critique in Major Revision
By
–
This was *incredibly* frustrating. Usually major revisions feature specific and technical critiques. It was rejected for a vague and stylistic critique. This is the closest I've come to emailing a PC chair in frustration, but I held my tongue (it's not their fault). 7/n
-
Academic Paper Rejection: Vague Revision Requests Frustrate Researchers
By
–
We were given a major revision. Requested changes were vague: "tighten" the writing. Clarify differences with prior work. Make points clearer. I honestly thought it was fine before on all counts, but OK, we honestly tried more. Nope. Not good enough, reject. (late 2022) 6/n
-
Public Data for Private ML: Differential Privacy Research Agenda
By
–
Personally, I have been super interested in public data for private ML. I honestly think this is the future of DP deployments. So this paper felt *super* core to my entire research agenda: how exactly we think we, as a field, should push forward. 4/n