A user testing my AI art tool caught a bias I totally missed

Last month I launched a simple portrait generator for a local artist collective in Pittsburgh. A beta tester ran a photo of her nephew and the tool lightened his skin tone, she called it out right away. I never checked skin tone variance because I trained it on my own album of mostly white faces. She said 'you are baking your own blind spots into the software' and I had to rebuild the color pipeline from scratch. Has anyone else gotten feedback like that that forced a full rework of your model?

0 comments

Log in to join the discussion

Log In

0 Comments

No comments yet

Be the first to share your thoughts on this discussion.