GUWAHATI: Google DeepMind researchers have unveiled a new AI model named Genie, capable of creating virtual worlds using text or image prompts, media reports stated.
ALSO READ: Prominent Congress Leader Rana Goswami Likely To Join BJP
This model, currently in a research preview, was trained on online gameplay and videos.
According to Google DeepMind's official blog post, Genie can generate a wide variety of playable worlds from synthetic images, photos, and sketches.
Click Here To Join Our WhatsApp Channel
The research paper, titled 'Genie: Generative Interactive Environments,' highlights that Genie is the first generative interactive model trained in an unsupervised manner from unlabeled internet videos. With 11 billion parameters, Genie includes a spatiotemporal video tokeniser, an autoregressive dynamics model, and a scalable latent action model, reports further stated.
As per reports, the Genie researchers said that the AI was trained from Internet videos that can "generate an endless variety of playable (action-controllable) worlds from synthetic images, photographs, and even sketches."
Furthermore, Genie can reportedly function on a frame-by-frame basis in generated environments, even without training, labels, or other specific requirements.
Unlike many other AI models, Genie can produce playable environments from just a single image prompt.
The researchers claim that Genie can be prompted with unseen images, like real-world photographs and sketches, allowing users to interact with their imagined virtual worlds, known as a foundation world model.
The model is trained to focus on 2D platform games and robotics videos but can adapt to various domains and larger internet datasets.