A glimpse into a potential future for LLMs:

Exciting to see that you can get AI at Meta's Llama 2 to call functions just like OpenAI's functions feature using llama.cpp's newly added grammar-based sampling. And it'll be even more exciting to see what happens once this fits on consumer hardware!

Answering by approximate retrieval or by understanding+reasoning are two ends of a spectrum. Humans are at various places on this spectrum, depending on the task, experience, and depth of understanding. We see this in physics or math students: some will study very hard, do lots

An awesome new package Conformal Tights adds conformal prediction of coherent quantiles and intervals to any scikit-learn regressor. 🔥🔥🔥🔥🔥 Features: 🍬 Meta-estimator: add prediction of quantiles and intervals to any scikit-learn regressor 🌡️ Conformally calibrated:

“Tank, load the coffee program.” youtu.be/e747SIBmWPE

“I don’t even see the R’s. All I see is 302, 1618, 19772, 198, 3504, 1134, 19772, 198, 101830, 198, 138322, 198, 1100, 302, 1618, 19772, 25644, 1100, 3504, 1134, 19772, 1100.”

"Europe needs regulatory certainty on AI" An open letter signed by Mark Zuckerberg, me, and a number of European CEOs and academics. The EU is well positioned to contribute to progress in AI and profit from its positive economic impact *if* regulations do not impair open

I'll get straight to the point. We trained 2 new models. Like BERT, but modern. ModernBERT. Not some hypey GenAI thing, but a proper workhorse model, for retrieval, classification, etc. Real practical stuff. It's much faster, more accurate, longer context, and more useful. 🧵