Ah. We’d had on our todo list “make it so post authors are prompted to tag their posts before publishing them”, but it hadn’t been super prioritized. This comment updates me to ship this sooner so there’ll hopefully be fewer un-tagged AI posts.
Logging on today, I noticed that all of the posts with this problem (today) were personal blogposts; are those treated differently? Also some of these are tagged ‘ML’, but that makes it through the AI filter, which.… I guess is intended behavior :/
The situation is that posts show up in the moderator-queue, moderators take a few hours to get to them, and in the meanwhile they are personal blogposts. So if you’re okay with hiding all personal blogposts you can solve the problem that way. This would probably also hide other posts you want to see.
I’m hoping we can ship a “authors are nudged to give their post a core-tag” feature soon, which should alleviate a lot of the problem, although might not solve it entirely.
Wouldn’t generating psuedo-tags which users can opt in to see/filter by mostly solve this problem? Like, I’d have thought even a pre-DL-revolution classifier or clustering algorithm or GDA or something would have worked. Let alone querying GPT-instruct (or whatever) on whether or not an article has to do with AI. The pricing is quite cheap for Goose.AI or other models.
Ah. We’d had on our todo list “make it so post authors are prompted to tag their posts before publishing them”, but it hadn’t been super prioritized. This comment updates me to ship this sooner so there’ll hopefully be fewer un-tagged AI posts.
Logging on today, I noticed that all of the posts with this problem (today) were personal blogposts; are those treated differently? Also some of these are tagged ‘ML’, but that makes it through the AI filter, which.… I guess is intended behavior :/
The situation is that posts show up in the moderator-queue, moderators take a few hours to get to them, and in the meanwhile they are personal blogposts. So if you’re okay with hiding all personal blogposts you can solve the problem that way. This would probably also hide other posts you want to see.
I’m hoping we can ship a “authors are nudged to give their post a core-tag” feature soon, which should alleviate a lot of the problem, although might not solve it entirely.
Wouldn’t generating psuedo-tags which users can opt in to see/filter by mostly solve this problem? Like, I’d have thought even a pre-DL-revolution classifier or clustering algorithm or GDA or something would have worked. Let alone querying GPT-instruct (or whatever) on whether or not an article has to do with AI. The pricing is quite cheap for Goose.AI or other models.
Could work, but it’s a bunch of upfront developer time.