Thanks for all the warm wishes, everyone! I couldn’t write my yearly AI research review, but here’s my list of bookmarked papers you might find useful: https://
magazine.sebastianraschka.com/p/llm-research
-papers-the-2024-list
… And no worries, I’m slowly but steadily recovering & already starting to feel bit better every day
@rasbt
-
2024 AI Research Papers Compilation and Recovery Update
By
–
-
12 Years of Python, ML, and AI Content Journey
By
–
Hey everyone!
I just wanted to say how incredible this journey here has been over the past 12 years, when I started sharing and creating Python, ML, and AI content. The positive feedback I’ve received from you all has been such a powerful motivator. I had some exciting plans -
Contribution guidance: Referencing Alpaca and LIMA code implementations
By
–
Thanks for your interest in contributing. I unfortunately currently don’t have access to a computer until likely next week (long story) but off the top of my head and as far as I can tell right now, I think you can copy most code from the Alpaca or LIMA (
https://
github.com/Lightning-AI/l
itgpt/blob/main/litgpt/data/lima.py
…) -
Understanding Multimodal LLMs Latest Papers Overview
By
–
I covered the latest multimodal LLM papers lately (but not the evals, as this would be a different beast): https://
magazine.sebastianraschka.com/p/understandin
g-multimodal-llms
… -
ML Students and CI Pipelines Using Common Practices
By
–
Every student of every ML class but also many CI pipelines, research baselines,…
-
Token Caching Challenges in Large Language Models
By
–
Not a dumb question at all. I think caching is the trickiest one here (obvious ones like kv-caching aside). Caching token embeddings of common words probably don’t really help much. And prompts are probably often diverse enough that caching those would be too expensive. Session
-
Serving ChatGPT to 100M users with 1k H100s efficiently
By
–
Load balancing, caching, dynamic batching… with 10% DAU you probably don’t need much more than 1k H100s to serve ChatGPT to 100 million customers.
-
AI as Brainstorming Aid for Medical Diagnosis Exploration
By
–
They don’t have to blindly rely on the results but could consider it as an additional brainstorming aide. I think the most interesting use case is something like: “given symptoms xyz, what could be the possible causes to further look into / follow-up with additional tests”
-
AI’s Potential Impact on Medical Practice and Doctor Adoption
By
–
And that’s the general purpose model. If you develop a finetuned version with and for doctors, it could have such a positive impact.
(Also, doctors not taking advantage of AI is kind of like doctors not consulting books; if used properly it doesn’t make sense not to) -
Deep Learning Jargon Explanations in Statistics Course Introduction
By
–
When I was a fresh statistics prof at UW-Madison & teaching Deep Learning class in the stats department, I had a bunch of “sorry for the DL jargon” slides about that in the intro lecture