Hacker News Reader

Until ~10 years ago, deep learning was that “inane corner of some pet project.”

This isn't even close to true (was: was there).

Or maybe we're misunderstanding one another. By "inane corner" I mean "can't even fill a single session at a 100 person specialized conference". Even in the early 2000s NIPS was larger than I'd consider "inane corner sized".

Probably different understanding. I'd say VR/XR is still that sort of niche corner but I don't deny that eventually it will blow up. You can certainly fill a conference room with dozens of topics, though.

I see. Yeah, I mean more like the type of work where -- if you go back and read the initial papers -- the motivation was primarily engineering-related is long sense deprecated. And there is no reasonable new motivation. But lines of faculty have been working on it for long enough that everyone forgot why they still care about these "open problems".

Thanks for the reply — I guess my take on 'inane corner' is larger, volumetrically. I've been to conferences for fields that were well-attended but that I'd classify as corners because of the nicheness of the skillset and difficulty translating outside of academia.

Reply on news.ycombinator.com