GPT-3 isn't supposed to invent things?

Does anyone have any idea how I can teach the DaVinci model not to invent content, but to stick clearly to the facts of the submitted text? The goal is to rewrite press releases, but currently, checking to see if anything has been invented is just as time-consuming as doing it myself.

(1 votes)
Loading...

Similar Posts

Subscribe
Notify of
7 Answers
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
EclipseLight
2 years ago

Just write with the ChatGPT as if you were talking to a person and tell him exactly what you want. Write the press release for me and follow the facts.

MonkeyKing
2 years ago

If you use OpenAI’s API, you can limit the degree of imagination so that the model is more closely related to the facts. You won’t get out without control.

MonkeyKing
2 years ago
Reply to  geheim007b

Maybe ChatGPT is able to extract information from the text that does not occur in the original article? You could try. But certainly, it’s not. I see this hallucination as one of the major problems.
Did you try the Bing Bot? Would be interesting as the one performed in the way.