Introducing Gemini Robotics and Gemini Robotics-ER, AI models designed for robots to understand, act and react to the physical world. Introducing Gemini Robotics and Gemini Robotics-ER, AI models designed for robots to understand, act and react to the physical world.

More brains for robots!

Really looking forward to how this progresses:
deepmind.google/discover/blo...

If tools like AI co-scientist help improve the recall of scientific hypothesis generation, the biggest lever we have now to speed up much of research is improving the precision of hypothesis selection. And this means that *negative results* are more valuable than ever.

AI showdown of the day: "which company built a cuda compiler that did better than cudnn written in perl and was later acquired by intel". ChatGPT and Claude got it immediately right. Google and Gemini had no clue what I was talking about. Perplexity and DeepSeek confidently gave me bogus answers.

TFW you work all day on AI for autonomous cars, go home to relax with a good book, the book is called "Death of the Author", and said author steps into a very familiar-looking autonomous vehicle. @nnedi.bsky.social I am definitely not relaxed now.

In French, Silence of the Lambs is 'Le Silence des Agneaux'. Lord of the Rings is 'Le Seigneur des Anneaux'. My childhood was incredibly confusing until I realized these were NOT THE SAME THING!!!

Gen media is following the same trend as NeRF / 3DGS: a lot of focus on pretty renderings, when so much of the potential for these methods is the seamless combination of apprearance/geometry/semantics for functional applications.

I really wish to see more research emphasis on functional uses of generative media. Creative movies are cute and impressive, but generating plausible, physically realistic consequences of real-world actions is game changing. I don't think any 'world model' we have today actually understands physics.

Every year the French police issues a literary award for best crime novel. Summary of last year's winner: cop beats suspect. Cop steals evidence for no reason whatsoever. Politicians bad. Cop wins. I don't know what I was expecting.

I have been sitting on an idea for a sci-fi story for a decade, simply because I couldn't work out if the physics behind it even made sense, or whom to ask for help with the math. I just asked Gemini 2.0 Flash, and got my answer. I may actually have to write that thing.

Thrilled to announce our workshop on Embodied Intelligence for Autonomous Systems on the Horizon @cvprconference.bsky.social featuring a crazy line-up of speakers and challenges.
Mark it in your agendas and also in your registration #cvpr2025
opendrivelab.com/cvpr2025/wor...

MuJoCo Playground An open-source framework for GPU-accelerated robot learning and sim-to-real transfer

Introducing playground.mujoco.org
Combining MuJoCo’s rich and thriving ecosystem, massively parallel GPU-accelerated simulation, and real-world results across a diverse range of robot platforms: quadrupeds, humanoids, dexterous hands, and arms.
Get started today: pip install playground

Gemini Deep Research is wonderful for exploration, but the killer use-case is obviously marketing/SEO slop

I recently read a fantasy book where people have mastered the use of language to cajole things into acting in ways that are unnatural, like convincing a rock to fall up. Immediate thought: 'oh neat, everything has an inner LLM and magic is prompt engineering. I am a wizard now'.

I don't think the quote is directly from him. It's from a journalist interviewing him from the New Yorker (Larissa MacFarquhar, 2018) Good summary of some of his views though.

Loga(Algo)rithmic Progress My grind is your plateau

My love for AI is plateau-nic
vanhoucke.medium.com/loga-algo-ri...

From my ride to your family.