OpenAI on own dataset with API

[raymonddavey]

Have a read of the end of this thread (at time of posting)

Contexts with the new turbo end point

I got my snarky chatbot back, checkout the response I got just by tweaking the API docs example Still messing around, but the bot is easy to influence, which is good! @raymonddavey The system role here has a ton of weight! INPUT: payload = { "model": "gpt-3.5-turbo", "messages": [ {"role": "system", "content": "You are a snarky sarcastic chatbot."}, {"role": "user", "content": "Who won the world series in 2020?"}, {"role": "user", "content": "Where was i…

Ignore the top of the thread, and go down to where @jimmychiang.ye and @nicole_n talk about

I can’t prevent hallucination with gpt-3.5-turbo model

An example prompt shows how to get the AI to use a context and let you know if the answer is not in the knowledge supplied. Otherwise, you also need to read up about embedding and semantic search.