Academician Zhao Qinping proposes 'JuRen Intelligence': human-centric, multi-form factor, companion-style AI

The First Immersive Consumption Carnival and the 2026 AI Terminal Innovation Development Conference were recently held at the Tongming Lake Development Center in Beijing's Yizhuang district. The event brought together academicians, experts, and representatives from industry chain enterprises for in-depth discussions on the evolution path of AI terminals.

A growing industry consensus has emerged: the AI industry is shifting from a machine-centric paradigm of embodied intelligence to a human-centric new paradigm of "human-oriented intelligence" (具人智能). On-device AI agents have become the core battleground for the next generation of human-computer interaction interfaces.

"Human-Oriented Intelligence" Makes AI More Understanding of People

"If 'embodied intelligence' answers the question of how to make machines smarter at doing things, then 'human-oriented intelligence' answers how to make intelligence better understand people, better couple with humans, and make people more powerful," said Zhao Qinping, an academician of the Chinese Academy of Engineering, in his keynote speech at the conference, introducing a new industry proposition.

In his view, embodied intelligence focuses more on humanoid robots, following a machine-centric technical route. In contrast, the larger-scale AI普及 over the next decade will be driven by various consumer terminals that accompany people closely, opening up a new human-centric AI赛道.

Zhao explained that "human-oriented intelligence" has three core tenets:

• Human-centric: AI terminals are no longer cold tools but partners that collaborate with humans following their intentions.
• Multi-form carriers: AI glasses, rings, wristbands, AIPC, and AI phones together form a贴身 perception matrix.
• Accompanying evolution as the goal: Devices continuously learn user habits, becoming more understanding and默契 over time.

Under this new paradigm, AI terminals will evolve toward diversified form factors, natural and seamless interaction, and cross-device interconnectivity and collaboration, while adhering to technological inclusivity so that the elderly and people with disabilities can also benefit from intelligent dividends. Zhao called for coordinated efforts from industry, academia, government, and society to strengthen foundational research in on-device large models and multimodal perception, address gaps in standards, applications, and talent, and leverage China's complete electronics industry chain to seize the first-mover advantage in developing "human-oriented intelligence."

Industry Enters the Second Half of AI Deployment

The implementation of human-oriented intelligence requires terminals to shift from passive response to proactive execution, with on-device AI agents serving as the technical载体 for this transformation. The competition over large model parameters has come to an end, and the industry has entered the second half of AI deployment—putting AI to practical use through inference.

Ji Zhaohui, Vice President of Marketing for AMD Greater China, stated in a special report that AI agents represent the watershed moment for AI evolving from conversational bots to productivity tools. "Given a target instruction, an AI agent can complete reasoning, break down subtasks, call tools, and ultimately output execution results. Token consumption can increase 5 to 30 times compared to ordinary conversations."

Ji noted that rising cloud-based token costs, network latency, privacy compliance requirements, and data sovereignty demands are driving a large number of inference tasks to migrate to the edge. He proposed a new product category—the "AI agent host"—which operates 7×24 locally running AI agents without requiring direct human interaction. "Traditional computers need keyboards, mice, and monitors. An AI agent host doesn't need them. It receives task instructions and works continuously in the background." Combined with lightweight on-device large models, such devices can reduce enterprise token costs by over 80%.

Tao Hongzhi, Chief Standards Officer of Lenovo Group, provided an industry benchmark from the standards perspective. He introduced that China has issued a national standard for AI terminal grading, which classifies intelligence levels from L1 to L4 based on five capabilities: perception, cognition, execution, memory, and learning. Current mainstream products are transitioning from L1 and L2 toward L3 (assistance level).

Jiang Wen, Vice President of ZTE Corporation, shared consumer product practices using the Nubia NaviX Ultra, the world's first AI agent phone, as an example. He proposed four core product capabilities: "understands speech, can execute tasks, remembers information, and ensures security." Leveraging fingerprint-authenticated AI buttons, multimodal perception, and layered permission management, the phone can execute complex tasks across multiple apps with a single voice command, freeing users from频繁 tapping and switching between applications.

The Ecosystem for Developing "Human-Oriented Intelligence" Is Maturing

While new paradigms and products continue to emerge, the large-scale普及 of AI terminals still faces multiple real-world challenges in underlying technologies and hardware form factors, requiring a closed loop across industry, academia, research, and application.

Academician Huang Wei of the Chinese Academy of Sciences emphasized at the conference that flexible electronics are the foundational underlying technology supporting the next generation of AI terminal innovation. "Advanced materials are the most core and foundational layer. Above that, one soft element is artificial intelligence, and one hard element is flexible electronics—the two empower each other." Huang stated that flexible electronics possess four key characteristics: lightweight, thin, flexible, and transparent. These properties can support wearable devices, brain-computer interfaces, and next-generation display iterations, providing the hardware foundation for the various贴身 terminals required by "human-oriented intelligence." He urged China to seize the window of opportunity for future industries, strengthen investment in foundational research, and open up new fields and tracks.

Shi Qing, General Manager of Yijia Virtual Reality Technology Co., Ltd., offered practical observations from the hardware manufacturing front line. He believes that AI glasses represent the long-term terminal form factor, but current bottlenecks in wearing comfort, appearance, and power consumption remain significant, making large-scale普及 difficult in the short term. Instead, diverse wearable devices such as wristbands, rings, pendants, and badges will develop in parallel, prioritizing three high-value scenarios: productivity and office work, health and stress management, and emotional and social interaction. The hardware industry is shifting from competition over single devices to competition over the entire end-edge-cloud system capability. All-day proactive perception, proactive memory, and proactive service are important evolution directions for next-generation wearable hardware.

The Beijing Economic-Technological Development Area (BDA) has already positioned itself as a global AI terminal product definition center and intelligent manufacturing hub. It has gathered over 600 core AI enterprises, with the industry scale exceeding 80 billion yuan. The BDA has established shared factories, pilot platforms, and an AI terminal global experience center, using real urban scenarios to validate new products.

Attendees generally agreed that China possesses unique advantages for developing the "human-oriented intelligence" ecosystem, including a complete manufacturing supply chain, a vast consumer market, and abundant testing scenarios. Going forward, the industry needs to fill gaps in core components,完善 interoperability standards, develop "killer" applications, balance capability iteration with privacy and security, and drive AI terminals beyond parameter competition—truly returning technology to serve people.