Liputan6.com, Jakarta Google DeepMind, a leading artificial intelligence research division, officially announced the launch of Genie 3 on August 5, 2025.
This world model marks a significant leap forward in AI technology, offering the ability to generate a diverse range of interactive environments like never before.
Genie 3 is the latest iteration of Google's AI world model, designed to create 3D spaces.
Within these virtual spaces, both humans and AI agents can move and interact seamlessly in real time, unlocking new potential for a wide range of applications.
Genie 3: What's New? Here Are Details and Abilities
Genie 3 demonstrates remarkable capabilities in generating dynamic virtual worlds from simple text commands.
The created environments can be navigated in real-time at 24 frames per second at 720p resolution, maintaining impressive consistency for several minutes.
A significant improvement over previous versions, Genie 3 is the first world model from Google DeepMind to enable direct, real-time interaction.
It also offers substantial improvements in consistency and realism compared to Genie 2.
One of Genie 3's standout features is its innovative "world memory," which ensures the environment remains consistent even when the user performs actions such as painting on a wall; these actions persist upon return.
The "Promptable World Events" feature allows users to dynamically change the scene through natural language commands, such as adding a herd of reindeer to a ski slope.
Genie 3 Applications and Potential
The Genie 3 world model has significant potential to revolutionize various sectors, particularly in AI training and robotics.
Google DeepMind states that this model can be used to train robots and autonomous vehicles in realistic recreations of environments, such as warehouses, allowing AI agents to train in a rich and unlimited curriculum of simulated environments.
Furthermore, Google DeepMind views world models like Genie 3 as a crucial step towards achieving Artificial General Intelligence (AGI).
The model's ability to create and simulate complex worlds is a crucial foundation for the development of more intelligent and adaptive AI.
Beyond the technical realm, Genie 3 also has the potential to enhance user experience and entertainment.
This model can help humans experience various simulations for training or exploration, replicating experiences like skiing or hiking around a mountain lake.
Potential uses for Genie 3 also include the development of next-generation gaming and immersive forms of interactive entertainment.
Genie 3 Limitations and Availability
Despite its advanced capabilities, Genie 3 still has several technical limitations that need to be addressed.
Handling physics in simulation remains challenging, with clear failure cases in intuitive physics experiments, such as towers of blocks.
Social and multi-agent interactions are also still difficult to handle, and one-on-one combat games don't work well.
Furthermore, the model still struggles to follow long instructions and simple combinatorial game logic, such as collecting points or keys. The available action space is also limited.
While promising, Genie 3 is still far from being a true game engine, but it does provide a clear glimpse into the future of this technology.
Regarding public availability, Google stated that Genie 3 is not yet ready for full release and did not provide a specific launch date.