I treat it like a *very* knowledgeable intelligence that gets things wrong about as often as humans, probably a bit less. I don’t get why people scoff that at, we all make mistakes.

chatgpt is best on the first prompt. If you can give it as much info as possible: entire source code, the data that was produced from the code and what was expected, etc, you will see its true power. People rarely experience it because they don’t treat it with the respect it deserves.

Reply to this note

Please Login to reply.

Discussion

So true. To get good output, you have to put work into the input.

Good point. Every tool is only as good as the person doing it. Act as ChatGPT's manager telling it a task to complete. It would be no different than if you poorly explain a task to complete to another person and they did it wrong.

nostr:npub1xtscya34g58tk0z605fvr788k263gsu6cy9x0mhnm87echrgufzsevkk5s I’m curious if you can share a tip with getting chatgpt to assist with a codebase that exists after it’s knowledge training cut off date?

Last I checked, they disabled, the web browsing function, so I’m curious if you are using a certain grouping of extensions, and if you use the API to get a larger context/token capability, then the default open AI web interface?

For instance, I would like to use gpt4 for to help analyze parts of the nostr protocol and it does not appear to have any training data on that so I would need to probably paste in large chunks of the code. I wonder if there’s an extension that you use to help?

#respectfordave