Microsoft launches new AI model for real-world robotic learning

Microsoft has unveiled Rho-alpha, a new AI model designed to enable robots to operate effectively in unpredictable, real-world environments beyond traditional factory settings. Developed by Microsoft Research as part of its Phi vision-language AI family, Rho-alpha represents a shift toward "physical AI," where intelligent agents interact directly with the physical world. Unlike conventional industrial robots that follow rigid scripts, Rho-alpha translates natural language instructions into control signals for complex, two-handed robotic manipulation tasks. The system incorporates tactile sensing alongside visual input, allowing robots to adapt their movements based on touch and learn from human corrective feedback in real time. Microsoft is currently testing Rho-alpha on dual-arm and humanoid robots and plans to expand its sensory capabilities in future versions. A key challenge addressed by Rho-alpha is the scarcity of training data for robotics. Microsoft combines physical robot demonstrations, simulated tasks, and large-scale visual question-answering datasets to train the model. The company leverages reinforcement learning pipelines and robotics simulation tools on Azure to generate synthetic