DeepMind's Keerthana on Gemini Robotics 2: manipulation and cross-embodiment are the real bottlenecks
The Cognitive Revolution · youtube · 2026-10-03
Keerthana Gopalakrishnan, Research Lead for Gemini Robotics at Google DeepMind, joins the Cognitive Revolution podcast to discuss the Gemini Robotics 2 release and the push toward generalist humanoids.
Key points:
- Architecture: pairs reasoning models like Gemini Robotics ER 2 with vision-language-action execution, aiming for "one brain, any body" cross-embodiment generalization
- Bottlenecks: multi-fingered manipulation and cross-embodiment remain far harder than bipedal locomotion; generalization vs. task mastery is a central trade-off
- Deployment: inference latency, sensor failures, and operational safety must be handled jointly since physical-AI errors carry immediate physical consequences
Chapters also cover simulation, whole-body humanoid control, robotic hands, safety, and the future of robotics data and fleet interaction.
More from Embodied
- ETH's Anymal robot completes 10km autonomous alpine hike with VLM-guided ROAM system — ChongZzZhang · 2026-11-19
- iFixit tears down Apple HomeView: lots of glue and 'endless frustration' — Resident-Fall4351 · 2026-11-19
- Removing human video drops robot OOD success from 60% to 7%: generalization comes from human video — gui_penedo · 2026-11-19
- PSYONIC's prosthetic Ability Hand now used by 90+ robotics orgs, raising Series A — DhowCIO · 2026-11-19
- Simulator misspecification breeds overconfidence: sim-to-real study on SPEED+ — Nikolay Dolgov · 2026-11-19
- Doppel world model renders the robot body and beats action-only baselines across 8 embodiments — mangahomanga · 2026-11-19