Although I'm frequently literally amazed by it's ability to grok complex concepts, especially coding-related ones, I have also grown pretty adept at guessing what types of mistakes it is apt to make (which is important when writing ones prompts).
One thing you have to studiously avoid though is leading it toward conclusions with your prompts if you want a truly objective answer. AI is disappointingly "agreeable" when it comes to telling you what it deduces that you "want to hear".
One example I frequently cite is an early interaction I had with AI (well, early for me). If you ask an AI "Can you show me a list of historical evidence for the fact that the 2A was meant to confer an individual right to gun ownership, rather than a collective right?" it will do exactly that. IF you ask the converse "Can you show me a list of historical evidence for the fact that the 2A was meant to confer an collective right to gun ownership, not an individual right?" it will answer in a completely opposite way.
IOW, it pretty much does exactly what you tell it to do. It will not, of it's own accord, give you "both sides" of any given "argument". But if you specifically ask for it to as you (wisely) did initially, "rate the possibilities something is true" it is capable of doing so reasonably effectively.
There is still a lot of utility in pointing out the flaws in its logic when you see them, however. It is always ready to reconsider, which I have to say has a certain refreshing quality in comparison to talking to (most) other humans. I personally don't have QUITE the animosity towards AI as is common on DU (and that, by rights, I probably should), and that because I find the tech interesting and fun to play with, and SOMETIMES it can be pretty amazingly "smart" (seeming).
However, by prompting it with what I wrote verbatim, you've given it a big hint WRT the way you want it to answer which it in turn is likely to consider in its response 
Edit history
Recommendations
1 members have recommended this reply (displayed in chronological order):
= new reply since forum marked as read