News for hackers. Stuff that matters.
Re:Imagen, a text-to-image diffusion model by joshcryer on Monday May 23, 2022 @06:15PM (#31485508)
The bulk of the trained data is from western technology, images, books, television, movies, photography, media. That's where the very real and recognized biases come from. They're the result of a gap in data nothing more.

Look at how DALL-E 2 produces little bears rather than bear sized bears. Because its data doesn't have a lot of context for how large bears are. So you wind up having to say "very large bear" to DALL-E 2.

Are DALL-E 2 bears just a "natural consequence of natural differences"? Or is the model not reflective of reality?

Re:Imagen, a text-to-image diffusion model by Shaanie on Monday May 23, 2022 @07:04PM (#31485938)
That's true for some things, but the "gender bias for some professions" is likely to just be reflecting reality.
Re:Imagen, a text-to-image diffusion model by joshcryer on Monday May 23, 2022 @08:38PM (#31486648)
Don't really know that, either. They said they didn't do an empirical analysis on it. For example, it may show a few male nurses for hundreds of prompts or it may show none for thousands. They don't give examples. Hopefully they release a paper showing the biases because that would be an interesting discussion.