AI Radar

Noteworthy AI conversations and writing, summarized from publisher notes.

Keerthana Gopalakrishnan on Gemini Robotics 2, Cross-Embodiment, and Humanoids

· The Cognitive Revolution

Keerthana Gopalakrishnan, Research Lead for Gemini Robotics at Google DeepMind, discusses the release of Gemini Robotics 2 and the pursuit of generalist humanoid robots. She details how DeepMind pairs reasoning models like Gemini Robotics ER 2 with vision-language-action execution to enable whole-body intelligence. The discussion highlights why multi-fingered manipulation and cross-embodiment continue to present harder bottlenecks than bipedal locomotion. Gopalakrishnan also addresses the practical realities of deploying physical AI in human spaces, including trade-offs across inference latency, sensor failures, and operational safety. Additional topics include simulation, generalization versus task mastery, robotic hand design, robotics training data, and managing robot fleets.

AI Radar wrote this summary from the publisher’s notes with AI, not from a transcript. How it works.