I was thinking of ICLR apparently? I don’t follow conf deadlines or procedures too closely but looks like ICLR is in the middle of getting the first review back or something. Everyone seems to time announcements around conferences (for PR, because of anonymity and quiet periods, or just finishing projects I guess), so I just assume if a bunch of stuff is suddenly being announced by a variety of groups, there’s probably a conference involved somehow. Even when they aren’t actually submitting a paper, I guess they want to be able to talk about it at the conference or have people talk about it there or were thinking of doing a paper? More active conference-goers than myself can speak to the exact mechanisms better than I can.
(On a meta note, I think that one unfortunate consequence of this is to distort perceptions of progress: people seem to saturate at 2 or 3 major announcements, and then the rest get ignored. So 3 or 4 times a year, for ICLR/ICML/NIPS, you wind up having long stretches of time where ‘nothing is happening, deep learning has hit a wall!’ and then a sudden flurry of announcements which wake people up and remind them that, no, DL progress continues to be rapid, but they simultaneously get overalarmed in the week or so the announcements happen - ‘oh my god oh my god oh my god this is really happening isn’t it’ - and then underalarmed a month later when the massed repetition fades out and the urgency can’t be maintained and now it’s the new normal. If everything dropped more evenly, the most upset people would be less upset and the less upset people would be more upset, and on net everyone would be less wrong.)
I was thinking of ICLR apparently? I don’t follow conf deadlines or procedures too closely but looks like ICLR is in the middle of getting the first review back or something. Everyone seems to time announcements around conferences (for PR, because of anonymity and quiet periods, or just finishing projects I guess), so I just assume if a bunch of stuff is suddenly being announced by a variety of groups, there’s probably a conference involved somehow. Even when they aren’t actually submitting a paper, I guess they want to be able to talk about it at the conference or have people talk about it there or were thinking of doing a paper? More active conference-goers than myself can speak to the exact mechanisms better than I can.
(On a meta note, I think that one unfortunate consequence of this is to distort perceptions of progress: people seem to saturate at 2 or 3 major announcements, and then the rest get ignored. So 3 or 4 times a year, for ICLR/ICML/NIPS, you wind up having long stretches of time where ‘nothing is happening, deep learning has hit a wall!’ and then a sudden flurry of announcements which wake people up and remind them that, no, DL progress continues to be rapid, but they simultaneously get overalarmed in the week or so the announcements happen - ‘oh my god oh my god oh my god this is really happening isn’t it’ - and then underalarmed a month later when the massed repetition fades out and the urgency can’t be maintained and now it’s the new normal. If everything dropped more evenly, the most upset people would be less upset and the less upset people would be more upset, and on net everyone would be less wrong.)