Post

We launched some updates to the Discover feed today. This likely won't affect existing users as much, but for new users it's a big improvement. Basically: the ML team managed to build an improved topic classifier, which means Discover is better at responding to selected interests.

The classifier uses BLIP2 embeddings with training on labels which we manually created (yep). The ML team has the exact numbers but anecdotally it seems to have jumped from 40%ish accuracy to more like 90%. I'm surprised at how much more consistent it is.

This affects new users more because they haven't followed or liked anything yet. As you do both, you get more specific postes. When you're new, we don't know much about you, so we use these broad topic classifiers.

We literally just need it to be chronological It doesn't fucking matter if it gets to us because we like it It matters that we see what gets posted in real time We have the very real problem of developers trained by Nazis to fix

This should never happen For any reason Ever

You're looking for the "Following" feed, not the "Discover" feed, then. The "Discover" feed isn't ever going to do what you want.

meaning less furries posts?

Thanks for the update, sounds like a significant improvement. Any idea how much this impacts racial / gender / etc biases?

Why would it? Is there any race/gender based topic?

Classification isn't always cut-and-dried, there's subjectivity - which leads to bias. On Twitter back in the day, topics like "news", "politics", and "tech" were mostly guys and mostly white. Posts by and women and people of color were more likely to be considered "entertainment".

Can you share more on what the labels are? Is it multiple categories like "politics", "art", "anime"?

If you go to Explore you should have a tool at the top for editing your interests. (If you dismissed it, then check somewhere in your settings I forget where) The interests are 1:1 to the topics right now

Ah, I see. So you rely on those classifications to match content to users for cold-start. How about established users? Do you not use the classifications and instead use the raw content embeddings and assign them to clusters?

Just curious why we can’t upload our own animated gifs, and like… have them animate. Can’t be a technical thing, y’all are too smart for that. So maybe a storage volume issue, or something contractual with Giphy that’s limiting implementation? 🤷🏼‍♀️ Just idle curiosity.

try using the video upload