Research scientist @DeepMind, here to see what's going on in the world and what @samuelpepys is up to
Joined November 2012
- Tweets204
- Following575
- Followers513
- Likes464
Cosmin Paduraru retweeted
Interested in using reinforcement learning to train LLMs for problems where there’s no room for error? Do you want to build massive data pipelines to transform how we interact with scientific knowledge?
We're hiring for multiple roles at Reliant:
apply.workable.com/reliant-a…
Cosmin Paduraru retweeted
Congrats team!
Thanks @TechCrunch for covering our $11.3M seed round, bringing next gen(AI) analytics to biopharma and beyond. techcrunch.com/2024/08/20/re…
Happy to have great investors on board with Tola Capital, @inovia and @mavolpi in additon to our amazing Angels from before.
Cosmin Paduraru retweeted
Love to build ML software?
Reliant AI is hiring in Montreal.
apply.workable.com/reliant-a…
This is such a powerful, emotional and important movie. Thanks Oz for everything you've done to tell the world what's happening in Ukraine
Really enjoyed working on this. Perfect problem fit for RL, technically challenging in many ways, and hopefully a step towards making tokamak fusion practical
Our paper on using RL for tokamak magnetic control has been recently published on the Fusion Engineering and Design journal. And while this is not about the latest LLMs, there are quite a few lessons learned on how to make RL work in applied domains
sciencedirect.com/science/ar…
Cosmin Paduraru retweeted
Student researcher position applications are open at Google Deepmind!
I'm hosting a SR in the intersection of bias and generative models. If you're an interested PhD student please reach out!
google.com/about/careers/…
Cosmin Paduraru retweeted
Today I am going to present our work on transformers for directed graphs in the Learning on Graphs and Geometry reading group.
Joint work with @liyuajia @DJ_Mankowitz @TaylanCemgilML @guennemann @CauseMean
Tomorrow in the reading group: @geisler_si will present his "Transformers Meet Directed Graphs" arxiv.org/abs/2302.00049 👌
Excellent for understanding Graph Transofmers better (and GT for DAGs :3)
Join on zoom at 11am EDT / 3pm UTC: m2d2.io/talks/logg/about/
Cosmin Paduraru retweeted
Are you interested in transformers and/or graphs and are at #ICML2023? Then visit me at poster session 1 (25 Jul 11 a.m.), where I present our @DeepMind paper Transformers Meet Directed Graphs.
Joint work work with @liyuajia @DJ_Mankowitz @TaylanCemgilML @guennemann @CauseMean
The transformer architecture powers recent AI tools like #ChatGPT or #GoogleBard.
In our @DeepMind #ICML2023 paper Transformers Meet Directed Graphs, we generalize transformers to more general inputs, namely directed graphs.
Here’s how we did it. 🧵 arxiv.org/abs/2302.00049
Cosmin Paduraru retweeted
The transformer architecture powers recent AI tools like #ChatGPT or #GoogleBard.
In our @DeepMind #ICML2023 paper Transformers Meet Directed Graphs, we generalize transformers to more general inputs, namely directed graphs.
Here’s how we did it. 🧵 arxiv.org/abs/2302.00049
Cosmin Paduraru retweeted
Our AI started with games. ♟️ But it didn’t end there. 🌐
Meet MuZero and AlphaZero, two powerful models which have evolved to transform computing itself. They’re already optimising data centres, improving the way we watch videos, and much more.
How? 🧵 dpmd.ai/optimising-computer-…
Our @Nature work on using #AlphaDev, an extension of AlphaZero, to improve the efficiency of fundamental algorithms such as sorting and hashing is out today! See @DeepMind post at dpmd.ai/alphadev-tw, and a few things I found interesting about it in 🧵
Our neural network was a relatively small transformer trained only on assembly code generated by AlphaZero during its search process, showing that AI can still produce breakthrough results without very large pretrained models
xitter.cf/DeepMind/status/166646… I would also like to thank everyone who contributed to this work. It was a privilege (and lots of fun) working with you all!
Work by: @DJ_Mankowitz, @andreamichi, @AZhernov, Marco Gelmi, @macselvi, @eleurent, @CauseMean, @shariqbal, Jean-Baptiste Lespiau, Alex Ahern, Thomas Koppe, Kevin Millikin, Stephen Gaffney, Sophie Elster, Jackson Broshear, Chris Gamble, Kieran Milan, Robert Tung, Minjae Hwang…
Great talk from @Jackstilgoe for #Pint23 on self-driving car policy, questioning underlying assumptions on inevitability & responsibility. Thanks to @SamanthaWork for hosting!
Cosmin Paduraru retweeted
How easy is it for adversaries to hide image content from classifiers through obfuscations? Our new benchmark allows you to evaluate this!
Dataset and evaluation code: github.com/deepmind/image_ob…
Paper: arxiv.org/abs/2301.12993.
Joint work between @DeepMind, @Google & @GoogleAI 🧵
Cosmin Paduraru retweeted
If you are at #aaai2023 come to today‘s workshop Deep Learning on Graphs (room 145B)! I will be presenting our work on transformers for directed graphs (poster 3:30 pm, contributed talk 4:30 pm).
Joint work with @liyuajia @DJ_Mankowitz @TaylanCemgilML @guennemann @CauseMean
Excited to share the details of our work at @DeepMind on using reinforcement learning to help large-scale commercial cooling systems save energy and run more efficiently: arxiv.org/abs/2211.07357.
Here’s what we found 🧵
🤔 So how did we deal with these?
The solutions required a strong core RL algorithm together with domain-specific additions such as sensitivity analysis, model unit tests, feature engineering, or heuristic action pruning (lots more detail in the paper).
🤝 This project has been a joint effort with great people from @DeepMind, @Google and @TraneCommercial, and I am grateful for the chance to work with them.
We will present this research at the @rl4reallife workshop at #NeurIPS2022 on December 3rd - looking forward to it!