I was exploring how ChatGPT could be used to generate images for social media to show support for different causes. Interestingly, it refused to make any images involving Sudan, but had no problem doing so for North Korea, Israel, or Palestine. I asked it why, and it said the "perceived acuteness" of these situations was different, which seems very subjective. When I asked whether it recognized the inconsistency in this application of policies, it said yes and suggested sharing this feedback with Open AI. I genuinely found this interesting and am curious about the technical explanation, as there are many similar concepts and situations I could imagine this will pop up in. This is not meant to be political at all, just something interesting I wanted to note.
#4o makes images of some humanitarian conflicts but not others.
1 messages · Page 1 of 1 (latest)