As an experiment, I once asked ChatGPT for details of a TV episode I already knew the contents of. After it provided false information, I corrected it. The algorithm then repeated its mistake several times, providing more false information.
Two theories:
1. The AI's knowledge hasn't been "trained" on this subject, and it lacks the necessary information to give a correct answer.
2. The algorithm is purposely obfuscating to avoid spoiling information on copyrighted material.
When you ask them questions, AI will form answers based on a composite of multiple sources—including false info, opinion pieces, and even speculative or fan-made content.
I look at them sort of like elderly relatives with dementia, lol.
That being said, sometimes I'll copy-paste an internet recipe into ChatGPT or Gemini. "Is this safe to use, or will I be temporarily blinded by the horseradish and mustard it calls for?" Sometimes, they'll make adjustments to ingredients, or suggest a garnish to elevate the dish.
A few times, I've created recipes from scratch, using my own inspiration. They usually turn out amazing.