On 520, 4 Million AI Enthusiasts Tuned into Nearly 20 Speeches & Dialogues at QbitAI's 4th China AIGC Industry Summit
The summit, held on May 20, 2026, brought together nearly 20 AI leaders to discuss the future of AI agents, multimodal AI, spatial intelligence, and the transition from AIGC to AIGA (AI-generated action). Key themes included agent-as-product, the rise of long-term memory, and the infrastructure shift from training to inference computing. The event attracted 4 million online viewers.
On May 20, 2026, QbitAI hosted the 4th China AIGC Industry Summit, a landmark event that drew nearly 4 million online viewers and over 1,000 on-site attendees. The summit featured nearly 20 speakers from leading AI companies and research institutions, discussing the most pressing topics in AI: agents, multimodal models, spatial intelligence, and the evolving AI infrastructure.
The summit kicked off with Fang Han, Chairman and CEO of Kunlun Inc., who spoke about how individuals and enterprises should cope with the impact of agents. He highlighted that agents are becoming ubiquitous, with token consumption skyrocketing—from millions per month for ordinary employees to billions for heavy agent users. He argued that five types of people are irreplaceable by AI: storytellers, idea creators, definers of beauty, system builders, and paradigm resetters. He also noted that in most industries, being second is more advantageous than being first, as the first bears high exploration costs.
Yi Zhengchao, CEO of Funshion Online, shared insights on how AI coding and AI video are transforming the entertainment industry. He emphasized that AI lowers costs and barriers, and that crowdsourcing will be the key lever for AI productivity. He described a future where companies consist of employees, digital workers, and external partners, forming a socialized ecosystem.
Lin Dahua, Executive Vice President and Chief Scientist of SenseTime, discussed the journey from multimodal unification to spatial intelligence. He noted that integrating diverse data types (charts, Excel, images, video) into a single AI system accounts for over 70% of enterprise costs. He introduced SenseNova U1, a model that unifies understanding, reasoning, and generation, enabling seamless switching between text and visuals. He stressed that true AI must bridge the digital and physical worlds, and that future agents will need to operate in both domains.
Deng Yafeng, Vice President of Shanda Group and CEO of EverMind, drew parallels between the agent product 'Longxia' (Lobster) and the iPhone 4, calling it a paradigm-defining product. He emphasized that long-term memory is the key to agent autonomy and self-evolution. He argued that as models become more capable, memory becomes the most differentiated asset, and that personal memory should be synchronized across different agents.
Wang Xiaoye, Technical Director at AWS, addressed the gap between powerful models and enterprise-grade AI agents. He proposed a five-layer framework for building agentic AI: from inference computing, multi-model selection, enterprise data and knowledge, agent-building platforms, to ready-to-use agent applications. He highlighted that coding agents are mature, and working agents are the next breakthrough.
A special GenAI Talk session featured Shen Yujun, Chief Scientist of Ant Lingbo Technology, in dialogue with QbitAI co-founder Li Gen. Shen introduced the concept of AIGA (AI-Generated Action), arguing that AI must move from the digital world into the physical world. He predicted that in 1-2 years, benchmark examples will emerge; in 2-3 years, mass replication; in 5-10 years, robots entering C-end; and after 10 years, home robots will become common.
Qiu Xipeng, Professor at Fudan University and founder of Mosi Smart, presented on multimodal models and inference optimization. He introduced MOSS-VL for video understanding and MOSS-Audio for speech and scene understanding, emphasizing real-time interaction and long context handling.
Other speakers included Hu Weiqi from MiniMax, Zhu Guangxiang from Baidu, Hong Yuan from TaiChu, Huang Chao from HKU, Dai Wenjun from JD, Ma Xiaowu from Qingsong Health, and Zhang Lu from Fusion Fund. They covered topics such as democratizing AI, the rise of OPC (one-person companies), the shift from training to inference computing, and the importance of data quality.
The summit also featured a roundtable discussion on agent products and non-consensus opportunities, with participants from Quwan Technology, ColaOS/MarsWave, and EvoMap. They discussed the rapid iteration of agent products, the failure of traditional metrics like DAU, and the need for agents to provide emotional value.
Finally, QbitAI released the '2026 AIGC Enterprises to Watch' and '2026 AIGC Products to Watch' lists, along with the '2026 China AI Application Panorama Report.' The summit underscored that AI is no longer just a tool but a productivity system, and that the era of agents is here.