Google DeepMind is pushing the boundaries of artificial intelligence with a new initiative that aims to create virtual copies of our physical world. This development could change how we train robots, test self-driving cars, and create interactive entertainment.
On January 6, Google DeepMind announced plans to build “massive” generative models designed to simulate real-world environments. Tim Brooks, who recently joined Google DeepMind from OpenAI in October, will lead this ambitious project.
“DeepMind has ambitious plans to make massive generative models that simulate the world,” Brooks shared in a public announcement. His team will work closely with Google’s existing AI projects, including Gemini (their main AI model), Veo (for video generation), and Genie (for world simulation).
These world models are computational frameworks that help AI systems understand and simulate real or virtual environments. Autonomous vehicles currently use these world models to simulate traffic and road conditions, and they can help train generalist AI robots in different environments.
Google has already made progress in this field with Genie 2, their latest simulation tool. This system can turn simple text and images into 3D worlds that respond to user actions. When using Genie 2, the worlds it creates can last up to a minute, with most in the 10- to 20-second range. The system displays object interactions, complex character animation, physics (such as gravity and splashing water effects), and behavior modeling of other agents.
The impact of this technology could reach far beyond gaming and entertainment. A common challenge in the field is finding rich, diverse, and safe training environments for embodied AI systems.
However, this advancement raises concerns in the creative industry. A study by the Animation Guild predicts that AI could affect over 100,000 U.S. film, television, and animation jobs by 2026. Some companies are trying to find middle ground – like 3D startup Odyssey, which plans to work with creators rather than replace them.
Google isn’t alone in this pursuit. World Labs, a startup led by Stanford AI expert Fei Fei Li, recently raised $230 million to develop similar technology. Their impressive supporter list includes AI pioneer Geoffrey Hinton, Salesforce CEO Marc Benioff, and former Google Chairman Eric Schmidt.
These developments are part of what many see as steps toward artificial general intelligence (AGI) – AI systems that could match human capabilities across different tasks. The job posting explicitly states that they believe scaling AI training on video and multimodal data is on the critical path to AGI.
The announcement adds to Google DeepMind’s recent achievements, including AlphaFold2, which helped solve a 50-year-old protein prediction challenge and earned recognition from the Nobel committee for CEO Demis Hassabis and John M. Jumper.
This technology development suggests potential changes in how we test and develop new technologies, create training environments, and experience interactive entertainment. However, it also highlights the need for careful consideration of how AI development affects jobs and industries.
