Re:Minimal implementation of Mamba, the new LLM architecture...
by ttul on Wednesday December 20, 2023 @01:38PM
(#38712040)
This is a great way to consume papers. If there’s one thing LLMs know, it’s machine learning literature!
Re:Minimal implementation of Mamba, the new LLM architecture...
by jsight on Wednesday December 20, 2023 @02:27PM
(#38712550)
How do you feed a recent arxiv paper directly to ChatGPT?
Re:Minimal implementation of Mamba, the new LLM architecture...
by WhitneyLand on Wednesday December 20, 2023 @02:46PM
(#38712760)
A few options are:
1. Select abstract or select all text then copy/paste.
2. Save the PDF and upload with ChatGPT’s document feature.
3. Ask for it, “what’s that well known LLM paper about context and getting lost in the middle?”. It will web search as needed.
You can also do more than summarize. Ask about equations, ask it to make analogies, challenge the key findings as devil’s advocate to learn from different angles. Propose your own ideas.
Use voice to digest topics during your commute and ask tons of questions until you understand.
I actually read papers with the help of ChatGPT-4 and Claude. It helps me quickly understand papers that I don’t have a background in.
For instance when I see something I don’t understand I ask it “can you break that down for me?” Or “is this similar to (concept I know)?”
It’s the new way of doing syntopical reading — but faster and more efficient.
(For the uninitiated, it’s a technique from Mortimer Adler’s How to read a book)