SenseTime Launches Wuneng Embodied AI Platform for Self-Evolving AI in the Physical World

Moving AI from the digital world into the physical world

As natural-language data on the internet approaches saturation, AI needs to move beyond the digital world and enter the physical world to achieve its next stage of growth.

SenseTime launched Wuneng, an embodied AI platform centered on the Kaiwu world model. The platform is designed to help robots and intelligent devices autonomously explore, learn, and evolve in physical environments.

World models as the bridge to embodied intelligence

The Wuneng platform uses world-model capabilities to connect perception, prediction, decision-making, and action. It aims to give embodied agents a richer understanding of physical causality and spatial interaction, enabling them to adapt to complex real-world tasks.

By combining native multimodal AI, world-model simulation, and embodied control, SenseTime is building a path for AI systems to understand not only digital content, but also the physical environments where robots and devices operate.

Toward self-evolving physical AI

Wuneng represents SenseTime's exploration of a new stage of AI: systems that can learn through interaction, improve through simulation and feedback, and eventually support a wider range of real-world automation scenarios.