For the technology company, games are “the research center” and “play a fundamental role in the world of AI research”, with works such as Atari games, advances such as AlphaStar and Alpha Go, among other developments.
Genie 2 allows the future agents train and enjoy evaluating a program unlimited with the creating new worlds in 3D and the development of prototypes interactive experiences.
How Genie 2 works
Genie 2 can simulate virtual worlds along with the impact of doing any action (swimming, jumping, running, etc.), since it was trained composed of video data on a large scale that allow with the others generative modelsperform interactions and animations.
He operation is based on receiving a single image generated by Image 3the model of text to image next-generation GDM, which means anyone can upload their favorite idea, creation, and representation of their world, through text.
After completing this process, people can interact with that space created or even ask someone AI agent to train and work on it. Genie 2 has the ability to generate worlds for up to one minutewhich demonstrates an innovative and captivating gaming idea for users, creating their own adventure.
What creation alternatives exist?
Users have to entertain themselves because the world of odds with Genie 2 it is truly infinite due to its power to recreate spaces, places, objects and people. The benefit also opens great opportunities for creators and artists, since they have the chance to generate prototypes in an agile, dynamic and realistic way.
Some of the many characteristics and options it offers are: object interaction, character animation, creation of other agents, water and smoke effects, creation of gravity, different perspectives, lighting and prototypes.
Google DeepMind demonstrated his great capacity for creation and development of Genie 2, to make visible the potential of world models through 3D environments and the acceleration of agent investigationwhose developments are still in their first stages.
Source: Ambito
