Launch Dexterity · ASRock Rack Dexterity and ASRock Rack are embedding custom edge AI supercomputers into Dexterity's Mech warehouse robot to enable real-time manipulation without cloud dependency. Dexterity and ASRock Rack Partner to Deliver Physical AI Supercomputers for Superhumanoid Edge Intelligence · Dexterity
Research Physical Intelligence — Marcel Torne, Karl Pertsch, Homer Walke Physical Intelligence introduces a memory architecture for vision-language models that enables robots to execute multi-stage tasks like kitchen cleaning by maintaining coherent understanding of object locations and past actions across extended durations. VLAs with Long and Short-Term Memory · Physical Intelligence
Research Physical Intelligence — Benjie Holson Physical Intelligence demonstrated its π 0.6 model solving three of five Robot Olympics manipulation tasks, including peanut butter spreading and lock-turning, showing progress on the hard-easy inversion that defines Moravec's paradox. Moravec's Paradox and the Robot Olympics · Physical Intelligence
Research Physical Intelligence — Sergey Levine, Chelsea Finn Physical Intelligence releases π * 0.6, a visual language model for robotic manipulation that learns from human demonstration, correction, and self-practice to achieve high-reliability task execution. π * 0.6 : a VLA that Learns from Experience · Physical Intelligence
Research Physical Intelligence — Kevin Black, Manuel Y. Galliker, Sergey Levine Physical Intelligence published a research paper on real-time action chunking for vision-language-action models, enabling robots to operate with lower latency by executing sequences of predicted actions while the model reasons about the next ones. Real-Time Action Chunking with Large Models · Physical Intelligence