Google DeepMind, the tech giant’s AI research lab, announced the launch of Genie 3, which can create 3D environments at 720p resolution from text prompts.
This level of quality in the new Genie model is an upgrade from Genie 2, which operated at 10 to 20 frames per second instead of Genie 3’s 24 frames per second.
Genie 3’s environment generations can stay physically consistent over time because the model remembers what was previously generated.
Using the model, users will be able to infinitely adapt virtual worlds for industries such as gaming, robotics, training, disaster preparedness, education, and for creating immersive experiences.
A First Step Towards AGI
One of the standout features in the product is real-time interactivity – which enables users to change virtual worlds in real-time, and dynamic environment modification – which allows users to reshape virtual worlds continuously.
Google says the Genie 3 model could be used to train robots and autonomous vehicles to interact with realistic environments.
The company also highlights the advent of the model being a first step to achieving artificial general intelligence (AGI) – an advanced form of AI that would be able to reason and carry out tasks at the same level of humans.
This is because the model could change the way AI interacts with and understands the world.
During a brief, Jack Parker-Holder, a research scientist at DeepMind, said: “We think world models are key on the path to AGI, specifically for embodied agents, where simulating real world scenarios is particularly challenging.”
