Google is forming a new team to develop AI models capable of simulating the physical world. This team, which will operate under Google DeepMind, will be led by Brooks, a former leader of the Sora video generation project at OpenAI. Brooks announced the hiring of new staff for a team aimed at creating “massive generative models that simulate the world” in a post on X (formerly Twitter).
The new Google team, building on the success of projects like Gemini (Google’s flagship model for image analysis and text generation), Video (Google’s video generation model), and Genie (a real-time 3D environment simulation model), will focus on solving “critical new problems” and elevating the models to the “highest computational levels.” The Genie model, introduced last December, can generate a wide range of 3D game worlds.
One of Google’s Main Goals in its New Project, Led by Brooks, is to Develop ‘Real-Time Interactive Generation Tools
A key goal of Google’s new project under Brooks’ leadership is the development of “real-time interactive generation tools” based on the created models and examining how they can be integrated with existing multimodal models, such as the previously mentioned Gemini. Google believes that developing these models using video and multimodal data is a crucial step toward achieving Artificial General Intelligence (AGI).
The hiring for this team comes at a time when several companies and startups, such as World Labs, Decart, and Odyssey, are also working in this area. They believe that real-world models could be used in the future to create interactive media like video games and films, as well as to run realistic simulations such as training environments for robots.
However, the Technologies Have Raised Concerns Among the Artistic Community and Creative Teams
The technologies mentioned have also faced concerns from the artistic community and creativity-based teams. Research shows that gaming studios like Activision Blizzard are using AI to reduce costs and increase efficiency, which has led to job losses for some individuals. A 2024 study by the Animation Guild reveals that by 2026, more than 100,000 jobs in the film, television, and animation industries in the United States will be impacted by AI.
Copyright Issues Are Also at Play, as Some Real-World Models Are Trained Using Video Game Clips
Copyright issues are also a concern, as some real-world models are being trained using video game clips, which could lead to legal action against the companies developing these models.
Google, as the owner of YouTube, claims that it has the necessary license to train its models using YouTube videos based on the platform’s terms of service, but has not provided details about the specific videos used.