CyberRobo

We developed an RL method for fine-tuning our models for precise tasks in just a few hours or even minutes. Instead of training the whole model, we add an “RL token” output to π-0.6, our latest model, which is used by a tiny actor and critic to learn quickly with RL.

BCI+ Humanoid Robot in the Treatment of Children with Autism Fujian Agriculture and Forestry University, in collaboration with several institutions, has developed a wheeled humanoid robot, Fuxiaozhi F1-D, and combined it with brain-computer interface (BCI) technology to assist in rehabilitation training for patients with autism and other neurodevelopmental disorders. When autistic children wear the BCI device, without speaking or making gestures, Fuxiaozhi F1-D can raise its hand, grasp, and deliver objects based on brain signals, engaging in immersive interactive emotional experiences with the child. This awakens the child's initiative (crucially). The humanoid robot converts the child's attention and motor imagination into recordable data, which is more continuous and effective than traditional commands. This standardized work will significantly free up the hands of human rehabilitation therapists, allowing them to focus more on emotional guidance, relationship building, and personalized treatment design, forming a human-machine collaborative relationship. It's worth noting that this can also be extended to scenarios requiring active participation in training, such as stroke rehabilitation and spinal cord injury treatment.

Had a wonderful time unveiling the Torso 3 at the Abundance Summit this week! Video coming soon.

Two new humanoid robots from Beijing-based startup TARS demonstrate the company's ambitious plans to create flexible, precision manufacturing applications and general-purpose humanoid robots. The video showcases their potential in complex tasks such as wire harness assembly and embroidery. Embroidery, in particular, highlights the dexterity and precision of their hands and their ability to perform intricate movements. Founded in February of this year, TARS has already raised $250 million in funding. The company claims to be building a closed-loop system around autonomous humanoid robots, encompassing real-world data, foundational models, humanoid robots, and multi-scenario applications (factories, services, and daily life). While humanoid robots are not yet widely used in actual production lines, warehouses, or complex service environments, all manufacturers believe they will be the champions in the emerging field of robotic intelligence.