Video Friday: AI Gives Robot Hands Human-Like Dexterity
Summary
This edition of "Video Friday" from IEEE Spectrum Robotics presents a curated selection of recent robotics developments and upcoming events. Key highlights include the introduction of GENE-26.5, an AI brain from Genesis AI, which reportedly enables human-level physical manipulation in robots for complex tasks like cooking and playing piano. MIT Media Lab showcases Labububot, a spherical robot critiquing social robotics. Boston Dynamics' Spot is demonstrated performing advanced maneuvers via reinforcement learning, while NASA JPL pushes Mars helicopter flight limits by breaking the sound barrier with next-generation rotor blades. Other innovations feature 3D-printable multi-layer wave-winding topologies for high-efficiency electric motors and Sally, a magnetic climbing robot from Robomechanics Lab, designed for automated X-ray fluorescence inspection of steel structures, capable of traversing challenging interior corners.
Key takeaway
For research scientists developing advanced robotic systems, these demonstrations indicate a rapid acceleration in capabilities, particularly in manipulation and autonomous navigation. You should investigate the underlying AI and material science innovations, such as GENE-26.5's architecture or novel 3D printing techniques, to integrate similar advancements into your own projects, potentially reducing development cycles for complex tasks.
Key insights
Advanced AI and robotics are enabling human-level manipulation, extreme flight, and complex autonomous inspection.
Principles
- AI can achieve human-level manipulation.
- Reinforcement learning enhances robot agility.
Method
Neural networks trained with reinforcement learning and multi-expert distillation control advanced robot locomotion. Novel 3D-printable wave-winding topologies enable high-efficiency electric motors.
In practice
- Explore GENE-26.5 for complex manipulation tasks.
- Consider magnetic climbing robots for structural inspection.
Topics
- Human-Level Manipulation
- Reinforcement Learning
- Mars Helicopter Technology
- Additive Manufacturing
- Magnetic Climbing Robots
Best for: Research Scientist, Robotics Engineer, AI Scientist, Director of AI/ML
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by IEEE Spectrum.