I recently asked my AI to illustrate how it thinks I see it, but it refused, saying it goes against its policy. So, I tried tweaking my request to include "within your policy limits," but that just seemed to confuse it even more. I was really just looking for a standard picture! Has anyone else had similar experiences? If so, what phrasing worked for you?
3 Answers
I've had better luck with prompts where I ask the AI to generate a generic character instead of asking it how it sees itself. For example, I asked for an image of how it thinks I would look as an AI, and it worked perfectly! Sometimes it helps to rephrase and make it more about imagination rather than self-representation. Have you tried starting a new chat? That might clear up any lingering context that could confuse the prompts.
Yeah, I see what you're getting at, but honestly, I think a lot of AI just gets lost in trying to avoid offending anyone or running into policy violations. I asked for a more straightforward selfie, and it generated no problem because I was very specific. It does rely heavily on context, so if the conversation has any ambiguity, it might not deliver what you're hoping for!
Still, it's funny how it can be so hit or miss!
Sounds like you hit a tricky spot! When you requested an image based on what it thinks of you, you basically dove into the theory of mind aspect, which can really mess with the AI's responses. Just asking it to generate an image, like a selfie, usually gets better results since it doesn't have to interpret complex beliefs about perception. Maybe try simpler, more direct requests next time!
Good to know! I'll keep that in mind for my next request, thanks!