A new round of technological revolution and industrial transformation continues to evolve, with breakthroughs in artificial intelligence technology accelerating and its application scope constantly expanding. This is reshaping economic operations and altering the traditional landscape of industrial development. This year's government work report proposed the creation of new forms of the intelligent economy. Currently, the nation is at a critical stage of high-quality development. Accelerating the empowerment of various industries with AI and rapidly establishing new forms of the intelligent economy represent a crucial pathway for cultivating new drivers of development and an inherent requirement for advancing the modernization drive.
Consolidating the intelligent foundation requires building a robust support system that integrates cloud, network, edge, and device. This involves implementing a computing power layout project that coordinates training and inference. It is essential to plan the allocation of training and inference computing power in a unified manner, promoting a shift in computing power supply from "hardware stacking" to "service delivery." Efforts should be made to make computing resources perceivable, measurable, schedulable, and billable. Service level agreements need to be improved, with clear standards for availability, response latency, throughput capacity, and fault recovery. Rules for peak-valley complementarity and cross-regional coordination should be refined to guide training tasks towards off-peak hours and low-cost regions.
The capacity for inclusive supply from public clouds must be enhanced. Accelerating the coordinated development of cloud, network, edge, and device is key to building a universally accessible computing power supply system for all industries. Strengthening the integration of cloud and network, along with optimizing the layout of edge nodes, will improve capabilities for dedicated access, low-latency transmission, and localized inference. Policy tools such as computing vouchers should be utilized effectively to lower the barriers for enterprises to adopt cloud services and computing power.
Promoting the synergistic development of computing power and the power grid is crucial. Planning for computing infrastructure must be integrated with power supply guarantees and green energy provision. The construction of intelligent computing centers should be coordinated with power grid planning, the integration of new energy sources, and energy storage for peak shaving.
Activating data as a factor of production is fundamental to establishing efficient circulation and high-quality supply. The foundational systems for data elements must be perfected. This involves accelerating the improvement of basic systems in areas such as data property rights, circulation and trading, profit distribution, and security governance. Clear definitions of data ownership and detailed rules and responsibilities for collection, processing, sharing, opening, and trading are needed. Developing compliance templates, standard contracts, and operational guidelines for industries will help reduce the transaction costs borne by enterprises.
Streamlining the chain for compliant data circulation is essential. By focusing on key areas like trusted measurement, pricing and settlement, and profit distribution, pathways for the market-based allocation of data elements should be explored to enhance circulation efficiency. Building a strong system for high-quality data supply is a priority. Efforts should concentrate on key industries and public service sectors to rapidly develop high-quality datasets and industry-specific corpora. Clear standards for collection methods, annotation specifications, and evaluation benchmarks will improve data usability, comparability, and reusability, continuously enhancing the quality of data supply.
Implementing a robust data quality evaluation mechanism is necessary. This requires establishing a scientific and unified set of data quality indicators. Regular testing and scoring across multiple dimensions, including completeness and accuracy, will generate quality profiles that allow for comparison and benchmarking.
Expanding application scenarios is vital to propel artificial intelligence towards large-scale commercial use. Accelerating the adoption of intelligent terminals and agents is key. Promoting the coordinated development of next-generation smart terminals and agents will create a supply structure based on device-cloud synergy. In the consumer sector, the widespread application of AI phones, AI PCs, and smart homes should be accelerated, driving simultaneous upgrades in on-device inference, voice interaction, multimodal capabilities, and application ecosystems. In the industrial sector, integrating intelligent agents into high-frequency scenarios such as collaborative office work, R&D management, supply chain operations, customer service, and maintenance inspections will facilitate end-to-end task execution.
Concurrently, capability inventories should be refined, specifying requirements for data, computing power, models, and system integration to enhance the standardization and precision of scenario implementation. Improving third-party evaluation and gradual rollout mechanisms will boost the sustainable operational capability and production conversion rate of AI applications.
Perfecting the mechanism for achieving a commercial closed loop is important. Exploring models such as subscription fees, payment-by-results, and savings-sharing should be encouraged. Promoting the productization, standardization, and scalable replication of solutions will accelerate the transition of AI from demonstrative applications to commercial, large-scale use, truly converting it into new drivers of development and increments of new quality productive forces.
Fostering an innovation ecosystem is essential for cultivating intelligent-native business formats and governance capabilities. Supporting businesses in using AI to reshape underlying architectures and operational logic is crucial. Focusing on key links like R&D design, production organization, supply chain collaboration, customer service, and risk control will accelerate the development of intelligent-native products, services, and business models.
Cultivating an open-source ecosystem for artificial intelligence is necessary. The significant role of open source in lowering innovation barriers, accelerating technological iteration, and promoting the diffusion of成果 should be fully leveraged. Collaborative development of open-source toolchains, model components, evaluation benchmarks, and security capabilities should be promoted.
Enhancing the governance of artificial intelligence is paramount. Adhering to a people-oriented approach and ensuring AI is used for good requires establishing a governance system that covers the entire lifecycle of data, models, and applications. High-risk scenarios should be managed through classification and tiering. Strengthening pre-deployment security assessments and standardized management, clarifying responsible entities, and improving mechanisms for risk identification, emergency response, and复盘改进 are vital.
Continuous post-deployment monitoring and periodic re-evaluation should be strengthened, with a focus on preventing risks such as model hallucination, algorithmic bias, data leakage, and unauthorized access, forming a closed loop of monitoring, early warning,处置,复盘, and improvement.
Comments