Twitter | Search | |
Suchi Saria
NeurIPS needs 4500 (qualified) reviewers this year! Looks infeasible. it may be worth trying a process w/ two rounds. The first round—like top journals—rules out papers unlikely to score more than 3. The papers that make the cut get multiple reviewers.
Reply Retweet Like More
Hugo Larochelle Feb 6
Replying to @suchisaria
It will take work, but I think we can do it and find that many qualified reviewers. We’re not giving up! :-)
Reply Retweet Like
Suchi Saria Feb 6
Replying to @hugo_larochelle
This may not just be about will: I think the quality of the reviews for the >3 papers will go up and the final decisions will be much less noisy. This is something we can verify using a simulation; it would take a very large pool (>10K qualified) for this to not hold.
Reply Retweet Like
Jon Bloom Feb 6
It’s a resource-constrained predictive task. How could you, of all people, not apply active learning?!
Reply Retweet Like
Christopher Peters Feb 6
Heh, set a neural network on it! 😜
Reply Retweet Like
Matt Miller Feb 6
I think you should probably give up. That’s what I would do.
Reply Retweet Like
Daniel Roy Feb 6
I see someone volunteering for program chair 2020!
Reply Retweet Like
Jasper Feb 6
When the going gets tough, gets going?
Reply Retweet Like
Matt Miller Feb 6
Yes I am both cowardly and very very lazy.
Reply Retweet Like
OneBigOh Feb 6
Somebody needs to submit a paper titled “Deep neural networks for reviewing NeurIPS papers”
Reply Retweet Like
John Mannisi Feb 7
Need AI reviewers!
Reply Retweet Like
Suchi Saria Feb 9
2022 or 2222 sounds like better years than 2020. Also, ideas for simple experiments is the least critical skill needed to be a PC :-)
Reply Retweet Like