Roblox launches and open 3D cubes, the first version of its fundamental AI model to generate 3D objects.
“Thanks to Cube, we intend to increase the performance of 3D creation,” says Roblox In a press release. “By generating 3D mesh, programmers can quickly examine new creative directions and increase their performance by undertaking quickly deciding what you can move forward with.”
The model is trained on 3D facilities, as Roblox explains in a press release:
To achieve this, we aroused inspiration from the latest models trained on text tokens (or sets of characters) so that they could predict the next token to create a sentence. Our innovation is based on the same basic idea. We have built the ability to tokenize 3D objects and understand shapes as tokens and trained 3D cube to predict another shape token to build a complete 3D object.
The actual data used to train the model includes “a combination of licensed and publicly available data sets, as well as data on the Roblox ecosystem,” says Samantha Spielman spokeswoman The Verge.
Down the Cube 3D line will be able to generate objects using images as input data. “Ultimately, it will be a multimodal model, trained in the field of text, images, video and other types of input data – and integrates with our existing tools to create artificial intelligence,” according to Roblox.