Tsinghua-backed team builds distributed predictive world model, secures hundreds of millions in Series A funding, deployed on tens of thousands of terminal devices | Hard Science First Release
Qianjue Technology, a startup founded by Tsinghua University alumni, has secured several hundred million yuan in Series A funding to develop its embodied AI world model. The company's approach focuses on predictive world models rather than generative ones, aiming to enable robots to understand and anticipate physical world changes by predicting low-dimensional state evolution trajectories. This method is designed to avoid the feature pollution issues found in pixel-level prediction and allows for faster, more adaptable robot behavior, with their technology already deployed on over 100,000 devices across various applications like hotel cleaning and service robots. AI
IMPACT Predictive world models could accelerate robot autonomy and adaptability, potentially reducing training data needs and improving real-world performance.