[Former ByteDance Intern Tian Keyu, Sued for 8 Million Yuan, Reveals Startup Plan: 200,000 Visual Tokens to Cut Video Costs by 90%] Beating AI News Flash: Former ByteDance intern Tian Keyu has disclosed his world model startup plan. His team has raised nearly $30 million in funding from institutions such as FiveYuan Capital and IDG Capital, with a post-money valuation of $200 million. The team currently consists of about 10 members and has not yet publicly released a product. The world model aims to enable AI to learn how objects move and change in the real world, with applications in video generation, robotics, and other fields. Tian Keyu is the first author of the Best Paper at NeurIPS 2024 and has previously researched how AI can progressively generate images from coarse to fine. The new team is building on this direction by designing a "vocabulary" of 200,000 visual tokens for video. These tokens function similarly to the tokens used by language models to process text, allowing AI to handle video in a way that is more computationally efficient. The team plans to train the model using 100 million hours of video. Tian Keyu claims that this approach could reduce the cost of generating video per second to one-tenth or even less of the current level. The team intends to first release a public demonstration and launch the full model in 2027. He also admitted to Bloomberg that he had previously shut down another intern's program, claiming it was to free up computing resources for other researchers, and acknowledged that his actions at the time were inappropriate. ByteDance dismissed him in 2024 and subsequently filed a lawsuit seeking 8 million yuan in damages. According to a judgment reviewed by Bloomberg, the court ordered Tian Keyu to pay 500,000 yuan for breach of contract. [Original Link]