DeepMind's Genie 2 is the next-gen model capable of generating diverse 3D worlds from a single image and text description, offering endless interaction possibilities.
Genie 2 can create consistent, playable environments that mimic AAA video games, leveraging a large dataset of playthroughs, though details of its training remain undisclosed.
The model demonstrates nuanced understanding of object interactions, intelligently responding to user inputs, and making distinctions between characters and environment elements.
Concerns arise over potential intellectual property issues surrounding Genie 2's training, particularly regarding its access to YouTube for model development.
Collection
[
|
...
]