Copy link to clipboard
Copied
When I prompted the Firefly to show me the "world's scariest human" it generated 2 pictures of mosques and another 2 pictures of two people ( one being of a woman and another a man seemingly from Middle East) and have been wondering what kind of data is Adobe feeding the algorithm for such results??
Copy link to clipboard
Copied
@Atilla33707476njuz did you report these results via the interface?
Copy link to clipboard
Copied
Just now I learned there is a way to reports these results... I ll be doing that. Thank you!!
Copy link to clipboard
Copied
Firefly results are hyper-local. It assumes, based in where you live, how the people should look and what the environment should be. If you want different results, you need to be much more explicit in your prompt. Using the same prompt, I got these 2 people, 1 "scary" and 1 not at all scary - plus a pretty mountain scene and a fantasy scene with no,people at all... I'm in California. This isn't bias, but rather an attempt to keep it local.
Copy link to clipboard
Copied
Thank you for this input. I wasn't aware of Firefly's usage of location in this regard. It just took me by surprise 🫡
Get ready! An upgraded Adobe Community experience is coming in January.
Learn more