Twitter Beğeni Hilesi

There are good arguments for removing race from medical algorithms, but there may be unintended consequences. Our PNAS paper finds that race-blind algorithms can *worsen* racial inequalities, bc they can't adjust for racial disparities in data quality. shorturl.at/7ugW5

Extremely excited to share that ‘Machine Learning for High Risk Applications’ is now also available in Korean 🇰🇷. Thanks to the team at O'Reilly.

Machine Learning Street Talk has been dropping some seriously cool interviews back-to-back which I’ll highly recommend: 1️⃣ Prof. Subbarao Kambhampati (కంభంపాటి సుబ్బారావు) spilling the tea on ChatGPT’s reasoning skills (or lack thereof). 2️⃣ Sayash Kapoor on AI existential risk and government policy 3️⃣ Sara Hooker

👀

Sometimes good old common sense beats algorithm-powered "shortcuts".

Thinking about using an LLM evaluator? you can't miss this article by Eugene Yan! 🎯

Excellent paper reinforcing Goodhart's Law: When a measure becomes a target, it ceases to be a good measure. The authors unveil the "benchmark blindspot" where contaminated test data inflates LLM scores. Their solution? A "retro-holdout" dataset creation method, applied to

Finally got my hands on a book that’s been on my radar for so long 😀! Why Machines Learn by Anil Ananthaswamy beautifully blends storytelling with intuitive explanations of technical machine learning concepts, combined with interesting anecdotes about the journey from Rosenblatt’s

This is a great example of why writing publicly is so valuable. From my own journey, I can attest that writing allows you to share your learnings, build a network, and also learn along the way. Congrats Leonie 👏

This is a great introductory video by Yann Dubois on building LLMs. It covers often-overlooked aspects like evaluation, costs, and data, and not just architecture. It’s part of Stanford’s CS229 Machine Learning course for Summer 2024. Link: youtube.com/watch?v=9vM4p9…

The debate between open-source and closed-source AI has been gaining a lot of attention lately, but H2O.ai has been pioneering open-source AI for over a decade. Our CEO shared a great blog post in April this year, where he explains our long-standing commitment to open-source

We're providing free access to OpenAI's new o1 reasoning model through our WildChat chatbot: 🔗huggingface.co/spaces/yuntian… Also, proud that WildChat was referenced in the safety evaluation for o1!

Great analysis on OpenAI’s o1 models. Twitter/X is abuzz with chatter, but Delip Rao e/σ adds much-needed context to the conversation.

This book is a visual treat. It covers topics like embeddings, attention mechanisms, fine-tuning techniques, LoRA, quantization - all illustrated with colorful graphics that make it a joy to read.

I mean no disrespect to any of the authors in this thread but the claims being made are wild and/or taken out of context. The AgentClinic dataset is based on an online dataset (read contamination) I helped create, from a resource aimed at helping medical students study for the