Main information
- Genie 3 of Deepmind, the Google AI tool, generates interactive 3D environments that users can explore in real time.
- Users can interact with these worlds for several minutes and the model remembers space information for about a minute.
- Deepmind is currently distributing Genie 3 as a limited research-research premier due to the potential risks associated with technology.
Google Deepmind’s latest model of AI, Genie 3, represents a significant advance. This technology makes it possible to create interactive 3D environments that users can explore in real time. Unlike traditional video games whose elements are handcrafted, Genie 3 generates these worlds entirely thanks to artificial intelligence.
Applications and improvements
Genie 3 users can enter an prompt to generate a space in which they can navigate freely, as in a video game. While the previous versions, such as Genie 2, had limits in terms of interaction time (up to one minute) and visual loyalty, Genie 3 promises substantial improvements.
GENIE 3 allows users to interact with the worlds generated for several minutes continuously, which represents a notable increase compared to the limit of 10 to 20 seconds of Genie 2.. In addition, the model can keep spatial information for about a minute, which means that objects such as paint on a wall or writing on a black board will remain in their original position even if the user diverts the look.
Restrictions and future access
However, GENIE 3 is currently broadcast as a limited research-research, accessible only to a limited group of academics and creators. This approach allows Deepmind to better understand the potential risks associated with technology and to develop appropriate mitigation strategies.
Restrictions are in place, in particular limitations on user interaction methods in the generated worlds. Nevertheless, Google explores the possibilities of expanding access to Genie 3 in the future.