Training AIDungeon to give you answers is not the same thing as training GPT-3 in general. GPT-3 has an API where you give it a bunch of example prompts and answers and is not an entity that in general is interested in being told “you are X”.
I’m not sure I follow. Where does AI Dungeon come into this? Could you elaborate?
The GPT-3 answers I gave for “are ghosts real?” came from zero-shot prompting of OpenAI’s GPT-3 API (meaning no prior examples).
If you’re asking about the “You are a superintelligent AI that’s never wrong” trick, then the idea is that, by prefacing your question like this, you can get GPT-3 to write the text that the “superintelligent AI” would write, because GPT-3 thinks that’s the most likely continuation. GPT-3 is more likely to be right if it’s writing dialog for a character that’s always right.
People often give GPT-3 multiple examples of the task they want it to solve (multi-shot prompting), but wanted to keep things simple in the post. I’ll add some clarification there.
The finetuning scheme I proposed probably wouldn’t be as beneficial in the multi-shot setting as it would be in the zero-shot setting, but I still think it would be beneficial. Explicitly training GPT-3 to follow instructions also seems like a more straightforward way to tell GPT-3 that it’s supposed to follow your instructions than giving it enough examples that GPT-3 picks up on your intent. Working with GPT-3 would be far easier if we didn’t have to generate a list of examples of each task we wanted it to do.
I’m not sure I follow. Where does AI Dungeon come into this? Could you elaborate?
The times I have seen the “You are a superintelligent computer that’s always right...” trick was always in connection to getting AI Dungeon to do things.
Training AIDungeon to give you answers is not the same thing as training GPT-3 in general. GPT-3 has an API where you give it a bunch of example prompts and answers and is not an entity that in general is interested in being told “you are X”.
I’m not sure I follow. Where does AI Dungeon come into this? Could you elaborate?
The GPT-3 answers I gave for “are ghosts real?” came from zero-shot prompting of OpenAI’s GPT-3 API (meaning no prior examples).
If you’re asking about the “You are a superintelligent AI that’s never wrong” trick, then the idea is that, by prefacing your question like this, you can get GPT-3 to write the text that the “superintelligent AI” would write, because GPT-3 thinks that’s the most likely continuation. GPT-3 is more likely to be right if it’s writing dialog for a character that’s always right.
People often give GPT-3 multiple examples of the task they want it to solve (multi-shot prompting), but wanted to keep things simple in the post. I’ll add some clarification there.
The finetuning scheme I proposed probably wouldn’t be as beneficial in the multi-shot setting as it would be in the zero-shot setting, but I still think it would be beneficial. Explicitly training GPT-3 to follow instructions also seems like a more straightforward way to tell GPT-3 that it’s supposed to follow your instructions than giving it enough examples that GPT-3 picks up on your intent. Working with GPT-3 would be far easier if we didn’t have to generate a list of examples of each task we wanted it to do.
The times I have seen the “You are a superintelligent computer that’s always right...” trick was always in connection to getting AI Dungeon to do things.
I was using the API. The trick actually seemed to help a bit, but its responses were still inconsistent and not always “no”.