🔥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

My thesis on exchangeable BRUNO models is now online: irakorshunova.github.io/phd_iryna_kors…  Huge thanks to the fantastic Arthur Gretton, Joni Dambre and Ferenc Huszár for making my PhD and BRUNO possible!

We wrote a blogpost on why so many machine learning algorithms are so hard to tune, and how an old Nips'88 paper shows how to fix it. It has some nice graphics to visually explain what is going wrong, and the proposed solution is simple to implement. engraved.ghost.io/why-machine-le…

When large sequence models are used for interactive speech or control, they can suffer from causal self delusions. In a new paper, our team explores the reasoning behind this & shows that it can be resolved by treating actions as causal interventions: dpmd.ai/sequence-model… 1/2

agadmator We have no plans to, but LC0 is the spiritual successor to AlphaZero and I'm sure will provide lots of great analysis of the games. Looking forward to watching the match!

Very happy to share my secret project for the last 3 years: we were busy getting AI to run a nuclear fusion reactor. Spoiler alert: it works! Deep reinforcement learning is pretty great at working with sci-fi things where human intuitions tend to break down.

This, my friends, is a Very Big Deal. And, as if by magic, fusion is one of the topics of conversation on the next episode of the DeepMind podcast, hosted by yours truly: deepmind.com/learning-resou…

A blogpost with a more readable overview of the fusion work. And pretty videos of what shaping plasma actually looks like! On the left, you can see a camera looking into the tokamak, while a single NN controller goes through the different phases of a shot. deepmind.com/blog/article/A…

When your research gets a shout-out on Nerdland after being a loyal listener for years! Lieven Scheire Jeroen Baert Hetty Helsmoortel en Poncho.

Feeling proud and honoured that our recent work on Deep Reinforcement Learning for Tokamak Magnetic Control, performed in collaboration with DeepMind, has been selected as Finalist for the Falling Walls Science Breakthroughs of the…lnkd.in/eQgpT5T9 lnkd.in/eQ5Gu5x6

This week, I've embarked on a new adventure as a research engineer @DeepMind☺️

I think it's cool to be fused w a machine and I like the idea of open sourcing all art and killing copyright

$ openssl dgst -sha256 nostradamus2024data.csv SHA256(nostradamus2024data.csv)= 6154d3cd8ef51e93d47fd64363fd69545be2b24ca95bacbe7c8192f1ff12d331 $ openssl dgst -md5 nostradamus2024data.csv MD5(nostradamus2024data.csv)= 6ff707736600c8d7354340d9ee35578d

Lots to look forward to Google IO tomorrow 👀

We’re sharing Project Astra: our new project focused on building a future AI assistant that can be truly helpful in everyday life. 🤝 Watch it in action, with two parts - each was captured in a single take, in real time. ↓ #GoogleIO

Gemini and I also got a chance to watch the OpenAI live announcement of gpt4o, using Project Astra! Congrats to the OpenAI team, super impressive work!

Over 2 years of our blood, sweat and tears and here we are: an interactive agent, situated, aware of its environment. Ears, eyes and speech. This is going to be TARS, this will be HER one day, becoming HAL, JARVIS, ...

Excited to announce the release of TORAX, a tokamak transport simulator from our Google DeepMind Fusion team! #fusionenergy - Open-source: github.com/google-deepmin… - Uses JAX: fast, differentiable - Easy coupling of ML-surrogates Hot off the press → arxiv.org/abs/2406.06718

Things never change: 2016: An End-to-end Differentiable Physics Engine for Deep Learning in Robotics, open source in Theano arxiv.org/abs/1611.01652 2024: End-to-end differentiable simulator of tokamak heat transport, open source in Jax arxiv.org/abs/2406.06718