As artificial intelligence grows from a "smart tool" to an "intelligent partner" capable of autonomously planning tasks, a profound productivity revolution is quietly happening around us. On December 13th, at the "2026 China Academy of Information and Communications Technology Deep Observation Report Meeting", experts and scholars depicted not only an AI technology roadmap, but also a new picture of the intelligent economy reconstructing productivity and production relations. Trillion scale AI has become a new engine for economic development. Artificial intelligence is upgrading from a tool to a partner, fully empowering high-quality development. ”The opening remarks by Wei Kai, Director of the Institute of Artificial Intelligence at the China Academy of Information and Communications Technology, set the tone for this observation. A set of data confirms this judgment: by 2024, the scale of China's core artificial intelligence industry will exceed 900 billion yuan, with a growth rate of 24%; It is expected to cross the 1.2 trillion yuan mark by 2025. This is not only a quantitative increase, but also a qualitative leap. At the technical level, large models are experiencing an explosive period of capability. Wei Kai pointed out that "from the test data of the Institute of Information and Communications Technology, the model has improved its language and multimodal understanding abilities by 30% and 50% respectively, and its reasoning, programming, and other abilities have achieved 'good and fast' development." Basic innovations such as linear attention mechanism and environment oriented reinforcement learning are paving the way for the practical application of large models. However, what deserves more attention is the structural changes in the application ecosystem. As the main carrier for the implementation of large models, intelligent agents have shown the embryonic form of "digital labor". General intelligent agents perform outstandingly in scenarios such as web interaction and information integration, while specialized intelligent agents accurately break through vertical fields such as law and programming. But Wei Kai also pointed out soberly: "Whether it is specialized or universal, intelligent agents are still in the early stages of development, and the reliability of task planning, the complexity of integration with business systems, and other challenges are still prominent." The new breakthrough robot is moving from the laboratory to reality. If the big model is the "brain" of AI, then embodied intelligence is to equip AI with a "body". Xu Zhiyuan, Deputy Chief Engineer of China Academy of Information and Communications Technology, said, "Currently, embodied intelligence has achieved a dual breakthrough in cognitive intelligence and physical intelligence." On the one hand, large models enable robots to possess intelligent features such as "perceivable, thought-provoking, and interactive"; On the other hand, reinforcement learning and other technologies enable humanoid robots to rapidly improve their physical abilities such as walking on complex terrains and performing fine operations. However, Xu Zhiyuan emphasized that "despite continuous technological breakthroughs, large-scale implementation is still in the early stages." Behind this judgment is actually a dispute over the three core routes. One is the debate over model roadmap: can the paradigm of large models be directly transferred to robot control? The second is the debate over data paradigms: which type is the key, real machine data, simulation data, and human video data? The third is the debate over the form route: is it a fully humanoid route or a compromise solution of "wheel arm"? Xu Zhiyuan said, "Multiple 'wheel arm hybrid robots' have emerged in China this year, emphasizing the feasibility of engineering implementation." This pragmatic approach forms an interesting contrast with the vision of' universal humanoid robots' adhered to by companies such as Tesla, reflecting the different priorities of the AI industry in China and the United States at different stages of development. The technological breakthrough in the reconstruction of economic forms through the three "double" characteristics of intelligent economy ultimately serves economic development. Sun Ke, Deputy Director of the Policy and Economic Research Institute of the China Academy of Information and Communications Technology, proposed a concept with a more macro perspective: "The intelligent economy is a new economic form driven by artificial intelligence technology, which has significantly improved total factor productivity." In his view, the intelligent economy has three major characteristics: firstly, technology and application promote each other. AI helps businesses improve efficiency, and enterprise data makes AI smarter. The second is the upgrading of both traditional industries and the emergence of new AI products. For example, the sales of new products such as AI glasses and smartwatches are growing rapidly. Thirdly, human-machine collaboration has become the norm. Sun Ke specifically pointed out that "now 'one person+AI' can form an efficient team, and this change is giving birth to a new working mode. ”Data shows that there are 498 unicorn companies in the global AI field, and China has built over 35000 high-quality datasets. Sun Ke said, "We often overestimate the short-term impact of technology and underestimate its long-term impact. The future of AI may far exceed imagination. ”A set of data reveals the depth of change: as of August 2025, there are 498 global AI unicorn companies with a total valuation of $2.7 trillion; China has built over 35000 high-quality datasets with a total volume of over 400PB. Sun Ke said, "Human predictions about the impact of new technologies on economic growth are generally overestimated in the short term and underestimated in the long term." Manufacturing upgrading AI is deeply integrated into the core production process, and as the main battlefield of the real economy, the manufacturing industry is becoming the most profound testing ground for AI landing. Liu Mo, Director of the Institute of Information Technology and Industrialization Integration at the China Academy of Information and Communications Technology, pointed out that "if digital transformation has solved the problem of improving quality, reducing costs, and increasing efficiency in the past decade, digital intelligence will need to address innovation and resilience issues in the next five years." A key signal is the penetration and changes of AI in the manufacturing value chain. The data shared by Wei Kai shows that the distribution of AI applications in the manufacturing industry is extending from a smile curve of "high at both ends and low in the middle" to the core links of production and manufacturing. This year, the proportion of production and manufacturing cases has increased from 19.9% to 25.9%. Although this change may seem small, it means that AI is breaking through the "comfort zone" of research and development, design, and marketing services, and advancing towards the production line with more complex processes and higher reliability requirements. Liu Mo proposed a more forward-looking architecture concept: "Building a new architecture of 'basic model+digital twin+intelligent agent' will greatly enhance the intelligent analysis and decision-making capabilities of the manufacturing industry." In this architecture, the basic model is responsible for knowledge management and reasoning, the digital twin ensures high reliability analysis, and the intelligent agent realizes autonomous execution - the three work together, which is expected to achieve a qualitative change from automation to autonomy in the manufacturing system. The "cold" thinking in the heat summarizes the views of experts and scholars. The reporter found that the development of AI in China has presented three new obvious characteristics: firstly, from "tools" to "partners": AI is no longer just an auxiliary tool, but has begun to become a work partner capable of independently completing tasks. Secondly, from "usable" to "easy to use": after technological breakthroughs, how to make AI work stably and reliably has become a new challenge. Finally, from "improving efficiency" to "creating value": AI is becoming a new type of "labor force" that directly creates value. However, we still need to be clear headed amidst the craze. Wei Kai reminds that AI poses risks such as generating false information and data security, and cutting-edge technologies may even produce unexpected behavior. ”Therefore, establishing an effective security management mechanism is crucial. Looking ahead to the future, China's AI industry is entering a new stage of "deepening and solidifying" after exceeding one trillion yuan in scale. As Sun Ke said, "The unknown of AI is far greater than the known. ”In this era full of opportunities, we must embrace innovation and steadily promote it, making artificial intelligence a reliable partner for promoting high-quality development. (New Society)