Initially, Google had introduced the world to Imagen, a competitor to DALL-E 2 that has some startingly good results. However, the tech giant has currently taken Imagen a step ahead by making a text-to-3D A.I. that could turn the words into full 3D models.

HIGHLIGHTS
Imagen moved ahead to make text-to-3D A.I. that turns words into full 3D model
The text-to-3D A.I. image generator has presently being referred as DreamFusion
The possibilities of this A.I. taking over the world were slim

Google Brain’s researcher Ben Poole had shared the news of the new text-to-3D A.I. on Twitter. Poole had mentioned that he had worked alongside Ben Mildenhall, Ajay Jain, and Jon Barron to form the AI, that depends heavily on Google’s Imagen image generator.

As per the paper that the researchers have shared on the arXiv servers, the program has been a proof-of-concept. The text-to-3D A.I. image generator has presently being referred as DreamFusion, along with an evolution of Dream Fields, a text-to-3D generator that Google had revealed back in 2021.

The biggest distinction here between the text-to-3D A.I. found in DreamFusion and Dream Fields was that Dream Fields relied on OpenAI’s CLIP tech. OpenAI had made some spectacular things within the past, too, such as fake news generator that it had been too afraid to unleash.

Though, the researchers have produced the 3D models just by using a pre-trained text-to-image diffusion model, Poole explained in his tweet. This removes the necessity for any 3D information.

It has been unclear what kind of overall use a product like DreamFusion, or extremely any text-to-3D A.I. image generation may had it overall. Well, Google’s proof of concept has been intriguing and with a lot of research and studying, it might prove worthwhile for expansion. Currently, the results that we were seeing were extremely exciting and would no doubt push different A.I. generators to follow the same route in the future.

Of course, the possibilities of this A.I. taking over the world were slim. However, this may create some parts of model generation to be much easier and simpler in the future.

Also Read: Ex-Google ad boss of Indian origin builds tracker-free search engine