Dr. Simon Müller | AI Track | Compass AI & Data Summit 2025
summary
Overview of how AI crosses from screens into stuff: recommends building an MCP server for a real device (Node-RED or Python), shows CAD generation from specialized LLMs for 3D printing, and cites visual-language-action models plus heavy use of simulation to turn short teleoperation recordings into millions of training runs. Notes current humanoids often rely on teleoperation ($20k 1X robot advertised with remote operators), early hobby bots like L robot cost under $200 and work for hands‑on experiments.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:17AI interfaces evolution
- 5:53Digital and physical AI
- 9:05MCP servers as interfaces
- 15:22Humanoid robots limits
- 21:48Humanoid robotics progress
- 27:14AI agents hype cycles
- 31:30Future outlook for physical AI
- 32:32Q&A
- 35:22Specialized vs general robots
- 38:09Humanoid robot safety
- 41:43AGI implications
- 44:58Digital twins role
related talks
18mRobotics: why now?
15mPhysical AI in a World of Humans
16mReal-World Robotics as the Next Frontier for AI?
26mPanel: AI Future - Where Investments Are Being Made
14mTowards Unobtrusive Physical AI: Augmenting Everyday Objects with Intelligence and Robotic Moveme...
10mMeet the Speakers
33mThe New Generation of Robots
27mThe Cutting Edge of AI: From LLMs to Multimodal Intelligence