Forum Discussion
Correcting Bias in AI Image Generation
I love the new AI features!
I do find though that when it comes to AI image generation of people, there's a bias towards white people. If you don't specify race, you only get white people (at least as far as I can tell) and if you do specify race it can generate some quite problematic images.
Would be nice if it would generate a mix of people with more ease and less racism.
I won't add an example image.
Hello KylaPearce-97eb!
I appreciate you taking the time to share this feedback! I've shared this with our Product team as our goal is to continue improving Articulate AI, including image generation. Thank you for pointing this out!
5 Replies
Hello KylaPearce-97eb!
I appreciate you taking the time to share this feedback! I've shared this with our Product team as our goal is to continue improving Articulate AI, including image generation. Thank you for pointing this out!
Racist AI lol...
Any now it's complete opposite. Ask for a person in your image and it almost always gives a brown and black women and if you ask for male it's a black or brown man. You literally need to specify a race and gender to get any variety. And by variety a mean a white person...
The obvious workaround here is to look for images yourself or create them yourself. It's really not difficult to find an image that's close and modify it as needed to fit. Time to skill up and stop depending on what AI spoonfeeds you.
You kinda missed the point — both of this post and what these tools are for.
The thread’s about how the woke mind virus crept into image generation, not a lecture on doing things manually.
"AI" isn’t just used to “spoon-feed” — it’s there to make repetitive, time-wasting stuff faster and easier if you can grasp that leap.
The image is just there to support the text it's sitting beside. Why waste time hunting through Google and editing when I can have it generated in seconds?
Telling me to “just make or find it yourself” kinda proves my point — sounds like you’re missing out.
PS AI wrote the response above 😉
