Digital News

　　From ChatGPT to DeepSeek, and from virtual algorithms to large-scale multimodal models, AI is pushing the boundaries of human cognition and reshaping countless industries. At the 2025 World Artificial Intelligence Conference, more than 3,000 cutting-edge exhibits were unveiled, including over 40 large models, 50-plus AI terminal products, more than 60 intelligent robots, and upwards of 100 major new releases making their global or China debut. An application boom driven by the “AI+” wave has officially arrived. Amid this surge, a key priority for the successful implementation and future development of AI is ensuring its integration is robust, thorough, and secure.

　　The Ubiquitous “AI+”

　　As AI technology continues to evolve, its integration across industries is accelerating, and the concept of “AI+” is gaining widespread recognition. At the 2025 World Artificial Intelligence Conference, a broad range of AI-integrated scenarios, such as the integration of AI with culture and tourism, education, industry, finance, and office applications, alongside a variety of AI-enabled terminal products, drew significant attention from participants.

　　The Robot Exhibition Area is one of the key highlights at this World Artificial Intelligence Conference. At AgiBot’s booth, the setup resembled a fully functional “robotic workplace,” with each robot assigned a specific role and showcasing its professional capabilities through engaging interactions. On one side, the robot Genie-1 (G1) demonstrated its efficiency on an industrial production line—scanning package barcodes and expertly sorting them into designated areas. On the other side, the robots actively engaged with visitors. Some took part in human-robot curling games, demonstrating the precision of players with years of experience.

　　In the “Upper Limb and Cognitive Rehabilitation Zone” at the Fourier Exhibition Hall, GRx series robots engaged visitors in cognitive rehabilitation exercises. Meanwhile, at Unitree Robotics’ booth, a “boxing ring” was set up on-site where two robots staged an impressive battle.

　　In the areas essential for the people’s wellbeing, a variety of intelligent terminal products offer fresh and innovative solutions. The AI glasses developed by Linker Technology, equipped with the Om intelligent agent, are expected to become “AI eyes” for people with visual impairments. “These AI glasses are linked to our self-developed large model, enabling visually impaired users to perceive their environment and objects in real time. They also support smooth, natural conversations with users and proactively provide real-time alerts to help them avoid obstacles,” said Hu Xueqing, co-founder of Linker Technology.

　　Ant Group showcased its AI health application “AQ,” built on Ant Group’s Healthcare Large Model. The app provides users with comprehensive and expert health support, offering more than 100 AI-driven features such as health consultations, report interpretation, disease screening, and chronic disease management.

　　Large model companies are also increasingly linking real-life applications with their model designs. At StepFun’s exhibition booth, a smart cockpit called “Clever Egg” became one of the main attractions. Equipped with StepFun’s large model, this cockpit can not only sing and chat with passengers but also communicate fluently in regional dialects such as Sichuanese and Cantonese.

　　The Next Step for “AI+”

　　According to Wang Qiang, Director of the Frontier Technology Research Center at Tencent Research Institute, there is a strong motivation across various industries to participate in the AI revolution. “It’s driven both by the excitement of rapid technological progress and the anxiety of missing out or being disrupted by new technologies,” he noted.

　　“AI is a major trend, and many worry that without transformation, they’ll be left behind.” This is what most industry professionals are thinking today.

　　Wang Qiang noticed two typical contradictory attitudes toward AI. “On one hand, some companies are eager to roll out AI projects quickly, hoping for fast results and feeling optimistic about their potential. On the other hand, many lose momentum and enthusiasm after some initial progress because current applications are mostly limited to usage scenarios such as knowledge Q&A and simple customer services. Since results are difficult to assess, expectations run high but reality often falls short, causing significant disappointment.”

　　“Actually, we shouldn’t overestimate ‘AI+’ or rely solely on fleeting enthusiasm. Instead, it requires a profound shift in understanding and a transformative journey that involves everyone,” said Wang Qiang.

　　From the perspective of AI practitioners, achieving the widely envisioned vision of “AI+” depends on the coordinated development and seamless integration of both software and hardware. To transition from the virtual to the physical world and ensure smoother AI applications, multimodal research is essential.

