Topics

ETH Zurich researchers have developed muscle actuators that outperforms earlier systems by designing living muscle tissue and artificial scaffolds together as one system. The centimetre-scale biohybrid robots can jump, swim, walk and grasp using lab-grown mouse muscle cells combined with 3D-printed hydrogel structures.

Researchers have shown that autonomous AI systems can create their own communication systems without human guidance. The breakthrough could enable self-driving cars, robots and industrial sensors to share information that human languages cannot express.

Researchers at Seoul National University have developed Dynin-Robotics, a single AI model that combines language understanding, visual prediction, and robot control. In tests on a real robot, it achieved a 78.4% success rate, outperforming models from NVIDIA and Physical Intelligence.

University of Bristol researchers have demonstrated that lab-grown brain organoids can process tactile information from a robotic sensor to recognize Braille letters with 83% accuracy when three organoids work together, marking progress toward energy-efficient biological computing systems.

Researchers at KAIST and Microsoft Research Asia developed Neural Value Alignment, a system that reads EEG signals in real time to detect when AI misunderstands human intent. By distinguishing between reward prediction error and state prediction error, the technology enables machines to self-correct without explicit verbal feedback, pointing toward more natural human-AI collaboration.

A standardized dataset maps how magnetic fillers reshape the stiffness, stretchability and magnetic response of materials designed for soft robotics.

EPFL engineers built a sliding fiber motor that turns voltage into linear motion without gears, for soft robots and wearable devices.

The New Jersey facility combines direct movement recording with external robot evaluation, targeting the data requirements of physical AI.

EdgeCortix has introduced RAIDEN, a multi-die chiplet architecture designed to scale AI compute for robots, autonomous systems, and edge servers. The platform ranges from single-die to four-die configurations, with the flagship X4 offering up to 3.36 petaflops of FP4 compute and 256 GB of memory. The chiplet approach aims to solve the scaling limitations faced by physical AI systems that must perceive, reason, and act in real-time environments.