Qiming Star | Video Rebirth's Wei Liu: Video Generation Models Are the Optimal Path to Building World Models

The Qiming Venture Partners-hosted 2025 World Artificial Intelligence Conference (WAIC) "Qiming Venture Partners · Entrepreneurship and Investment Forum — Venture Capital Unleashing the Resonance Cycle of AI Technology and Applications" was successfully held on July 28 in the Blue Hall of Shanghai World Expo Center.

At this forum, Wei Liu, former Tencent Distinguished Scientist and technical lead of the Hunyuan large model, and now CEO of Video Rebirth, delivered a keynote speech titled From Video Generation to World Models.

Wei Liu, former Tencent Distinguished Scientist and technical lead of the Hunyuan large model, CEO of Video Rebirth

Liu stated that video generation models represent the optimal path to building world models, and this technical direction could become the critical breakthrough for AI's leap from perception to cognition.

Liu believes AI technology is evolving through four important stages: from symbolism to connectionism, and then to realism and generativism. Of particular note, OpenAI's o1 model released in September 2024 granted AI genuine reasoning capabilities for the first time, marking a significant breakthrough in 40 years of connectionist exploration. In 2025, AI stands at a critical inflection point: software is transitioning to agents, and intelligent hardware is evolving toward physical AI.

On the technical framework, Liu proposed that world models should possess three core capabilities: simulation for emulation functions, prediction for causal reasoning, and exploration to support planning and decision-making. Simulation belongs to fast-thinking perception, prediction enters the realm of slow-thinking cognition, and exploration represents active thinking — these three capabilities are precisely what define world models.

Although multimodal models like GPT-4o can process text, voice, and image inputs and outputs, such models remain in passive response mode, lacking holistic environmental modeling and predictive capabilities. By contrast, world models need to shift from passive to active thinking, moving from responding only when asked to being capable of serialized autonomous reasoning.

SORA's emergence provided important inspiration for world models. Liu pointed out that SORA first proved the feasibility of world models, achieving highly realistic spatiotemporal simulation through video generation. While the current version has certain limitations, it has already provided a practical technical starting point and viable path for constructing world models.

Video Rebirth defines the video-native world model as a combination of world simulator plus world predictor. Current mainstream DiT architectures, while possessing powerful spatiotemporal simulation capabilities, suffer from critical deficiencies including lack of causal reasoning and inability to interact or intervene. Going forward, the company is committed to a technical upgrade path to address these issues, building proprietary technical propositions and model paradigms, ushering in the "ChatGPT Moment" for the video generation field, and launching the first truly video-native world model.

Liu emphasized that AI needs not only grand narratives but also the creation of convincing reality. Video Rebirth's approach of entering world models through video generation could achieve important technical innovation during this critical period of breakthrough in AI cognitive capabilities, providing crucial support for the development of artificial general intelligence.

Source | IPO Zaozhidao

Past Coverage

Qiming Perspectives | Alex Zhou of Qiming Venture Partners: 2025 Will Be the Year AI Applications Land at Scale Qiming Stars | Multiple Qiming Venture Partners Portfolio Companies Appear at 2025 World Artificial Intelligence Conference Qiming Headlines | 2025 WAIC "Qiming Venture Partners · Entrepreneurship and Investment Forum — Venture Capital Unleashing the Resonance Cycle of AI Technology and Applications" Successfully Held

Founded in 2006, Qiming Venture Partners currently manages 11 USD funds and 7 RMB funds, with total committed capital reaching $9.5 billion. Since its inception, the firm has focused on investing in outstanding early and growth-stage companies in Technology and Healthcare.

To date, Qiming Venture Partners has invested in over 580 high-growth innovative companies, of which more than 210 have listed on the New York Stock Exchange, NASDAQ, Hong Kong Exchanges and Clearing Limited, Shanghai Stock Exchange, and Shenzhen Stock Exchange, or exited through M&A and other means. Over 80 portfolio companies have become recognized unicorns or super-unicorns.

Many Qiming Venture Partners portfolio companies have grown into the most influential companies in their respective fields, including Xiaomi (01810.HK), Meituan (03690.HK), Bilibili (NASDAQ:BILI, 09626.HK), Zhihu (NYSE:ZH, 02390.HK), Roborock (688169.SH), UBTECH (09880.HK), WeRide (NASDAQ:WRD), Insta360 (688775.SH), Gan & Lee Pharmaceuticals (603087.SH), Tigermed (300347.SZ, 03347.HK), Zai Lab (NASDAQ:ZLAB, 09688.HK), CanSino Biologics (688185.SH, 06185.HK), Schrödinger (NASDAQ:SDGR), MicroPort EP MedTech (688617.SH), Sanyou Medical (688085.SH), Amoy Diagnostics (300685.SZ), Berry Genomics (000710.SZ), Sinocelltech (688520.SH), Yuanxin Technology, Insilico Medicine, MediLink Therapeutics, LaNova Medicines, Zhipu AI, StepFun, Biren Technology, and others.