May I invite you to a fun thread about a delightful quirk of relativity theory? Starting with a simple fact about rotations, I’ll hope to give you some intuition about something that’s considered wildly counterintuitive: velocity addition. Intrigued? Buckle up!

Unfortunate article in The Nation, the venerable US lefty mag, that appears not to have been fact-checked, so let’s do that fact-checking, shall we? There are 10 main claims made by author Paul Hockenos, so it’ll take a bit of a chonky🧵. But here goes: thenation.com/article/world/…

I’m thrilled to announce that Carbon Re has raised a $4.8m seed round! We’ve been sitting on this news for a while to sort out all the legal bits, but we can finally share it around 🚀 It will help us grow the team, serve more customers, and kickstart an AI research group! 1/🧵

The #LLMs for Code Seminar is launching this week! 🧠 Join us online: bit.ly/LLMs-for-Code-… #AI4Code #LLMs4Code

Heard someone say “I don’t want to waste brain space on learning Chinese” PSA—that’s not how it works at all. Consistently *retrieving* information both deepens connections with the rest of your knowledge and frees up resources & working memory for more abstract thought. 🧵

Myself and Kyle Lo have just released peS2o 🍃🎓, a collection of 40M open-access papers carefully cleaned for LLM training. V1 has been used by @mosaicml to train MPT, and we have a V2 version! Hugging Face page: huggingface.co/datasets/allen… feedback? github.com/allenai/peS2o/…

Today we’re opening access to Adept Experiments 🧪, a new way to explore the technology we are developing at Adept. Each experiment is a self-contained mini-tool or demo that showcases a part of our underlying tech. adept.ai/blog/experimen…

Catch me if you can! Announcing Llama-rephraser: 13B models reaching GPT-4 performance in major benchmarks (MMLU/GSK-8K/HumanEval)! To validate results, we followed OpenAI's decontamination method and found no evidence of contamination...🤔 Blog: lmsys.org/blog/2023-11-1… [1/n]

Want to run LLMs locally on your Laptop?🤖💻 Here's a quick overview of the 5 best frameworks to run LLMs locally: 1. Ollama Ollama allows you to run LLMs locally through your command line and is probably the easiest framework to get started with. Just use the installer or the

🚀 Today, we're thrilled to introduce Cove, a groundbreaking interface for thinking brilliantly with AI, and announce our seed round led by Sequoia Capital Cove is not a chatbot. It’s AI unconstrained by a chat thread or document editor. See it in action:

Big congratulations to the team Cove !! They've put a lot of effort into creating a unique way to interact with LLMs.

I'm so *wildly* excited to finally be able to share that I have joined Pacific Fusion and I'm starting an AI team! PF is the product of a rare alignment of a bunch of exceptional events. The result? A realistic path to fusion energy in years instead of decades. Story time! 🧵

We’ve added /llms.txt and Markdown to Stripe docs: docs.stripe.com/llms.txt Use the .md pages to quickly move Stripe knowledge into your LLM of choice. 📄

Introducing the Stripe Economics of AI Fellowship: The economics of AI remains surprisingly understudied. The fellowship aims to help fill that gap, by supporting grad students and early-career researchers with $, data, a conference, and community –

The Stripe product keynote just finished! We announced a lot of new stuff. (We actually couldn't fit everything into the keynote.) Some highlights: • We've built and deployed the first general-purpose Payments Foundation Model. Trained on tens of billions of transactions and

TL;DR: We built a transformer-based payments foundation model. It works. For years, Stripe has been using machine learning models trained on discrete features (BIN, zip, payment method, etc.) to improve our products for users. And these feature-by-feature efforts have worked

1/n Multi-token prediction boosts LLMs (DeepSeek-V3), tackling key limitations of the next-token setup: • Short-term focus • Struggles with long-range decisions • Weaker supervision Prior methods add complexity (extra layers) 🔑 Our fix? Register tokens—elegant and powerful

Curious about how LLMs work? I discussed our recent research, and how to apply circuit tracing to answer arbitrary questions about how models work. We also snuck in a demo of an open source tool you can use today, and shared tips about getting into the field! Check it out!