　　“In terms of AI applications, ‘Agent’ is the most popular concept in 2025,” said Jiang Daxin, founder of StepFun. An Agent is generally defined as an intelligent entity capable of perceiving its environment, making independent decisions, and taking corresponding actions to achieve specific goals. “We are focusing on intelligent terminal agents as a critical driver for implementing large model technology, and collaborating extensively with partners across key edge scenarios, including automotive, mobile phones, embodied intelligence, and IoT, to jointly explore consumer-end application scenarios.”

　　Yan Junjie, founder and CEO of MiniMax, noted that AI applications will continue to expand, making it possible to analyze ancient scripts, simulate flight, design astronomical telescopes, and more.

　　The good news is that AI is becoming more accessible, with usage costs increasingly manageable. Yan Junjie observed that over the past year or so, the number of parameters in large models hasn’t increased significantly, while inference costs have steadily declined. A growing share of computing power is now being directed toward research and exploration. With enhanced capacity for innovation and ever-improving algorithms, AI development is no longer the money-burning endeavor it used to be. “Making AI affordable and accessible to everyone is both our vision for its development and the original motivation behind our startup,” said Yan Junjie.

　　In Wang Qiang’s view, achieving the integration of “AI+” requires long-term commitment and greater patience from the industry. “It’s much like the way bamboo grows. For the first four years, it seems to grow slowly, but beneath the surface, its roots are quietly spreading across hundreds of square meters. Then in the fifth year, it shoots up 30 centimeters a day. It may look like the bamboo forest appeared overnight, but in fact, it’s the result of years of quiet growth underground.”

　　Governance and Security as the Cornerstones of “AI+” Sustainability

　　There’s no denying that as technologies like AI models and generative AI are widely applied in various industries, hallucinated outputs, data bias, privacy disclosure, and other risks are coming into sharper focus. As AI continues to permeate every sector, ensuring its safe and reliable operation has become a pressing global governance challenge.

　　Many in the AI field have come to recognize that governance is not a constraint on innovation but a compass that ensures it stays on the right path. Zhang Wang, Vice President of SenseTime and Chairman of its AI Ethics and Governance Committee, noted that AI governance not only guides technological research and product development in the right direction but also promptly prevents innovation from veering off course.

　　“We believe that one of the key competitive fronts in the future AI landscape will be the accuracy and authenticity of AI-generated content. This is crucial for the integration of AI models with other industries,” said Li Jing, Vice President of StepFun. Recently, StepFun launched the industry’s first DeepResearch product featuring “in-depth verification” capabilities. They also rolled out the first expert research assistant with these verification features within their app, “StepFun AI,” which can identify AI hallucinations and Internet information pollution while cross-verifying the truth.

　　Geoffrey Hinton, the winner of the Nobel Prize in Physics 2024, spoke bluntly at the 2025 World Artificial Intelligence Conference: “AI is like a very cute tiger cub. When it grows up, if it wants to, it can easily kill you. If you choose to keep it, you have only two options: either get rid of it, or find a way to make sure it never wants to harm you.” He suggested that countries should work together on research aimed at preventing AI from wresting control away from humanity.

　　China has consistently taken a leading role in AI security governance. At the 2025 World Artificial Intelligence Conference, the Global Center for AI Innovation and Governance was officially launched, establishing a key platform for international collaboration on AI governance. The center is designed to promote global exchanges, foster cooperation, and support the development of standards, ensuring that AI technologies advance in a safe, reliable, and sustainable manner.

　　“AI governance follows the ‘buckets effect,’ where even a single weak point can compromise overall effectiveness,” said Zhong Junhao, Secretary-General of the Shanghai Artificial Intelligence Industry Association (SAIA).

　　“Governance and development always go hand in hand. Ensuring AI security governance depends on the collective efforts of society as a whole. With the advancement of the ‘AI+’ initiative, all sectors are better positioned to proactively identify potential issues and take early action, ensuring that AI is integrated into a wide range of sectors safely and healthily, and ultimately into people’s everyday lives,” said Hu Xueqing.