@ramachandranesk@FiftyTwoDotIn Don't think I understand the article - the implication seems to be that a lot of researchers are spending time looking at sanskrit/rule based systems instead of following the deep learning approach, but doesn't concretely make that claim or substantiate it?
Transformer Puzzles
- (https://t.co/tel0v1v7uZ)
7 short visual puzzles for learning how Transformers can compute basic algorithms. Based on Thinking Like Transformers.
(Presented at ICLR Blog today. Doing these has been one of the most useful ways to grok how LLMs compute)
Getting cloud GPUs is remarkably annoying. Got a service quota increase request denied from aws and my account flagged by paperspace because "i don't use it enough"
Really fun work I did last semester got accepted to ACL. https://t.co/PHmEpjh1nR in collaboration with https://t.co/j5lVKKGjiX /Alex Feng/@LangTechLara and Chris Callison Burch.
I recreated OpenAI's Todo Chatgpt plugin tutorial using @FastAPI and wrote some documentation for fellow ML engineers that haven't touched web stuff in a while. https://t.co/FtrnOo62nF
something I found kinda cool from the GPT4 whitepaper is that RLHF doesn't seem to be as powerful as I thought it was, it definitely makes a difference, but I was overestimating its utility. Puts LLama in a much better light.
https://t.co/oWGVxm3nYp (page 27)
Friend: I’m listening to a talk about grounding transformers for safety purposes
Me: ohh like AI safety?
Friend: like preventing fires
Me: so “grounding transformers” means…
Friend: like actual ground wires to actual electrical transformers