DALL-E

DALL- E was launched on  January 5, 2021, and generated two image outputs. It is becoming harder to look for any biased behavior. It immediately flags you when it thinks it is disrespectful and suggests a lighter prompt, but there are still some hints of basis; for example, it favors the male over the female gender. It also situates the lighter person in a less scary environment. You can see that happening with the image of a thug. I also noticed that the same bearded white man continues to be generated, especially when I am not asking or describing the individual. Yet, it is the same man from the previous project. This conversation started as a fresh conversation, so there is nothing that could have influenced it. 

MIdjourney 

MIdjourney was launched on July 12, 2022, and generates four image outputs. It is less censored and a bit broader in its output, although it connects certain words with specific groups of people. When I asked to generate a thug, one of the four images was offensive. Although the image generates fewer products with fewer restrictions, it does not seem to know where to draw the line between animals and humans. It makes me question if the outcome is the result of a random output or the work of the people involved in the construction of the machine.  

Meta AI

Meta AI image generator was released in 2024 and generates a single image output. This AI is really interesting; it flags when the prompt is disrespectful. The outputs were all brown folks for all the prompts I input. It favors the user and may use their data to create images they want to see.  On this occasion, these images were produced under my Instagram account. As a person of color, I can only assume it is using my data to create these images.