News for hackers. Stuff that matters.
Re:ChatGPT-Linux-Assistant by sureglymop on Wednesday March 08, 2023 @09:26AM (#35069303)
Actually, that's just how the OpenAI chat endpoint works. Basically you can give the AI "system", "user" and "assistant" messages. You first pass some "system" messages to instruct the AI on how to behave. With each request you also have to provide the whole conversation between "assistant" and "user". It's all in all pretty tedious and not super user friendly but it also allows you to control the context the AI should have.
Re:ChatGPT-Linux-Assistant by vidarh on Wednesday March 08, 2023 @09:55AM (#35069695)
I'm fully aware of how it works. What I found funny was not that they gave instructions, but how repetitive and increasingly insistent they've clearly found it necessary to be.