Instagram is the worst offender, I only want to receive message notifications, but I got notifications about inane random stuff I've tried to disable but it won't work. I ended up having to disable notifications altogether.
Instagram drove me so insane with that that I spent a while searching through the app to figure out how to disable it. There's a way to do it, and for a while it worked; I only got notifications about things like direct messages and posts from a few people I specifically told it to send notifications for, but I never got the "recommended" posts.
Then I got a replacement iPhone and reinstalled Instagram.app, and it defaulted to "show you notifications for everything we think you might be interested in" again, and I was too lazy to spend all that time relearning how to disable those notifications. I disabled the notifications entirely and now I open the app once a week or so to check in.
I had to do the same thing with Facebook years ago. Now I open it once a month to see who from high school got married in the last month and click the little "heart" icon and scroll until I get bored (~2 minutes). Can't trust Zuck with notifications.
Instagram run their notifications via an auction mechanism (which I suppose makes sense for an advertising company that likely built a lot of RTB systems).
My theory is that they rate limit that URL aggressively due to AI scrapers. At this point it's faster to just clone the repo and do your searching locally.
Surveillance gathered by an completely autonomous drone with no outside data, stationed far enough away to require refueling, close enough to enemy operations to be useful, that then needs to make its way back to origin, intact, through hostile territory, quickly enough for the gathered information to be useful, seems like a preeeetty big lift. Something a startup would promise to tackle with a star team of technologists over the course of like 10 years? Sure. Something they’d have designed within the past, like, year while getting shot at? I’d have to see that believe it.
> If you are interested in how well we do compared to demucs in particular, we can use the MUSDB18 dataset since that is the domain that demucs is trained to work well on. There our net win rate against demucs is ~17%, meaning we do perform better on the MUSDB18 test set. There are actually stronger competitors on both this domain and our "in-the-wild" instrument stem separation domain that we built for SAM Audio Bench, but we either match or beat all of the ones we tested (AudioShake, LalalAI, MoisesAI, etc.)
So ~20% better than demucs, better than the ones they tested, but the acknowledge there are better models out there even today. So not sure "competes against SOTA models" is right, but "getting close to compete against SOTA models" might be more accurate.
I don't know if you wrote it as a form of satire, but obviously thete is no such thing as "YouTube's front-page". Everyone gets recommended different videos, based on various signals, even when not authenticated.
reply