Uploaded November 2025 | Updated September 2026, 2 weeks ago
Generalizing AI for different types of robots, robots teaching each other, and entering an era of physical agents - tune in into today's video.
Catch these talks live and ask your own questions, join the BuzzRobot community: join.slack.com/t/buzzrobot/shared_invite/zt-37g5q0ao5-eMK_iDf0n4LAsh1d2qJYnQ
BuzzRobot interviewed Dhruv Shah and Ted Xiao, research scientists at Google DeepMind working in the Gemini Robotics team, about Gemini Robotics 1.5, Vision Language Action models (VLAs), and the progress in training AI to interact with the physical world through robots.
Gemini Robotics 1.5: Pushing the Frontier of
Generalist Robots with Advanced Embodied
Reasoning, Thinking, and Motion Transfer Report: arxiv.org/abs/2510.03342
Gemini Robotics 1.5 Report (PDF): storage.googleapis.com/deepmind-media/gemini-robotics/Gemini-Robotics-1-5-Tech-Report.pdf
Join BuzzRobot:
Newsletter: buzzrobot.substack.com
X: https://x.com/sopharicks
Slack: join.slack.com/t/buzzrobot/shared_invite/zt-37g5q0ao5-eMK_iDf0n4LAsh1d2qJYnQ
Support us: ko-fi.com/sophiaaryan
Timestamps:
0:00 Intro
1:02 VLAs and reasoning innovations
3:06 Knowledge transfer and scaling
05:54 Thinking before the robot acts
07:17 Robots learning from each other
8:07 Robotics landscape context
10:54 Robot knowledge transfer
12:52 Obstacles in generalization
15:11 Embodied reasoning model
22:18 Causality in robotics
25:24 Continual learning
32:06 What is imitation learning
34:49 Reward hacking
37:45 Data collection constraints
41:20 Tactile and mechanical feedback
42:37 Lighting training
43:59 The state of robotics
Generalizing AI for different types of robots, robots teaching each other, and entering an era of physical agents - tune in into today's video.
Catch these talks live and ask your own questions, join the BuzzRobot community: join.slack.com/t/buzzrobot/shared_invite/zt-37g5q0ao5-eMK_iDf0n4LAsh1d2qJYnQ
BuzzRobot interviewed Dhruv Shah and Ted Xiao, research scientists at Google DeepMind working in the Gemini Robotics team, about Gemini Robotics 1.5, Vision Language Action models (VLAs), and the progress in training AI to interact with the physical world through robots.
Gemini Robotics 1.5: Pushing the Frontier of
Generalist Robots with Advanced Embodied
Reasoning, Thinking, and Motion Transfer Report: arxiv.org/abs/2510.03342
Gemini Robotics 1.5 Report (PDF): storage.googleapis.com/deepmind-media/gemini-robotics/Gemini-Robotics-1-5-Tech-Report.pdf
Join BuzzRobot:
Newsletter: buzzrobot.substack.com
X: https://x.com/sopharicks
Slack: join.slack.com/t/buzzrobot/shared_invite/zt-37g5q0ao5-eMK_iDf0n4LAsh1d2qJYnQ
Support us: ko-fi.com/sophiaaryan
Timestamps:
0:00 Intro
1:02 VLAs and reasoning innovations
3:06 Knowledge transfer and scaling
05:54 Thinking before the robot acts
07:17 Robots learning from each other
8:07 Robotics landscape context
10:54 Robot knowledge transfer
12:52 Obstacles in generalization
15:11 Embodied reasoning model
22:18 Causality in robotics
25:24 Continual learning
32:06 What is imitation learning
34:49 Reward hacking
37:45 Data collection constraints
41:20 Tactile and mechanical feedback
42:37 Lighting training
43:59 The state of robotics
