AI Enabled Robotics - Stuart Bowers - Google Deepmind - Scaled ML 2026
Stuart Bowers (Google Deepmind) provides an introduction to AI-enabled robotics, specifically focusing on the development and applications of the Pupper robot and the transition from traditional heuristic-based control to modern AI-driven approaches like reinforcement learning (RL) at Scaled ML 2026 *Timestamps* **00:00 - 01:14**: Introduction and the goal of inspiring AI experts to explore the field of robotics. **01:15 - 02:40**: Overview of the Stanford CS123 curriculum, emphasizing its open-source nature and hands-on approach to AI robots. **02:41 - 04:30**: Introduction to the Pupper robot, its hardware design, and its accessibility as a low-cost, open-source platform. **04:31 - 06:01**: The "old way" of robotics: Heuristic control based on natural animal gaits like trotting. **06:02 - 08:20**: Technical breakdown of inverse kinematics and coordinate-based foot movement. **08:21 - 10:51**: Real-world impact: Using Pupper at Stanford Children's Health to reduce patient anxiety. **10:52 - 13:31**: The "new way" of robotics: Training walking policies using reinforcement learning in simulators. **13:32 - 15:06**: Key RL techniques for closing the "sim-to-real gap": System identification and domain randomization. **15:07 - 17:04**: Advanced RL: Reward shaping to achieve stable, natural-looking gaits suitable for clinical environments. **17:05 - 18:55**: Advancements in perception: Utilizing onboard neural accelerators and YOLO for high-speed object detection and chasing. **18:56 - 20:40**: Action planning: Moving from remote control to LLM-powered function calling for natural language interactions. **20:41 - 22:02**: Demonstration of sequenced actions, showing Pupper performing yoga moves through voice commands. **22:03 - 23:10**: Conclusion, acknowledgments, and resources for getting started in robotics. **23:11 - 26:22**: Q\&A session regarding natural dog movements in RL and the future of soft robotics.

Fully autonomous robots are much closer than you think – Sergey Levine

The World's Most Important Machine

The Hardest Problem AI Ever Solved, with Google DeepMind CEO

Robotics' End Game: Nvidia's Jim Fan

Neha Sharma and Nicolas Lidzborski

China Is About To Pop The AI Bubble

AlphaFold - The Most Useful Thing AI Has Ever Done

How to Build & Sell AI Agents: Ultimate Beginner’s Guide

The Uncomfortable Truth About AI “Reasoning” | World Science Festival

Physical AI for the Real World: A Vision From NVIDIA Robotics Research

Creator of C++: Bell Labs, Negative Overhead Abstraction, Mistakes | Bjarne Stroustrup

Training AI to Play Pokemon with Reinforcement Learning

Why AI Tokens are so Expensive - Computerphile

Humanoide Roboter 2026: Wer wirklich vorne liegt – und wann sie bei uns zuhause stehen

Ashok Elluswamy: Building Foundational Models for Robotics at Tesla

This is not the AI we were promised | The Royal Society

Meet the Former CIA Agent Who Wants to Abolish the CIA

Something is jamming GPS over Europe. Here's what we found

Yann LeCun: World Models: Enabling the next AI revolution

