China promotes transition toward a practice-based AI evaluation system

Experts and policymakers are calling for a shift in how artificial intelligence is evaluated, moving away from standardized benchmarks toward a system based on practical application and real-world utility. Following recent State Council guidelines on the “AI Plus” initiative, there is a growing emphasis on establishing evaluation frameworks that measure an AI system's ability to solve specific problems, manage costs, and control risks within actual production environments.

Economist Zhu Keli notes that China is transitioning from a digital economy to an “intelligent economy,” where AI serves as a core driver of autonomous decision-making rather than just a data tool. This shift requires reshaping production factors, organizational structures, and governance models. In regions like Ningbo, this transition is manifesting through “specialized, refined, unique, and innovative” (SRUI) enterprises that integrate AI into traditional manufacturing.

To support this evolution, suggestions include moving from general model testing to evaluating entire integrated systems. Furthermore, there are calls for more precise government support, such as providing “Token” vouchers for computing power and data instead of traditional cash subsidies, to better align with the needs of AI-native businesses.