35 Large State-Owned Enterprises Tested: Causality World Model Deployed

Chinese company Zhongshu Ruizhi, founded in 2020, has developed a causal world model designed to address the inability of current AI systems to participate in critical business decision-making. At the WAIC conference, the company presented its "AI for Reasoning" system, which includes the concept of "meta-causality" and a dynamic engine for constructing and evolving causal relationships. Unlike the traditional approach based on data correlation, the model uses business causality and mechanical rules, enabling not only the identification of risks but also the prediction of their development and the consequences of interventions. The system has already been deployed in over 35 large state-owned enterprises across industries such as oil and gas, energy, manufacturing, and finance, covering more than 800 scenarios. For example, in oil drilling, the model warns about blowouts and gas breakthroughs 15 to 20 minutes before an accident, with a 94% accuracy in identifying root causes of complex situations and a 40% reduction in false alarms. The company has also published a "White Paper on Causal World Model Technology," detailing its methodology: five pillars corresponding to five shortcomings of current language models — factual hallucinations, emphasis on description without prediction, lack of temporal dimension, inability to pre-validate, and accumulation of rules. Zhongshu Ruizhi's approach differs from the mainstream focus on physical models: the company emphasizes business causality rather than physical laws and has already achieved industrial deployment in scenarios with zero tolerance for errors.