|
Top-level
 I think we need to get the word out that doing an "image search" for reference images isn't good practice anymore if you care about anything being based on real creatures, plants or anything. And it may be very limiting for even fantasy subjects as you are getting normalized and reprocessed image pudding. 6 comments
 @futurebird @justafrog So basically, so-called #ArtificialIntelligence is ruining everything.  @BruceMirken @futurebird More precisely, people's inability to separate out fact from pleasing misinformation is ruining things. If rejecting that slop were universal, it wouldn't cause a lot of difficulty. Instead we've got millions of people who think they can finally outsource all their work. Not everyone has the same experience. There are areas where the slop works on each of us since we don't know enough to see what's wrong with it. For example AI slop cars? That would work on me. I don't care enough about cars to know a Toyota from an Audi. @futurebird @BruceMirken The best example I know of is with young peacocks. The fake ones have tiny adult plumage, the real ones have brown. And really short tail feathers which the young males will display anyway, because it's what dad does. Plenty adorbs without fakes, but the fake version is much prettier to look at. The fake one is in stock image collections, which makes it worse. @justafrog @futurebird But people looking for a picture of Thing A are often doing so because they're unsure what Thing A looks like. It seems unreasonable to expect them to be able to discern fake images. |
Gregory's Server
@futurebird Much to my sorrow, image searches now contain an extra workload for me, in that I need to analyze each one for "is this slop?" and I hate that so much.
Often, the slop appeals more to people, so you get things based on it, which were made by a human.
It seems very hard to stem the tide, if it's possible at all.
Even after LLMs get dropped and nvidia dies, the garbage will stay with us for decades.