@rentacowisgoogle

Some tips that have worked for me:
The language model has a hard time with negatives so try to always tell to it DO something instead of, rather than don't use X.
Elaborate on... is a good prompt to get it to start spit balling. example: "elaborate on why the concatenation wouldn't make it into the next prompt?" got me moving again after a hang up.
Try using "From the perspective of X" if you need it to suggest different things.
Of course it can't actually "audit" code. However, when I've asked it to "Compose a plausable comprehensive audit report of this code." it isn't too bad at pretending by giving a reasonable list of recommendations. Then tell it to take it's own advice :D
Last one is my favorite.
For explanations I usually just propose a topic with: "Explain in great detail X. Answer unambiguously.  Thoroughly explain the reasoning." possibly even "Respond with absolute maximum verbosity." if you don't want it leaving anything up for interpretation :P
I love it because It can be pretty hilarious when you propose a topic that no normal human would even want to consider trying to explain, let alone the reasoning as well.

@rentacowisgoogle

I had it walk me through making a console app in python so i could "chat" directly with GPT-3 over the API. It took about a week, a few rewrites (well, rethinking how to explain my code requirements accurately enough). Now I'm burning through my trial buxx, and will likely buy some more because face it sometimes you simply need the text without the guilt trip about morality and ethics in AI from ChatGPT, especially when you're specifically wanting to explore the models full capabilities.

The script is on pastebin right now, along with a VERY hefty beginner programmer level walkthrough explaining how it works. And a compare and contrast essay for GPT-3 API interactions vs. chatting with ChatGPT. No, I'm not joking... Tomorrow I'll make it spit out a comprehensive code audit report as well.

I KNEW NOTHING, when I started this other than: If you don't think it will be able to "figure out" what you mean, just ask anyway.

@DenaTollefson

Hey Will! I had heard a bit about ChatGPT but don't know many specifics and did not know the drawbacks and limitations. Your video is great! It was amazing how ChatGPT could whip out a sonnet LoL!  The whole topic of AI is fascinating, I know a lot of artists feel it is controversial, I agree with you AI can be used a tool, but not a replacement for humans

@rentacowisgoogle

How do you feel now that a valid debugging technique is: "Walk me through this code step-by-step like this is an episode of the magic school bus and you're Ms. Frizzle giving a plausible explanation to the class in extreme detail precisely what's going on while also using fun and relevant examples."