charles blundell

[Pushmeet Kohli @pushmeet]

I have always believed that advancing mathematics is about empowering mathematicians.

From our earliest days at @GoogleDeepMind, our focus has been on deep collaboration with the mathematical community, pioneering new tools alongside one another to push the boundaries of what’s possible.

[lcamtuf @lcamtuf]

PSA: for security, put your AI agent in a water-tight sandbox, such as a dedicated VM. Once this is done, you can maximize productivity by giving it your credit card number, email credentials, the ability to write and run arbitrary code, and unconstrained access to the internet

[Google DeepMind @GoogleDeepMind]

Human generated data has fueled incredible AI progress, but what comes next? 📈

On the latest episode of our podcast, @FryRsquared and David Silver, VP of Reinforcement Learning, talk about how we could move from the era of relying on human data to one where AI could learn for itself.

Watch now →

00:00 Introduction
01:50 Era of experience
03:45 AlphaZero
10:19 Move 37
15:20 Reinforcement learning and human feedback
24:30 AlphaProof
29:50 Math Olympiads
35:00 Experience based methods
42:56 Hannah's reflections
44:00 Fan Hui joins

[Google DeepMind @GoogleDeepMind]

Join our VP (Drastic) Research, Gemini co-Tech Lead @OriolVinyalsML and our podcast host @FryRsquared as they discuss the evolution of our AI models, from AlphaGo to Gemini.

They also cover agentic capabilities and why giving AI access to tools could lead to a new era of problem-solving.

Listen now ↓

Timecodes:
00:00 Intro
02:30 Games and early AI agents
04:28 Weights
09:27 Architectures and a digital brain
10:24 Agentic behaviour
13:31 Digital body
14:09 Scaling
19:02 Data
20:59 Complex understanding and knowledge
25:14 Post training challenges
30:43 Reasoning
33:11 Planning
34:19 Systems 2
37:00 Memory
40:54 Gemini and agentic capabilities

[Ivan Fratric 💙💛 @ifsecure]

New Project Zero blog post by Sergei Glazunov and Mark Brand: Project Naptime: Evaluating Offensive Security Capabilities of Large Language Models https://t.co/txvkXH5oCC

[Theo Weber @theophaneweber]

Interested in a project at the intersection of large language models, self-improvement, reasoning + tool use, and computer security / code vulnerability? Our project is looking for a student researcher in that area (position is in London)! Please reach out and apply at https://t.co/5BtIWhXnbX

[Tim Rocktäschel @_rockt]

I am really excited to reveal what @GoogleDeepMind's Open Endedness Team has been up to 🚀. We introduce Genie 🧞, a foundation world model trained exclusively from Internet videos that can generate an endless variety of action-controllable 2D worlds given image prompts.

[Tivadar Danka @TivadarDanka]

I described some of the most beautiful and famous mathematical theorems to Midjourney.

Here is how it imagined them:

1. "The set of real numbers is uncountably infinite."

[Bernhard Schölkopf @bschoelkopf]

Another AI paradox: people are excited about LLMs, some even think that AGI is just around the corner. But some students are depressed how they can still get a PhD. Is it becoming pointless?
Some personal notes on this. (1/8)

[Ian Osband @IanOsband]

Excited to share some of our recent work!

Fine-Tuning Language Models via Epistemic Neural Networks
https://t.co/6hC9CQDhme

TL;DR: prioritise getting labels for your most *uncertain* inputs, match performance in 2x less data & better final performance

Discussion (1/n)

[Allison Tam @allisontam_]

New paper! Language and large foundation models come together to drive semantically meaningful exploration. This idea helps RL agents learn faster in 3D environments, even when language annotations are unavailable (https://t.co/ez0PGuwFXC)

Read on 🔎⬇️

[Google DeepMind @GoogleDeepMind]

Today in @nature, with @EPFL, the first deep reinforcement learning system that can keep nuclear fusion plasma stable inside its tokamaks, opening new avenues to advance nuclear fusion research.

Paper: https://t.co/UdIIfNgaap

[Tim Behrens @behrenstimb]

Oh dear. Look at this ridiculous attempt at summarising all the recent hippocampal state space models. Laughable.
(👀@jcrwhittington, @d_mccaffary,@BakermansJJW)

[2202.01682] How to build a cognitive map: insights from models of the hippocampal formation
https://t.co/sJU7RUWvIh

[Jason Scott @textfiles]

I think the way that David Crane fit 255 room configurations into an Atari 2600 cartridge is a story that should be told every Christmas

https://t.co/hyPl9u4HfV

[James Whittington @jcrwhittington]

Hello! Very happy to share that The Tolman-Eichenbaum Machine (TEM) has been published. Many thanks to @behrenstimb and my other co-authors @TimMuller1,@ShirMarkLey,@GuifenChen,@caswellcaswell, @NeilBurgess10 for their help and support along the way! 1/15 https://t.co/PlOoKqyGNO

[Google DeepMind @GoogleDeepMind]

Out-of-sample generalisation of memory in deep reinforcement learning agents: we demonstrate how two different kinds of memory help out-of-distribution generalisation, and propose a memory task suite. https://t.co/azXUCByeKk

[Berkeley AI Research @berkeley_ai]

#TransferLearning is crucial for general #AI, and understanding what transfers to what is crucial for #TransferLearning. Taskonomy (#CVPR18 oral) is one step towards understanding transferability among #perception tasks. Live demo and more: https://t.co/jGJSA4oIzM

[Richard Socher @RichardSocher]

Very excited to announce the natural language decathlon benchmark and the first single joint deep learning model to do well on ten different nlp tasks including question answering, translation, summarization, sentiment analysis, ++
https://t.co/R5wbnAQcC3