Content pfp
Content
@
https://warpcast.com/~/channel/ai-art
0 reply
0 recast
0 reaction

Michael Neff pfp
Michael Neff
@heyneff
Have you experienced bias in your gen AI prompt results? I’ve found that vague prompts often result in weirdly banal responses that show patterns, some of which verge on bias. Do you have experiences like this?
1 reply
0 recast
2 reactions

littlecakes pfp
littlecakes
@littlecakes
basically always
1 reply
0 recast
1 reaction

Michael Neff pfp
Michael Neff
@heyneff
Any example outputs you’re willing to share? (I’m not asking for anyone to reveal their secret sauce, so don’t feel compelled to do that.)
1 reply
0 recast
1 reaction

littlecakes pfp
littlecakes
@littlecakes
well i'm not sure it'd be obvious even if i showed you something but basically, a lot of what i prompt that is supposed to be japanese doesn't come out japanese, like the data has been lumped as east asian - which is how i've experienced just living in the west anyway. or it might be correctly japanese but i can tell that the dataset it's working from is shallow on that subject because it can't give me a good range. stuff like that happens a lot while i'm working
1 reply
0 recast
1 reaction

Michael Neff pfp
Michael Neff
@heyneff
It feels weird to ♥️ that as a comment, but I appreciate your candor. I’ve been keeping an eye on bias from the beginning and it’s been obvious along the way and still is, even as there has been clear effort to make broad prompts return more diverse outputs.
0 reply
0 recast
1 reaction