World Labs: 3D world generation model
Here is the translation of the contents:
The article discusses what World Labs company is doing, which uses 3D model generation and spatial intelligence technology to create an interactive and controllable virtual world. The company’s founder is Jia Li, a former professor at Stanford University.
Jia Li’s team has been trying to teach computers “how to act in 3D worlds”, such as using large language models to let robots execute tasks. Now they are exploring spatial intelligence.
The article mentions that World Labs has raised $230 million in venture capital, valued over $1 billion. It quotes Jia Li saying that spatial intelligence is the key puzzle for the AI problem, which can allow AI systems to understand complex physical worlds and relationships between objects within them.
The article also mentions OpenAI’s Sora model, although it can generate videos, it remains 2D and lacks a deep understanding of three-dimensional space. Jia Li says that this lack of understanding prevents it from performing complex operations such as generating videos with cameras placed behind characters.
In summary, the article discusses the spatial intelligence technology being explored by World Labs company, which they hope to create an interactive and controllable virtual world, and believe it is a key solution to the AI problem.
Translation
文章讨论了World Labs公司正在做的事情，它使用3D生成模型和空间智能技术创造出一个可以与用户交互的、可控性强的虚拟世界。该公司的-founder是李飞飞，前斯坦福大学教授。
李飞飞团队早就开始试图教会计算机“如何在3D世界中行动”，如使用大语言模型让机械臂执行任务。现在他们正在探索空间智能。
文章提到World Labs已经筹集了2.3亿美元的风投资金，估值超过10亿美元。它引述了李飞飞关于空间智能的说法，认为这是人工智能难题的关键拼图，可以让AI系统理解复杂的物理世界及其内部物体相互关系。
文章还提到了OpenAI的Sora模型虽然能够生成视频，但仍然是二维的，没有对三维空间的深入理解。李飞飞称这缺乏了对三维空间的理解，无法进行像将摄像机放在女子背后的视频生成等复杂操作。
总之，文章讨论的是World Labs公司正在探索的空间智能技术，它们希望创造一个可以与用户交互的、可控性强的虚拟世界，并认为这是人工智能难题的关键解决方案。