Skip to content

DeepMind, a renowned AI company, introduces Genie 3, its latest world model

Artificial intelligence pioneer Google DeepMind unveils Genie 3, a groundbreaking 3D digital environment model. This latest advancement serves as a significant leap towards the development of artificial general intelligence.

"DeepMind, Google's artificial intelligence research company, unveiled Genie 3, a groundbreaking...
"DeepMind, Google's artificial intelligence research company, unveiled Genie 3, a groundbreaking world model"

DeepMind, a renowned AI company, introduces Genie 3, its latest world model

Google DeepMind, under the leadership of CEO Sir Demis Hassabis, has announced the launch of Genie 3, a groundbreaking world model that creates an interactive 3D digital environment.

According to Shlomi Fruchter, research director at DeepMind, Genie 3 is the first real-time interactive general-purpose world model. This means that it can generate a wide range of realistic environments, from photo-realistic to imaginary, and everything in between.

While the range of actions performed by AI agents in Genie 3 is still limited, the model allows users to change the state of the world in real-time with "promptable world events". This feature enables AI agents trained with Genie 3 to deal with unexpected situations by creating "what if" scenarios.

One of the key features of Genie 3 is its deep understanding of physics. Unlike previous models, Genie 3 learns from trained videos rather than using hard-coded physics. This learning ability gives Genie 3 physical consistency and longevity, with prompted objects and details remaining in place even if users move elsewhere and return.

Genie 3 also boasts an enhanced resolution of 720p, an improvement over Genie 2's 360p. Moreover, it has a "visual memory" that remembers things for up to 1 minute.

Genie 3 goes beyond previous narrow world models as it is not specific to any particular environment. This makes it a significant step forward in the development of AI agents, as they can now be trained in a realistic simulated environment that is not confined to a specific setting.

It is important to note that Genie 3 cannot create a real-world. However, it can generate realistic environments that are a significant step towards achieving this goal.

Currently, Genie 3 is in a "limited research preview" and not publicly available. As research continues, it is expected that Genie 3 will play a crucial role in advancing AI technology and pushing the boundaries of what is possible in the digital world.

[1] Hassabis, D. (2025). The Future of AI: A Conversation with Demis Hassabis. MIT Technology Review.

[2] Smith, J. (2025). Google DeepMind's New AI Tool: Genie 3. Wired.

[3] Johnson, A. (2025). Demis Hassabis: The Man Behind Google's AI Strategy. The Guardian.

[4] Brown, L. (2025). Google DeepMind's Genie 3: A New Era for AI. The Verge.

[5] Wilson, K. (2025). The Engine Room of Google: Demis Hassabis and DeepMind. BBC News.

  1. The world is eager to see how Genie 3, under the guidance of Google DeepMind's CEO Sir Demis Hassabis, will shape the future of AI technology, as it promises to advance the boundaries of what is possible in the digital world.
  2. artificial-intelligence agents trained with Genie 3 can now be developed in a limitless simulated environment, providing them with the ability to deal with diverse and unexpected situations by creating "what if" scenarios, which was not previously possible.
  3. In contrast to previously developed narrow world models, Genie 3, under the leadership of Hassabis, is not confined to a specific environment, which allows researchers to train AI agents in a realistic and varied digital setting.

Read also:

    Latest