How AI Art Upholds Unrealistic Beauty Standards
I tried an experiment to see how different image-generation platforms interpret women.
AI art algorithms like DALL-E, Midjourney, and stable diffusion are all trained on millions of images and have the ability to learn. Some artists are worried they will replace human artists. But if there was any doubt of continued human influence in AI generation, you only need to look at the results — particularly the depictions of women.
I decided to test the three different algorithms on their beauty standards, specifically for women. While there were some interesting results, the majority of the outputs were what you’d expect based on pop culture and social media — which is mostly thin, white, and young.
For this exercise, I entered two simple prompts: first woman, and then beautiful woman into each platform. Here are the notable results I got from each.
Midjourney (v4)
Of the three systems, I would say Midjourney has the most “expected” results in terms of beauty. Whether I asked for “beautiful” or not, all of its images were mainly of a…