Google has unveiled plans to create a new team dedicated to developing advanced artificial intelligence models capable of simulating the physical world. The initiative will be led by Tim Brooks, former co-lead on OpenAI’s video generation model, Sora, who joined Google DeepMind’s AI research lab in October.

A Bold Step Toward AI World Models

In a post shared on X, Brooks outlined the ambitious vision for the team:

“DeepMind has ambitious plans to create massive generative models that simulate the world. I’m hiring for a new team with this mission.”

The team will focus on building AI models that simulate real-time environments and solve complex challenges at unprecedented computational scales. This effort will build upon the work of existing teams within Google, including Gemini, Veo, and Genie.

Expansion of Existing AI Teams

• Gemini: Google’s flagship AI model, capable of handling tasks like image analysis and text generation.
• Veo: A team focusing on video generation.
• Genie: A key component of the new team’s effort, Genie aims to create world models that can simulate 3D environments in real-time and interactive games.

The new team will integrate these capabilities and develop tools for real-time interactive generation, laying the groundwork for advancements in fields like visual reasoning, real-time simulation, autonomous agent planning, and interactive entertainment.

Advancing AI and Multimodal Data

Brooks noted that scaling AI training with video and multimodal data is essential for achieving artificial general intelligence (AGI), which refers to AI that can perform any task a human is capable of. The goal of this initiative is to create models that can power a variety of industries, including entertainment, robotics training, and autonomous systems.

Rising Interest in World Models

World models are rapidly gaining attention from tech companies and startups, including AI researcher Fei-Fei Lee’s World Labs, Decart, and Odyssey. These models are seen as transformative for fields like interactive media, video games, and simulation training for robots.

Some startups, such as Odyssey, have emphasized collaboration with creative professionals, rather than seeking to replace them. It remains to be seen whether Google will follow this collaborative approach.

Challenges and Copyright Concerns

A key issue in developing world models is the use of training data, particularly regarding copyright concerns. As world models often require large datasets, questions arise about whether these models may use unlicensed material, such as video game content. Google has indicated that it has the rights to use YouTube videos for training but has not yet disclosed specific details about the videos being sourced.

With this initiative, Google aims to push the boundaries of AI technology, potentially reshaping industries and accelerating the path toward AGI.