XNews

Google Robotics Interview: Integrating LLMs and RL for Robots, Data Bottlenecks, and Future Timelines

Interview with Google robotics lead Tang Jie covering two paradigm shifts (reinforcement learning and LLMs), Gemini Robotics 1.5 innovations (textual 'thinking' traces and cross-body motion transfer), data as the core bottleneck with simulation and generated video as scaling paths, importance of touch for dexterous hands, and a multi-year timeline: 2–3 years for perceivable breakthroughs, 5 years for vertical deployment, 10 years for household arrival.

Source posts