I’m currently experimenting with text-davinci-003 to do some simple language translation to Spanish. There’s no conversation, just translate an unconnected phrase, then done. It doesn’t need any updated knowledge about the world.
I see text-davinci-003 is going away soon. Any suggestions on what I should use instead?
The -instruct model is the replacement, in that the only thing one needs to change for continued use of completion endpoint is the model name, and then you get an AI that can be instructed in four tokens and doesn’t chat or backtalk: