>If AI is artificially limited from considering women in comedic situations it will end up having unpredictable results when the model will have to consider women in comedic situations as part of some other task given to AI.

So one thing I will note now, just because AI is blocked from giving you a sexist joke doesn't mean it couldn't have trained on them to be able to understand them.

>An example would be if you were to have AI solve crime situation, but said situation would have aspect to it that included what humans would find comedic.

This feels like a very flimsy example. The AI is now employed as a detective rather than a chatbot, which is very much not the purpose of the ChatGPT but sure. Now ignoring like I said that the AI could be trained on sexist jokes and just refuse to make them, I still find it unlikely that understanding a sexist joke is going to be overly critical to solving a crime.