@michaelddennis.bsky.social

RS DeepMind. Works on Unsupervised Environment Design, Problem Specification, Game/Decision Theory, RL, AIS. prev CHAI_Berkeley

2,587 Followers 264 Following 12 Posts Joined Nov 2024
1 day ago

Are there any things that would have been core tasks if not for a huge amount of these now small-automatable annoyances?

1 0 1 0
3 weeks ago

Nice!! Thanks for the pointers! So many feeds to try πŸ™Œ

3 0 0 0
3 weeks ago

Anyone have advice on getting the most out of bluesky? Any feeds/clients I should use? Mostly interested in the ML/RL space

10 0 5 0
3 weeks ago
OWASP Top 10:2025 OWASP Top 10:2025

Could be worth skimming owasp.org/Top10/2025/ to see if any of the most common issues sound like a possibility. maybe also from prior years.

0 0 0 0
1 month ago

Huge congrats to the Genie Team on the launch!! πŸš€

Glad more people can playing this now!

3 0 0 0
8 months ago
Video thumbnail

*New Paper*

🚨 Goal misgeneralization occurs when AI agents learn the wrong reward function, instead of the human's intended goal.

πŸ˜‡ We show that training with a minimax regret objective provably mitigates it, promoting safer and better-aligned RL policies!

9 2 1 0
1 year ago
YouTube
Oskar StΓ₯lberg: Landscapes of Hex and Square - Summer School on AI and Games 2023 YouTube video by Summer School on AI and Games TV

a very interesting multi-criteria optimisation problem for procedural level generation in 3D games

could one be inspiration for the environment design as a general sum game or a coalition game, going beyond zero-sum stuff?

www.youtube.com/watch?v=Npfo...

cc @michaelddennis.bsky.social

2 1 1 0
1 year ago
Post image

In our recent paper, we outline 15 fully risk-agnostic, process-based, evidence-seeking policy objectives. None of them limit *what* developers can do -- they just affect *reporting* and *visibility.*
arxiv.org/abs/2502.096...

7 1 0 0
1 year ago
Preview
A theory of appropriateness with applications to generative artificial intelligence What is appropriateness? Humans navigate a multi-scale mosaic of interlocking notions of what is appropriate for different situations. We act one way with our friends, another with our family, and yet...

Very happy to announce the publication of our latest paper:

A theory of appropriateness with applications to generative artificial intelligence
arxiv.org/abs/2412.19010

And happy new year everyone!

31 7 2 1
1 year ago

Not at NeurIPS but would love to chat about it sometime! sounds right, and I think a lot of weird stuff happens when you take it to the extreme

0 0 0 0
1 year ago

Excited to reveal Genie 2, our most capable foundation world model that, given a single prompt image, can generate an endless variety of action-controllable, playable 3D worlds. Fantastic cross-team effort by the Open-Endedness Team and many other teams at Google DeepMind! 🧞

97 18 3 3
1 year ago

Thrilled to share Genie 2! Endless environments created by text or images, a key to open-ended/AI-Generating Algorithms.

Genie 1 showed it's possible. 9 months later, Genie 2 shows jaw-dropping progress.🀯 Witness the magic of scale, again. πŸ“ˆπŸš€ Thx to all team members @deep-mind.bsky.social!

43 4 1 0
1 year ago

It’s been a crazy 2 years seeing so many amazingly talented researchers bring GENerative Interactive Environments alive in Genie 1 and 2.

the future is agents in generative environments

30 3 0 0
1 year ago

The secret to doing good research is always to be a little underemployed. You waste years by not being able to waste hours. - Amos Tversky

38 3 0 1
1 year ago

Welcome! @natashajaques.bsky.social. πŸ‘‹

I think you may find one or two people who share your sentiment here... 😁

6 1 0 0
1 year ago

🀣 really appreciate the support ❀️

0 0 0 0
1 year ago

Thanks Eugene! I just have to get around to starting a blog already to post my weirdest takes πŸ˜…

3 0 1 0
1 year ago

just joined bluesky too! πŸ™‚

7 0 0 0
1 year ago

Always love a classic β€œmore environments” when they admit the current envs prove the method works

0 0 0 0
1 year ago

I have seen the silliest. strengths, weaknesses, and questions all a variant on the same sentence

4 0 1 0