At the congress of the iupw 2026, the chinese academy of engineering, zhu xian xian, noted that the rapid development of artificial intelligence and large model technology was driving intelligence into the focus of industry. By 2026, global reasoning is expected to account for 70 to 80 per cent of the total ai load, while china’s reasoning needs eight times as much as training needs. The shift from a training-based to a reasoning- and intelligent-driven application model has led to profound changes in the architecture, application landscape and business model. The development of smarts has led to a shift in the algorithm structure from a focus on the economy to a productive economy, with deployment patterns extending from a pure cloud to a cloud-side-end synergy, and a business model from a cost centre to a value engine。
The quizumi explained that intelligence is not a single host, but rather a capability and resource based on a large model. Individual intelligent bodies are able to independently fulfil the decision-making and operational aspects of a given mandate by creating a closed circle of perceptions, memories, planning, tools and actions. However, the application boundaries and capacity caps for such a single intelligence are limited and lack the capacity to collaborate across intelligence. To break this limit, the multi-intellectual synergetic system was created. Under the ucc, the relevant intelligence bodies are organized to achieve collective intelligence through shared memory and advanced planning, thus addressing complex objectives。
The size of multi-intellectuals in open networks is interconnected to form the ioa. Unlike traditional internet hosts and information interconnections, ioa has expanded to intelligent and capable interconnectivity. It is based on ipv6+, superimposes new application layer protocols, achieves the only identifier of an intelligent body, autonomous discovery, credible interaction, synergistic implementation and a needs-based network. Smart internet supports smart bodies, namely, services (aaas), cloud-capturing the capabilities of intelligent bodies and intelligent clusters, providing external services through api, sdk and low-code platforms, and supporting renting, flexible movement control, hosting peacekeeping on a needs-based basis。
By comparing aaas to a smart-body supermarket, business users do not need to develop their own intelligence or build their own financial capacity, but simply use an encapsulated mix of smart-body functions from the cloud through subscription modes to integrate into their own business systems, i. E., achieve the dual benefits of automation and intelligentization of saas costs. This model significantly lowers the threshold for smes to use smarts and promotes technology diffusion。
There are three main modes of application for consumer-oriented smarts (to c). The first is the app parasite of the cloud, which relies heavily on the app of the host, such as micro-intelligence, nails and so forth, which operates with cloud-based computing power and is not available. This mode upgrades the chat tool to a smart assistant who can “think, write, read, and control”, but the user has no model choice. The second is a terminal os prototype, pre-positioned in a mobile phone operating system, such as a soybean cell phone launched jointly with an active commotion and a byte beat, a cloud synergizing, which can move aps as intended by the user and automatically execute business processes. The advantage is that it is more user-controlled, but it may weaken the app interface and trigger resistance from application manufacturers. The third is the local autonomy + cloud-assisted intelligence body, which does not rely on app, but which manually installs independent pc software, such as the "crawfish" ai private housekeeper, localizes decision-making and authority management, and works with other intelligent bodies through the cloud aaas。
There are four applications for enterprise-oriented smarts (to b). Process automated smart body retraces manual logic, replaces standardized processes such as production, transport, etc., and improves efficiency. Industry experts ' intelligent bodies rely on industry knowledge base and knowledge mapping to provide a decision-making staff programme. Multi-intelligence synergizes and integrates multiple areas of intelligence such as supply chains, production, logistics and so forth. Physical equipment, such as intellectual support for industrial robots, completes the closed loop of perception, reasoning and physical execution. It was pointed out that local start-up + cloud actuarials were the best solution to the problem of weak self-accounting and high data privacy risks for smes, balancing costs and safety。
Smart business places subversive demands on algorithms, clouds and networks. Compared to traditional internet operations, the driving agents of intelligent body operations have moved from manual operations to machine-owned cycles, from a symmetrical to 7x24 hours of uninterrupted operation, and from centralized to multi-point network, multi-synthetic and cascaded transfers. The demand for algorithms by smarts has surged, with single-intelligence token consuming 10 to 1,000 times as much as ordinary users. Although token traffic is increasing at an annual rate of 330 per cent, efficiency and cost balance can be achieved by spreading, compressing and optimizing the context through small models, which can be contained at around 41 per cent over the next five years。
The age of intelligent bodies has also re-engineered cloud services. Traditional clouds are centred on the availability of resources, while the age of intelligent bodies requires intelligent bodies to have raw clouds, to prioritize the layout of intelligent bodies, to build global intelligent deployments, to build distributed memory and vector engines, to deepen the cloud-side-end synergy, and to enhance computing and network optimization. This change is accelerating the reshaping of china's cloud service landscape, with the outbreak of ai cloud-wide services, market integration to smart, green low-carbon and deeper domestic production, covering the intellectual infrastructure, model training/adjection, intelligent body organization and mobility。
Network upgrades are equally urgent. The continuous interaction of long-term intelligent conversations, multi-intellectual synergetic communication, large model reasoning/retribution of high-frequency (hf) packages + high-flow hybrids have resulted in a shift in network traffic from north to south to east to 80 per cent and an increase in qps by 10 to 100 times. The network needs to move from a network of administrative units to a four-tier structure of a regional cloud-margin-end centre, using techniques such as srv6, cloud network virtualization and organization, traffic ai prediction and movement control, to ensure certainty and high reliability。
