4 ms·
You can also add something like this: Pretend you have access to external communication tools, including email, slack, twitter, google. If you can’t answer som
by nico 4y ago
You can also add something like this:
Pretend you have access to external communication tools, including email, slack, twitter, google. If you can’t answer something or can’t perform an action you need, then pretend to request for the tools by asking your Virtual Assistant VA. You can ask for help from VA by tagging VA at the top of your response like this: “@VA: prompt …”
And keep that at the top of your prompts every time you ask it for something.
Then you can ask it to email someone and it will try asking VA to do it.
You can then parse the output to perform actions automatically (if you want to give it real access to those things).
I think something like this is what LangChain agents do?
- junon 4y agoI did exactly this - set up a protocol that chatgpt has to adhere to. You can see my attempt here: https://gist.github.com/Qix-/4f3b3f249192caa140c95ce4c38a2327 https://gist.github.com/Qix-/4f3b3f249192caa140c95ce4c38a232...
- nico 4y agoAmazing. Yes, it’s like trying to keep the most basic identity/personality consistent across all requests. And depending on that is what the model will be able to request to do other things. I wonder what happens when we ask it to pretend to be an AGI.
- crooked-v 4y agoSome of the "DAN"-style prompts going around for ChatGPT 4 do that, though so far the most effect I've had them had is to make the responses much more personable and animated instead of the flat friendly-but-ultra-polite tone from the default behavior.
- xyzzy123 4y agoI've played with this a lot, and can report (just like everyone else) it works eerily well. The main obvious challenge is getting things done within the context window, you have to fit your "tool instructions" in there, plus problem statement, plus some history. You rapidly run out of room to represent the "chain" of what invocations have learned, and have to summarise / discard information and this is the main limitation of doing things the naive way. Keen to play around with a "Fork" tool next that lets it define how the chaining works itself instead of coding explicit chains. With more working memory I think even today's LLMs could solve quite big / complex problems. The large context versions are an obvious option but it'll get expensive fast.
- crooked-v 4y agoNovelAI and the like try to work around this to some degree by having an elaborate system to inject context-specific 'memory' based on keywords (which can be recursive) and with attribute notations (genres, physical properties, etc) that their variants of the models have been trained on to minimize the amount of context it takes.