LogFAQs > #978980052

TheGoldenEel posted...

On the other hand maybe an ai that collects images of people and aggregates them just inherently believes humans are racially diverse, because they literally are

That's not really how these systems work though. Like, a prompt involving Nazis is presumably going to use media from Nazi Germany and not every picture of every person on the planet.

Edit: I'll also point out that for the system to be able to tell you users that is unable to generate from a certain prompt, they have to be doing some degree of preprocessing to identify words or phrases that they want to avoid. It's not like they're handing the generative AI a prompt and it's coming back with "I'm not comfortable doing that." There is very definitely a manual process in between the prompt submission and triggering the generative system.