Is it just me?
RAG seems more like a workaround than an ultimate solution. It mitigates the problems of LLMs through retrieval (or search) without solving them.
These are some of the directions I think RAG can take:
1) Retrieval or search will continue to be critical, but adapting them to LLMs instead of humans could be interesting
2) LLMs (agents) learn to use tools such as retrievers instead of trying to solve all problems by themselves
3) Try to introduce stronger reasoning mechanisms to LLMs than the current seemingly shallow reasoning capacities
Thoughts?
We just summarized a 2 hr+ intro Harvard lecture with our new Share feature so that you don't have to.
Click the link below; no sign-up, no commitment required 👇
https://t.co/5JbDVVw642
Introducing @whiteboardai - ✨Video AI Tutor✨
We just super-powered our intern's 2-hour calc lecture:
- 🧠 Get instant answers with AI
- 🕐 Find what you need w/ magic timestamps
- 🎥 Works with YouTube, Zoom, & more!
- 💅 Sexy formatting and colors
🧵(1/4)
Can you believe that on average, more than 20% of students say they skip classes to make time for other activities?
We certainly do.
That's why Whiteboard is here to help you speed through your lectures with personalized help and review! 🧠🕑📚
https://t.co/EV0jIGS40w
Estou testando uma ferramenta gratuita chamada @whiteboardai. É tudo de bom.
Você cola o link de uma palestra do YouTube / Zoom, e ele gera esboços, palavras-chave, e tópicos de estudo.
Claro que usei uma palestra do @JonasMadureira para testar. Saca só: