Jane Wang

😂

There's also nothing in the way we train AI that would ever allow it to optimize for "survival above all else", so unless we start training them in a completely new way, I'm not worried

Thanks for explaining, but I'm not convinced that power dynamics favor intelligence. There's already a range of intelligence among people, and people already self-improve (learn), but I'd say there's only a loose correlation between intelligence and power among humans

I put social media into the same category as smoking and lead paint. Some time decades from now, we'll be shocked we allowed ourselves and our kids to be so freely exposed to such a toxic substance

A "wasteland of fake nonsense" describes my X feed to a T 😭 I can't even open it anymore it's horrible

My main point is about comparison with social media. How emotionally attached would you say people are today to tiktok, Instagram, Facebook, etc? A small percentage of people can get attached to anything, but only social media was expressly designed to get you addicted to mindlessly scrolling

Social media is optimised for engagement. Chatbots are optimised to be useful. That makes all the difference

We are far from understanding just how much social media has damaged us. While people are getting whole PhDs on AI safety

Mark Zuckerberg was fine with allowing his platform to be used to push misinformation and affect elections around the world

Yes bad actors can start deploying AI to unleash deepfakes, or put a bunch of spam on the internet, but bots already existed, and so did Photoshop. The very fabric of public discourse has already been shredded by social media companies, for the sake of clicks and likes.

It'll come with the risks that any major new technological innovation will involve, but nothing like the "extinction-level" threat I keep hearing about. I have no idea where these people are coming from. If someone wants to explain, I'm happy to listen!

AGI will be created with the world's attention on it, and very thoughtful people that I work with every day are thinking about ways to make it beneficial for everyone

One major reason I believe this: when the likes of Facebook and Twitter were being set up, NO ONE was concerned about safety, ethics, or regulation of these companies, until it was far too late, and the damage was done (and is still being done)

This might be a hot take, but I think social media has done much more damage to humanity than AGI has the potential to do. At this point, I think AGI will likely be a net positive by ameliorating some of the catastrophic effects of social media.

Genie 3: A New Frontier for World Models Today we are announcing Genie 3, a general purpose world model that can generate an unprecedented diversity of interactive environments. Given a text prompt, Genie 3 can generate dynamic worlds...

Genie 3: A general world model

Google announced Genie 3, a world model that can generate 3D scenes in real-time, meaning that it can be used to create 3D experiences that you can immediately use. Like a dynamic video game limited by your own imagination

deepmind.google/discover/blo...

eurips.cc A NeurIPS-endorsed conference in Europe held in Copenhagen, Denmark

NeurIPS is endorsing EurIPS, an independently-organized meeting which will offer researchers an opportunity to additionally present NeurIPS work in Europe concurrently with NeurIPS.

Read more in our blog post and on the EurIPS website:
blog.neurips.cc/2025/07/16/n...
eurips.cc

Key-value memory in the brain Classical models of memory in psychology and neuroscience rely on similarity-based retrieval of stored patterns, where similarity is a function of retrieval cues and the stored patterns. While parsimo...

Key-value memory is an important concept in modern machine learning (e.g., transformers). Ila Fiete, Kazuki Irie, and I have written a paper showing how key-value memory provides a way of thinking about memory organization in the brain:
arxiv.org/abs/2501.02950

bye, felix

kyunghyuncho.me/bye-felix/

Felix Hill and some other DMers and I after cold water swimming at Parliament Hill Lido a few years ago

Felix Hill was such an incredible mentor — and occasional cold water swimming partner — to me. He's a huge part of why I joined DeepMind and how I've come to approach research. Even a month later, it's still hard to believe he's gone.

A brilliant colleague and wonderful soul Felix Hill recently passed away. This was a shock and in an effort to sort some things out, I wrote them down. Maybe this will help someone else, but at the very least it helped me. Rest in peace, Felix, you will be missed. www.janexwang.com/blog/2025/1/...

💔

Merry Christmas Bluesky! 💙

Looking forward to speaking at the CALM workshop today at 930! I'll be talking about causal reasoning in foundation agents #neurips2024
calm-workshop-2024.github.io

(From Feb. 10, 2021):

Introducing a benchmark task for meta-reinforcement learning: Alchemy! Train agents or play it yourself, this environment comes with interesting latent structure that encourages experimentation and structure learning.

github.com/google-deepm...
neurips.cc/virtual/2021...

Introducing 🧞Genie 2 🧞 - our most capable large-scale foundation world model, which can generate a diverse array of consistent worlds, playable for up to a minute. We believe Genie 2 could unlock the next wave of capabilities for embodied agents 🧠.

When will Bluesky get a "trending" feature? I always enjoyed looking at what an entire country was thinking about at any given moment 😄

I'm so tired of living in Trump's fevered dream

Applications are open for student researcher positions at Google DeepMind for 2025! Due date Dec 13

www.google.com/about/career...

Hi Ida! I haven't heard of blue ark, what's that?

The speed of growth on this platform is insane! Really feels like it's hit a critical point. At what point can we just consider it a clone of Twitter?