No Lies Detected: how to make your AI more useful

No Lies Detected: Metaprompting, Red Teaming and other tricks to make your AI more useful. Talk details: Metaprompting is easy, straightforward and it will work out the first time you try it. Right? Right?...Said nobody ever. Turns out instructing a generative AI model how to perform tasks, also called metaprompting, is equal parts art and science. In this talk we will explore my 5 proven steps of how to fail at metaprompting, and my 5 lessons to transform your technique from real bad to excellent. If you are new to this topic and want to learn how to create metaprompts, red-team, how to establish measurement and then mitigate, then this talk is for you.
#microsoftreactor #metaprompting #prompting #ai #redteaming
[eventID:23056]