I don't see the racism in these pictures. Because the people depicted are black? But that is true for police as well as those being arrested? Is it the art styl

I don't see the racism in these pictures. Because the people depicted are black? But that is true for police as well as those being arrested? Is it the art style? That said, AI image generation is in the process of being lobotomized very strongly. Just look at stable diffusion 2.0. 2.1 was better, but 2.0 just generated something random if it could be anything offensive. I think 1.5 is still the most used model because of the strategy to inhibit certain themes, even if 2.1 is trained with higher resolution images. I think the focus on generating politically correct images is a wrong one. I can understand that some don't want to generate as many naked asian women as the general internet userbase wants to, but I think any "safe" tool, will be one that is less capable in a fundamental way. Businesses need to be aware any creative process can generate something offensive. If that is a no-go, they want a robot instead of an AI. Aside from speed, I think stable diffusion is the only real contender for wider adoption and success. Just because you can run it on low end machines yourself and the platform is open. It has thousands of people that gladly train models and while other models can keep up with the standard themes, they are useless on concepts they aren't trained in.