How can we learn a foundational world model directly from Internet-scale videos without any action annotations? Yuge Shi (Jimmy), Ashley Edwards and Jack Parker-Holder from Google DeepMind's Open-Endedness Team are presenting 'Genie: Generative Interactive Environments' at the UCL DARK Seminar

Huge pleasure to visit Tim Tim Rocktäschel and UCL DARK, and a great start for my London trip. Lots of interesting questions/points about LLMs and agents! And thanks indeed for the free pizza! :)

Great to be back with the first in-person UCL DARK Seminar Series this year with Associate Professor Stefano V. Albrecht (@UoE_Agents)'s talk 'From Deep Reinforcement learning to LLM-based Agents: Perspectives on Current Research'

Thanks Lakera for sponsoring this event!

I am honoured to have been awarded an Amazon Research Award for our proposal 'Compute-only Scaling of Large Language Models' (i.e. Q* before it was cool!). Thanks to Amazon Science amazon.science/research-awards and to my amazing students Eltayeb Ahmed & Jonny Cook!

#AmazonResearchAwards

Looking forward to my trip to London next week to talk about 'From Deep Reinforcement Learning to LLM-based Agents: Perspectives on Current Research', starting at UCL UCL DARK on Monday 29th. Trip details and event links in posts below.

Stefano V. Albrecht (Stefano Albrecht (UoE Agents Group)) from The University of Edinburgh is the next speaker of UCL DARK speaker series!

Title: From Deep Reinforcement Learning to LLM-based Agents: Perspectives on Current Research

Sign up: eventbrite.co.uk/e/ucl-dark-spe…

📌 90 High Holborn, UCL Centre for Artificial Intelligence
📅 Mon 29 Apr, 16:00

We're excited to announce that the Genie Team from Google DeepMind will be our next invited speakers!

Title: Genie: Generative Interactive Environments

Speakers: Ashley Edwards, Jack Parker-Holder, Yuge Shi (Jimmy)

Sign up: eventbrite.co.uk/e/ucl-dark-spe…

📌 90 High Holborn
📅 Tue 30 Apr, 17:00

Save the Date!🤖1st of May 17:00

Next seminar will be with Jake Bruce Google DeepMind presenting Genie! a foundation world model that generates action-controllable 2D worlds.

📍Imperial Computing, hybrid event with in-person networking. Registration & details
eventbrite.com/e/imperial-col…

Actually the accept rate decreases monotonically with number of 1st author submissions: the more prolific the first author is, the lower the quality of their paper.

🤯📽️🏃Annoucning controllable video generation workshop at ICML2024! Super excited for speaker line up, and looking forward to seeing submissions exploring controllability in video generation.

Submission deadline is 31st May AOE (but why wait till then if you can submit now?!)

The amazing Ashley Edwards will talk about Genie at @RAAIS 2024!

Diffusion is an extremely powerful and general purpose approach - here we combine it with _policy guidance_ to improve the distribution mismatch in offline RL, which in turn offers the chance to bring RL to the real world without having to collect online data.

Super exciting to see improved techniques for generating synthetic data for agents! Awesome work from Matthew Jackson and team, plenty more to be done in this space 🚀🚀🚀

This is also a great description of research and producing research breakthroughs.

This is by far the best non-technical Natural and Artificial Intelligence book anyone could read. This comprehensive, well-researched, crisply clear, sharply focused and illuminating book is a thing of beauty. It is the book I wish I had had when I started my AI career 30 years

'Language models can autonomously identify and prioritize domains rich in knowledge, optimizing their storage capacity.' — that's a really interesting finding.

This is art.

If you are attending ICLR 2024, make sure to check out Dominik Schmidt's & Minqi Jiang's outstanding work!

Team worked hard to address the feedback from the open community to improve the model. Kudos to Robert Dadashi and colleagues for the hard work. Let us know how it is.

Watching AIDE autonomously tackle ML problems by designing, implementing, and iterating on code, conducting experiments, and evaluating results has been nothing short of fascinating. Can't wait to see how it will revolutionize DS and ML workflows, enabling a broader range of