NEW blog post! Edwin Zhang and Eran Malach show how generative models in #chess can surpass the performance of their training data. Read more on our #KempnerInstitute blog and check out the preprint: bit.ly/3XotNqz #transcendence

This paper seems very interesting: say you train an LLM to play chess using only transcripts of games of players up to 1000 elo. Is it possible that the model plays better than 1000 elo? (i.e. "transcends" the training data performance?). It seems you get something from nothing,

New paper: How can you tell if a transformer has the right world model? We trained a transformer to predict directions for NYC taxi rides. The model was good. It could find shortest paths between new points But had it built a map of NYC? We reconstructed its map and found this:

NEW #KempnerInstitute blog: Rosie Zhao, Depen Morwani, David Brandfonbrener, Nikhil Vyas & Sham Kakade study a variety of #LLM training optimizers and find they are all fairly similar except for SGD, which is notably worse. Read more: bit.ly/3S5PmZk #ML #AI

Please apply and spread the word! These positions are pretty great, and it's a wonderful community to study the most exciting AI and neuro questions!

Learn more about the #KempnerInstitute’s GPU “co-workers” (and human ones) making this research possible: kempnerinstitute.harvard.edu/compute/ Max Shad Marinka Zitnik

I'm recruiting PhD students this year with interest in machine learning, embodied AI, or AI for science! If you are interested in constructing fundamental tools to improve Generative AI and exploring how these tools can be used for intelligent embodied agents and science,

1/5⚡Introducing Flash Inference: an *exact* method cutting inference time for Long Convolution Sequence Models (LCSMs) to near-linear O(L log² L) complexity! Faster inference, same precision. Learn how we accelerate LCSM inference.

Attention is all they need! #ginger #mango

We're hiring! Our Associate Director of Educational Programs oversees all aspects of the #KempnerInstitute's fellowship and training programs for undergraduate, post-baccalaureate, and graduate students. Apply today: bit.ly/3C4IV3P #AI #education Harvard University

I will be at #SC24 next week in Atlanta. Happy to chat about #supercomputing in #AI if you’re attending!

Attending SC25 this week in Atlanta? Be sure to stop by the MGHPCC booth to learn about the #KempnerInstitute AI cluster, and more about our educational programs & other opportunities. Hope to see you then! Max Shad Yasin Mazloumi bit.ly/4fPrDWN

We're here at SC25 this week! Stop by the MGHPCC booth to learn about our AI cluster and say hello to some of the #KempnerInstitute's research & engineering team. Yasin Mazloumi Max Shad #SC24

Compute power supporting Kempner Institute at Harvard University research! Now on TOP500 and Green500 list. #SC24

Eleven years ago, I named our first HPC cluster "Ludwig" to honor his contributions to science. With its computing power, we simulated complex fluid dynamics using Lattice Boltzmann Method! #science #hpc #fluid

Congrats to #KempnerInstitute's Marinka Zitnik & colleagues whose work is featured as a spotlight paper at #NeurIPS2024! Check out "Generalized Protein Pocket Generation with Prior-Informed Flow Matching" at the 11a session today. Zaixi Zhang Qi Liu buff.ly/4fgzlsp

A lot to learn from Kempner Institute’s Computing Handbook. Check out the materials on distributed #GPU computing.

Great opportunity - apply if you're interested or share with your network! Kempner Institute at Harvard University #Harvard #Kempner #AI

Learn distributed training from our GPU engineering expert Yasin Mazloumi and get the opportunity to earn an NVIDIA DLI certification! Kempner Institute at Harvard University NVIDIA #GPU

New in the #DeeperLearningBlog: Kempner researchers Nikhil Anand (Nikhil Anand) and Chloe Su (Chloe H. Su) discuss new work on how numerical precision can impact the accuracy and stability of #LLMs. kempnerinstitute.harvard.edu/research/deepe… #AI (1/2)