RE: LeoThread 2024-10-31 10:28

avatar

You are viewing a single comment's thread:

wait so this is how prompt engineering works, you first have to tell gpt who it is before instructing it? What a trick 🤔

leofinance
0
0
0.000
5 comments
avatar

You don't necessarily need to do it exactly like this, but yep, it's pretty neat. A more general version of this "trick" is to simply aim to provide the chatbot with as much context as possible.

For instance: Let's say you have a meeting transcript and want to generate a meeting summary based off of it. You could probably get away with a prompt like

Based on the attached transcript, write me a summary

But you would probably get much further with something more like this:

https://inleo.io/threads/view/mightpossibly/re-taskmaster4450le-24wdbzaae?referral=mightpossibly

0
0
0.000
Reply
avatar

I never gave gpt prompts longer than 2 line. I n. The more data you feed it with your instructions the better the results. Thank you for the lessons I'm going to practice right away

0
0
0.000
Reply
avatar

Awesome, I'm very happy to hear that. Do share whatever interesting stuff you figure out with us :) Also, feel free to tag me if you have questions or issues about anything in this regard. I'll help if I can

0
0
0.000
Reply