>>11337283>Is that why there are so few papers?Paper submission and acceptation are at an all-time high, causing severe problems in the review process. The reason why I suggest the years 2010 to 2015 for papers is related to this: there are so many papers per reviewer at nips and the like that the reviewers are handing review duties to undergrad interns who've never seen a paper in their lives, resulting in garbage-tier review process that lets absolute dogshit through. I think it was last year that they accepted a paper that literally said "we used a normal convnet except we added position index and it didn't work".
>Is it true when people said we've hit a brick wall in AI?Yes. It's part of the growth process of new scientific fields. We're past the "wild west" phase, now we actually need concerted efforts to make breakthroughs. There has not been anything innovative since 2016 except neural odes, whereas we used to have something new and exciting every other month back then.
On the plus side, the theory (which has significantly lagged behind until now) is finally starting to catch up because of this wall, so maybe it will reach far enough along that we'll get a new nesterov momentum-tier paper.