In August, we previewed Genie 3, a general-purpose world mannequin able to producing numerous, interactive environments. Even on this early type, trusted testers had been capable of create a formidable vary of fascinating worlds and experiences, and uncovered fully new methods to make use of it. The following step is to broaden entry by a devoted, interactive prototype centered on immersive world creation.
Beginning in the present day, we’re rolling out entry to Undertaking Genie for Google AI Extremely subscribers within the U.S (18+). This experimental analysis prototype lets customers create, discover and remix their very own interactive worlds.
How we’re advancing world fashions
A world mannequin simulates the dynamics of an setting, predicting how they evolve and the way actions have an effect on them. Whereas Google DeepMind has a historical past of brokers for particular environments like Chess or Go, constructing AGI requires programs that navigate the range of the actual world.
To satisfy this problem and help our AGI mission, we developed Genie 3. In contrast to explorable experiences in static 3D snapshots, Genie 3 generates the trail forward in actual time as you progress and work together with the world. It simulates physics and interactions for dynamic worlds, whereas its breakthrough consistency permits the simulation of any real-world situation — from robotics and modelling animation and fiction, to exploring areas and historic settings.
Constructing on our mannequin analysis with trusted testers from throughout industries and domains, we’re taking the subsequent step with an experimental analysis prototype: Undertaking Genie.
How Undertaking Genie works
Undertaking Genie is a prototype internet app powered by Genie 3, Nano Banana Professional and Gemini, which permits customers to experiment with the immersive experiences of our world mannequin firsthand. The expertise is centred on three core capabilities:
