Introduction to AI-Generated Worlds
AI companies have been experimenting with creating interactive, AI-generated worlds, similar to popular video games like Quake and Minecraft. Google DeepMind is also working on developing models that simulate the real world. A startup called Odyssey, backed by Pixar co-founder Edwin Catmull, is taking this concept to the next level with its "interactive video" technology.
What is Interactive Video?
Interactive video, as described by Odyssey, is "video you can both watch and interact with, imagined entirely by AI in real-time." This means that users can engage with the video in a way that feels like a first-person video game, but with environments that resemble the real world rather than being made up of polygons. Odyssey compares this technology to an "early version of the Holodeck," a futuristic concept from Star Trek that allows users to experience immersive, interactive simulations.
Exploring Odyssey’s Interactive Worlds
The interactive videos created by Odyssey feel like walking through a blurry version of Google Street View. Users can navigate through the generated worlds using the WASD keys, similar to a video game. There are several different worlds to explore, including a wooded area with a cabin, a shopping mall, and a parking lot in front of a large building. Each time the user enters a world, it is slightly different, as the system regenerates the environment in real-time. However, the picture quality is currently quite fuzzy.
Technical Details
Odyssey uses clusters of H100 GPUs in the US and Europe to generate the interactive videos. The company’s model uses input and frame history to generate the next frame, which is then streamed back to the user in real-time. This process can happen in as little as 40 milliseconds. The current preview is limited to two and a half minutes, after which the user must reload to continue exploring.
Limitations and Future Development
The current preview is not without its issues. Objects sometimes lack collision, and the environment can behave strangely when the user is standing still. The image quality is also noisy, with textures on buildings and other objects appearing blurry. However, Edwin Catmull believes that Odyssey is on the "leading edge" of this technology and that the community is continually improving. He acknowledges that the images are still noisy but suggests that applying neural network filters can help solve these issues.
Conclusion
While Odyssey’s interactive video technology is not yet a replacement for video games or movies, it shows promise. With the rapid evolution of AI tools, it’s not hard to imagine a version of this technology that is more polished and immersive. However, there is still a way to go before AI video can reach its full potential. As the technology continues to develop, we may see new and innovative applications for interactive video, from entertainment to education and beyond. For now, Odyssey’s preview offers a glimpse into a future where AI-generated worlds can be explored and interacted with in new and exciting ways.
