1/ The first big announcement: Google DeepMind has unveiled Genie 3
A new frontier for world models, the first AI that generates interactive 3D worlds in real time at 24 fps.
The breakthrough: minutes of consistency at 720p using only text descriptions. No pre-built 3D models required.
Let's break it down - with some examples:
2/ How does it work? Genie 3 creates worlds frame by frame, remembering up to a minute of past scenes. You enter a text prompt, navigate in real time, and the AI dynamically adapts the world to your movements. https://t.co/RrUSf1WZp1
3/ Particularly impressive: “Promptable World Events.” While navigating, you can change the weather, introduce new objects, or spawn characters—all via text commands. This opens up completely new possibilities for training AI agents. https://t.co/K6ZAuSgyAT
4/ The technical breakthrough lies in auto-regressive generation. Each new frame must take into account the entire trajectory up to that point. At 24 fps, this means 24 complex calculations per second that must access minutes of context. https://t.co/O8X9RVC7mO
5/ Practical application: DeepMind has already tested its SIMA agent in Genie 3 worlds. Longer action sequences and more complex goals are now possible. This could revolutionize robotics training and autonomous systems. https://t.co/ERDMGuyIic
6/ Limitations remain: limited scope for action, only a few minutes of interaction, imperfect representation of real locations. But the leap from static videos to interactive worlds marks a turning point for world models and AGI development. https://t.co/ZY3PyzJQVl
7/ https://t.co/sbQEO6pnHU
