Check out our new work on learning to drive in new cities without collecting any new demonstrations in that city. All we need is a simple reward function and readily available map and meta-information about the city. Self-play reinforcement learning does the rest!

I am hiring for two roles on my team at Waymo, based in London or Oxford. Come work at the cutting edge of autonomous driving! careers.withwaymo.com/jobs/machine... and careers.withwaymo.com/jobs/senior-...

I was interviewed for the Robot Talk Podcast.

Check out our new work on using low-rank perturbations to make evolution strategies work for billion-parameter models.

The robotics revolution is here.

Waymo is coming to London next year.

Got questions about self-driving cars? 🤖 🚘 Later this season, I'll be chatting to Prof. Shimon Whiteson (@shimon8282.bsky.social) from @ox.ac.uk and @waymo.bsky.social about machine learning for autonomous vehicles. Send me your questions for Shimon in the comments below! #Robots #Robotics #AI

I'm hiring a Machine Learning Engineer for my team at Waymo Research. Come help us tackle the biggest challenges in autonomous vehicles! careers.withwaymo.com/jobs/machine...

"In the fight against injustice, winning is a moral duty." www.nytimes.com/2025/05/17/m...

Our survey on meta reinforcement learning has now been published by Foundations and Trends in Machine Learning: nowpublishers.com/article/Deta...

Just another day at the office.

In order to set a good example for my students, I feel morally obliged to do as little work as possible.

A new version of the paper Counterfactual Multi-Agent Policy Gradients, that I first published in 2017 with @jfoerst.bsky.social, Greg Farquhar, and others, is now available on arXiv (arxiv.org/abs/1705.08926).

My team at Waymo is hiring two London-based interns for Summer 2025. Please consider applying! careers.withwaymo.com/jobs/2025-su... careers.withwaymo.com/jobs/2025-su...

What if instead of viewing AI as a separate entity, we treat the human and computer as one cognitive unit that together outperforms either on its own? My prefrontal cortex doesn't fret about delegating visual processing to my visual cortex. I don't fret about delegating navigation to Google Maps.

Who says Mountain Car is not a real world problem?