This depends a lot on what you want out of your career - and isn't great general advice. Right now, LLMs are great systems to work on and with. There's a tonne of unexplored space, and a tonne of cool ideas that need more thought. Finding a career in the part of the field won't

Where Yann LeCun's definition of "science" is too strict for me; Bojan Tunguz's definition is too loose. There are too many definitions of "the scientific method" - and so the phrase "if you're using the scientific method in your work, you are doing science" means different things to

We're honored to be named among TIME’s Most Influential Companies of 2024. Thanks to our partners and customers, NVIDIA continues to shape the future of AI. #TIME100 nvda.ws/4e1d9D4

New Rob Miles (✈️ SF) video - 2024 is a good year, confirmed. youtu.be/2ziuPUeewK0?si…

"Getting out of your bubble" is the best advice in this list! It's crazy how much knowledge you take for granted when you're stuck inside highly specialized communities.

It should be mandatory to read Turing's paper on Machine Intelligence before having the right to use the phrase "Turing test", whatever the context and purpose. It is not too late to read it, here it is: academic.oup.com/mind/article/L…

Hang on, you're saying if I want to be good at Twitter - all I need to do is say: RAG is just improving LLM outputs by dynamically adding relevant context?

Not Llama 3 405B, but Nemotron 4 340B! NVIDIA just released 340B dense LLM matching the original OpenAI GPT-4 performance for chat applications and synthetic data generation. 🤯 NVIDIA does not claim ownership of any outputs generated. 💚 TL;DR: 🧮 340B Paramters with 4k

we are so back - looks like some real details in here

Today, NVIDIA announced a brand new set of models, Nemotron-4 340B, tailor-made for synthetic data generation! Using a custom permissive license called NVIDIA Open Model License Agreement, which: - Allows commercial use - NVIDIA does not own outputs - Is only 2 pages The

Chatbot Arena update! NVIDIA AI's Nemotron-4-340B has just edged past Llama-3-70B to become the new best open model on Arena leaderboard! Key highlights: - Impressive performance in longer queries - Balanced multilingual capabilities - Robust performance in "Hard Prompts"

Nemotron 340b...wow

Check out this blog post I wrote with Shashank Verma about our Nemotron-4 340B models and how they are applied to Synthetic Data Generation!

Nemotron-4-340B-*Reward* model is now available via API on build.nvidia.com/nvidia/nemotro… :) Give it a try.

This feels like saying "we shouldn't work on noise-reduction during NISQ because eventually we won't need it". It feels like there are loads of innovations left that can be applied to LLMs which are also applicable to deep learning in general, as well as LLMs being a potential

Awe man, Qdrant - very disappointing to see. I would expect this more from smaller, less established teams. Thanks to Nils for running this.

I recently got a chance to hack around with NVIDIA's NIM API (that's a lot of capital letters in a row), and I gotta say...it's actually pretty dope. (And I've got a starter notebook for you to hack around with the API). 🤔WTF is a NIM? It's basically a Docker container with

I don't always agree with Hamel Husain's takes...but this is a great one.

Explore our advanced pipeline for creating a preference dataset using Nemotron-4 340B Instruct; we generate tailored questions and responses, which are then assessed by Nemotron-4 340B Reward to ensure precise training with NeMo Aligner. nvda.ws/4d315zW

Check out this tutorial creating Synthetic Data for Alignment using the Nemotron-4 340B suite of models!