TheBitCast

Embodied AI 101

Autor: Shaoqing Tan

Embodied AI 101 tracks the latest research at the intersection of AI and physical intelligence. Robotics, manipulation, world models, and the path from digital intelligence to embodied agents — each episode breaks down recent papers and advances in the field. We examine the methodology, real-world implications, and path to deployment. Covering robot learning, perception, control, and the fundamental challenge of grounding intelligence in the physical world. New episodes daily.

Odcinki

  1. Axis Robotics Browser Simulation: Does Sim-to-Real Manipulation Work? – Embodied AI 101
  2. HexaShell: Can Debugging Code Generalize RGB Robot Manipulation? – Embodied AI 101
  3. Niantic Spatial Gaussian Splats: Better Sim-to-Real Robot Training? – Embodied AI 101
  4. Gemini 4 Argon for Robotics: What the Benchmarks Actually Show – Embodied AI 101
  5. GreenWAM: Can One World-Action Model Learn Across Robot Bodies? – Embodied AI 101
  6. 3DCodeVerse: Can Code Generation Build Better Worlds for Robotics? – Embodied AI 101
  7. Motion- and Gravity-Aware 3D Reconstruction: Do Offline Priors Help? – Embodied AI 101
  8. ASPIRE: Can Code Refinement Make Robot Manipulation Generalize? – Embodied AI 101
  9. UAV-ESPrune: Can Visual Token Pruning Preserve UAV Reasoning? – Embodied AI 101
  10. Gemini Robotics 2: How General Is Its Whole-Body Robot Control? – Embodied AI 101
  11. RoboJEPA to Long-WAM: Do New Embodied AI Models Generalize? – Embodied AI 101
  12. VLA-ACL: Can Token Pruning Speed Up Robot Manipulation? – Embodied AI 101
  13. Juno: Do Predictive Latents Make VLA Manipulation More Robust? – Embodied AI 101
  14. Can Rephrasing Fix VLA Language Sensitivity Without Retraining? – Embodied AI 101
  15. Can CriticGPT Replace Hand-Designed Rewards for Robot Manipulation? – Embodied AI 101
  16. Urbex: Can Agentic VLM Search Improve City-Scale 3D Grounding? – Embodied AI 101
  17. VPP2: Does Better Video Prediction Improve Robot Manipulation? – Embodied AI 101
  18. PointWAM: Does 3D World Modeling Improve Dexterous Manipulation? – Embodied AI 101
  19. Mulligan: Can Failure-Focused Data Improve Robot Imitation Learning? – Embodied AI 101
  20. SOLE-R1: Can Video-Language Rewards Train Manipulation Policies? – Embodied AI 101
  21. Odyssey-3: Does a Physics-IQ World Model Transfer to Real Robots? – Embodied AI 101
  22. Long-WAM: Does Longer Visual Context Improve Real-Time Robot Control? – Embodied AI 101
  23. WR1.0: Does Whole-Body Humanoid Control Prove Generalization? – Embodied AI 101
  24. WorldGuide: Better Video Planning, Not Proven Robot Control – Embodied AI 101
  25. R2SGen: Can Constraint Solving Improve Real-to-Sim Driving Scenarios? – Embodied AI 101
  26. Humanoid Scooter Control: Does Sim-to-Real Reinforcement Learning Work? – Embodied AI 101
  27. ENPIRE: Can Coding Agents Improve Real-World Robot Manipulation? – Embodied AI 101
  28. GAS-Robo: Can Grid Actions Improve LLM-Driven Manipulation? – Embodied AI 101
  29. Four Keypoints, One Careful Pour: Connecting Language to Precise Robot Motion – Embodied AI 101
  30. PEAfowl: Helping Two Arms Agree on What They See – Embodied AI 101