Sakana AI and NYU Launch Dream-Cubed to Generate Playable Minecraft Worlds From Tens of Billions of Cubes
Sakana AI and NYU released Dream-Cubed, an artificial intelligence system trained on tens of billions of Minecraft cubes to generate playable, editable 3D environments. The project pairs a large-scale dataset with transformer models that treat discrete blocks as tokens, enabling users to mold terrain, structures and maps with block-level control.
The researchers applied discrete and continuous diffusion training objectives to the model, enabling targeted inpainting and large-scale outpainting. The dataset, code and research paper are publicly available to the research community.
From the sources (4 posts)
@sakanaailabsWe are excited to share our latest work, together with @nyuniversity: "Dream-Cubed: Controllable Generative Modeling in Minecraft by Training on Billions of Cubes." Blog: Paper: Code: https:/
@hardmaruDreaming in Voxels: How AI is Generating Playable Minecraft Worlds Generative AI has conquered images, video, text. But what about interactive 3D environments? We trained models on billions of cubes to generate fully playable, structured w
@sakanaailabsRT @hardmaru: Dreaming in Voxels: How AI is Generating Playable Minecraft Worlds Generative AI has conquered images, video, text. But what…
@togeliusText and pixels are boring. How about cubes? In Dream-Cubed, a project in collaboration with @SakanaAILabs, we train large generative models on Minecraft maps, and achieve controllable generation. A new frontier for human-AI co-creativity!