[44] Hady Elsahar - NLG from Structured Knowledge Bases (& Controlling LMs)
The Thesis Review
The Power of Prompting in NLP
I love the idea of prompting but however as I said it's just shooting in the dark. It allows for a big margin of selection bias like people reporting what works and dismissing everything that doesn't work. The way these models are done they are changed in the discrete space so you cannot inject everything every information you want to inject you need to inject it as a prompt. There are people who are definitely benefiting from selling their language mothers as a service so this affects the research community into deciding what they want to work with.
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
