SenseTime's Lin Dahua: AI Progress Doesn't Depend Only on Stacking Large Models
Lin Dahua, co-founder and chief scientist of SenseTime, said at CITIC Securities' 33rd Investor Forum that the key to AI progress is not just scaling up models, but also smarter ways of building them. He emphasized that the "intelligence density" of data determines learning outcomes, which is why SenseTime developed a native multimodal model that integrates multimodal capabilities into the architecture and training from the very beginning. Lin Dahua expects embodied intelligence to accelerate over the next one to two years, and SenseTime is also continuing to explore world models.
Body
𝗕𝗲𝘆𝗼𝗻𝗱 𝗕𝗶𝗴𝗴𝗲𝗿 𝗠𝗼𝗱𝗲𝗹𝘀: 𝗪𝗵𝗮𝘁 𝗗𝗿𝗶𝘃𝗲𝘀 𝗔𝗜 𝗣𝗿𝗼𝗴𝗿𝗲𝘀𝘀 AI progress isn't just about scaling up. It's about building smarter.
At CITIC Securities's 33rd Investors' Forum, our Co-Founder and Chief Scientist Dr @lindahua shared his views on where AI is heading next.
𝗠𝗼𝗿𝗲 𝘃𝗮𝗹𝘂𝗮𝗯𝗹𝗲 𝗱𝗮𝘁𝗮 𝗺𝗲𝗮𝗻𝘀 𝗯𝗲𝘁𝘁𝗲𝗿 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴. The value of data lies in its “density of intelligence” — how well it captures the steps involved in complex tasks. By turning industry data into realistic training scenarios, AI agents can learn and improve faster. At SenseTime, over a decade of working closely with customers has helped us continuously improve our models through real-world feedback.
𝗠𝘂𝗹𝘁𝗶𝗺𝗼𝗱𝗮𝗹𝗶𝘁𝘆 𝗶𝘀 𝗯𝗲𝗰𝗼𝗺𝗶𝗻𝗴 𝗳𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻𝗮𝗹. Describing an image in text can miss subtle but crucial details. That's why we develop native multimodal models, with multimodality built into their architecture and training from the start.
𝗔𝗜 𝗶𝘀 𝗺𝗼𝘃𝗶𝗻𝗴 𝗯𝗲𝘆𝗼𝗻𝗱 𝘁𝗵𝗲 𝘀𝗰𝗿𝗲𝗲𝗻. Much of the real economy operates in the physical world. Combining high-level intelligence with the ability to act in the physical world could accelerate embodied AI, which Dr. Lin expects to gain momentum in the next one to two years. We continue to explore world models that could help AI better understand and predict the physical world.
Ultimately, the true measure of AI progress is how reliably it performs in the real world and the value it creates.