@AndrewYNg Why would you do the planning with an LLM instead of a real planning algorithm? The LLM can certainly help formalize the intention of the user based on their input but why ask it to do every part of the process? It’s a weird constraint. Symbolic AI does better.
Interesting trend in AI: the best results are increasingly obtained by compound systems, not monolithic models.
AlphaCode, ChatGPT+, Gemini are examples.
In this post, we discuss why this is and emerging research on designing & optimizing such systems.
https://t.co/tfnNuoTNNY