Questions from a beginner.

Hi! Before, sorry for my English. I’m using the translator.

I’m trying to understand the logic before going to the API.
So far I have understood that it is important to provide as much information as possible via prompt so that the attendant knows how to behave and what to respond.

I also saw that there is a possibility to send a file with as much information as possible. For example, submitting all the content on my site so that AI knows how to answer everything about my site.
Through playgrount I listed a series of relevant information and davinci was able to answer perfectly (the others left the topic or gave very incorrect answers, ignoring the instructions).

But I have a question, which I still can’t understand the logic.
As I understand it, AI learns from conversation. But I didn’t understand how she learns from the conversation. Does she read only written on text/answer? In other words, with each new message I must send the entire conversation so that she understands the context and doesn’t forget what we said or does she read what is printed on the screen?

For example, the difference would be:
If it reads what’s on the screen, I could store the answers in the database and then just print all texts.
If she doesn’t read what’s on the screen, then does it mean that for every response, I should post everything we’ve already talked about and she analyzes the entire conversation to see how to respond?

As I’m still a little confused, I can’t be any clearer. So sorry if my text is confused.

Appreciate!

1 Like

Hi!
Thanks for the answer.
After all, I got it. The “completions” considers everything that we sent before. She/he complements the sentence following the pattern (hence the prompt training before). In other words, with each new post, we need to send everything we’ve talked about before so that she knows what it’s about.
I believe that now I have the logic and I can go to the API.
Thank you!

2 Likes