🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

If you are a PhD candidate interested in the interesection of #reinforcementlearning, #llms and #Biology, my colleague Juanjo Garau is looking for interns as well!

I am looking for PhD research interns at the InstaDeep Paris office for the coming summer. Join us if you are interested in DNA LLMs and regulatory genomics! We have different projects available and are flexible to find one that fits your PhD. Link: instadeep.com/internship-off…

CodeWhisperer team hires a few interns to work on a broad topic of code generation. We are an applied team publishing some of our work. Our product is a Copilot-like system that is GA now. If you are interested, please, send a resume/CV to lboytsov at our company domain.

We released the weights and downstream task datasets of the agroNT, our 1B parameters DNA foundation model for plant genomics 🧬🪴. Check out our github (github.com/instadeepai/nu…) for Jax lovers⚡❤️and our Hugging Face space (huggingface.co/InstaDeepAI) for the pytorch version🤗.

The things you learn on the Internet! 😂. Amazing thread by Latif Nasser on Quasimoons.

This painting is 115 years old. It's by Zinaida Serebriakova, a remarkable artist with a remarkable life involving both World Wars, the Russian Revolution, Paris, and personal tragedy. This is her story...

🧵Now it's officially published as a Transactions on Machine Learning Research paper. 🥮🍾🎈 I am very grateful for TMLR to exist for reasons that I explain in this thread. 1. We reproduced an important method using only OSS generative model whose training data & process was well-documented. ⏭️

A couple of years ago, we saw something weird when we ran Auto-attacks on models with front end defenses - Auto-attacks fail when applied to front end defenses. Leo Boytsov explains more in the thread below...

🎲Excited to release the first version of ⚡️cfrx⚡️, a lightweight open-source library designed for efficient implementation of counterfactual regret minimization (CFR) algorithms using JAX. github.com/Egiob/cfrx

Congrats to NYU's very own @NivCohenHuji and team! SaTML Conference NYU Tandon

Phenomenal evening yesterday ParisAI 🇫🇷 The French tech is on the rise and InstaDeep is playing its part! 🚀 Kudos to my fellow speakers, Camille Bouget of Scienta Lab, Gabriel Synnaeve (Gabriel Synnaeve) of AI at Meta and David Pfau (David Pfau) of Google DeepMind, for their inspiring

In 2019, a new drug called Trikafta came out. Within hours, people started coughing up all the mucus in their lungs. This was "the Purge." (Btw, I didn't get into this but the Purge wasn't just from the lungs. Think about all the other places in your body that have mucus...)

✨ Meet SegmentNT: the first-ever LLM capable of annotating DNA sequences at single nucleotide resolution. Built on top of our Nucleotide Transformer, SegmentNT offers precise genome annotation surpassing traditional methods and can offer deeper insights into our genome 🧬

I’m happy to present SegmentNT, our new genomics LLM annotating DNA sequences at a single nucleotide resolution! 🔎🧬 Building on our previous work on the Nucleotide Transformer (NT), we’ve designed SegmentNT to process sequences as long as 50,000 nucleotides (50 kbp) to predict

i’m happy to share that my friend Karim Beguir will join us as a speaker at RAAIS 2024 i met karim thru the brilliant Shakir Mohamed (who also spoke at raais!) in late 2018 InstaDeep has gone from a small ambitious underdog team to a leader in ai with BioNTech SE join us!

I encountered a delightful little astrodynamics proof last week when an astronaut casually stated it as a fact, and when I looked skeptical, he just smiled and said, "Check it yourself." Here's the statement: "It takes 2 hours to orbit at the surface of any object made of rock"

🎊Big News! In collaboration with Google Cloud and Google Cloud Partners, we’re launching DeepPCB Pro– the advanced version of our AI-powered platform for PCB design!🎯Join us on Sept 12th, 4:30 PM PDT at Google SF to be part of the future of hardware. RSVP here:

AI skeptics: LLMs are copy-paste engines, incapable of original thought, basically worthless. Professionals who track AI progress: We've worked with 60 mathematicians to build a hard test that modern systems get 2% on. Hope this benchmark lasts more than a couple of years.

Crazy finding: In (some?) neural networks there exist a super-weight whose pruning can destroy model's ability to generate text. One should really wonder if this survives through reproductions. arxiv.org/abs/2411.07191