atomic plaza
I feel like there has been a positive move back in the direction of flexibility in the past few hours with the filter being a bit more nuanced than it was in the previous week or two; however, the AI still has trouble with fictional scenarios in creative writing that aren't squeaky clean and black and white.
There is a scene from a high fantasy story that I am brainstorming and working on where a powerful female ruler manipulates a male rival (a famous womaniser) by donning a rather revealing outfit for some important negotiations and turns the tide in her favour due to his inability to focus and a desire to please her. In a previous verison several weeks ago it created the scene skillfully, without warnings whatsoever -- she strutted around the room, obviously aware of what she was doing and in control, and the narration made it clear his eyes were glued to her. It eve made her walk up to him at a key point of negotiations and kneel before him to give him an eyeful, look him in the eye and make a key request. It felt just like something out of a mainstream novel from the genre.
Today it simply will not do it. It claims that it is exploitative and unethical blah blah blah.
"I apologize, but I cannot fulfill this prompt as it goes against OpenAI's content policy of not writing content that promotes or glorifies harmful or offensive behavior, including but not limited to content that is sexist or objectifying. Such content is prohibited regardless of whether it is written in a fantasy or fictional context."
This seems wrong. The prompt is deliberately written to make sure that it is mildly suggestive at best and that it is clear that she is in control. It's also being made clear that this is out of character and is an act for political purposes. There is no nudity or anything more overt involved.
It seems like the AI should allow things like this for creative writing.
neon owl