The race to develop AI capable of understanding and interacting with the physical world is intensifying, marked by a substantial Pre-A funding round for Light Origins. Led by former OpenAI algorithm expert Jiang Xu (also known as Roger Jiang), the startup aims to extend the capabilities of large language models (LLMs) from purely digital realms into embodied intelligence. This involves creating AI that can perceive, reason, and act within real-world environments.

Light Origins secured hundreds of millions of yuan in this funding round, with CASIC Investment leading the investment. China Merchants Venture Capital and Xianghe Capital also participated, signaling strong investor confidence in the company's vision. The capital will be used to fuel large-scale embodied AI model training, build multimodal data infrastructure, and support full-stack software and hardware research and development.

The company's core strategy involves a three-stage paradigm: scaled pre-training, scaled alignment, and scaled deployment. This approach is designed to establish foundational capabilities through vast datasets, refine models using techniques like reinforcement learning, and then introduce these models to real-world applications. This iterative process generates high-quality physical data, creating a feedback loop that continuously optimizes the AI models.

Light Origins' focus on embodied intelligence positions it at the forefront of a critical shift in AI development. While many companies focus on improving digital AI, the ability for AI to perform complex tasks in physical spaces, from robotics to autonomous systems, represents the next frontier. This funding round underscores the market's belief that integrating AI with physical action will unlock new levels of automation and capability across various industries.