Google’s latest AI tool, Genie 2, can create entire playable 3D worlds from a single image prompt, complete with interactive objects like doors and explosive barrels. This “large-scale foundation world model” enables artists and designers to rapidly prototype environments, accelerating the creative process for environment design.
The tool’s capabilities include generating different perspectives, complex visual scenes, and physics effects such as smoke, gravity, and reflections. These can be quickly played by humans or AI agents using keyboard and mouse.
Genie 2’s out-of-distribution generalisation capabilities also allow concept art and drawings to be turned into fully interactive environments. This technology holds promise for training embodied agents safely while achieving the breadth and generality required for Artificial General Intelligence (AGI).
While still in its early stages, Genie 2 is seen as a crucial step towards solving the challenges of training AI agents safely while advancing towards AGI. The full report, including examples, can be found on Google’s Deepmind sub-site.
Source: https://www.gamesindustry.biz/googles-genie-2-ai-tool-can-generate-a-playable-3d-world-from-a-single-prompt-image