Google has announced the formation of a new team within its artificial intelligence research lab, DeepMind, to develop AI models capable of simulating the real world. The team will be led by Tim Brooks, a former leader of OpenAI’s video generation model, Sora, who joined DeepMind last October.
Brooks posted on X: “DeepMind has ambitious plans to build massive generative world models.” He added: “I am building a new team to achieve this goal.”
According to job listings shared by Brooks in his post, the team will collaborate with other internal Google groups, such as Gemini, Veo, and Genie, to develop models that tackle new challenges and expand AI capabilities. Gemini is Google’s flagship series of models specializing in image analysis and text generation, while Veo is Google’s video generation model, and Genie serves as a real-world simulation and 3D gaming model.
See also: American teen suicide case: mother sues chatbot company, with Google in the middle
Ambitious goals and the future of the virtual world
Through its new team, DeepMind aims to develop real-time interactive generation tools based on these models. The team will also collaborate with multimodal models, such as Gemini, to enhance integration between them.
One job posting indicated that “scaling AI training using multimodal and video data is a critical component to achieving artificial general intelligence.” AGI is defined by its ability to perform any task that humans can accomplish. The job description emphasized that real-world models will be used in numerous fields, such as visual reasoning, interactive robotics planning, and interactive entertainment.
Challenges and criticisms
Despite widespread optimism regarding the future of real-world models, significant debate surrounds their impact on creative jobs. According to an investigation by Wired magazine, gaming companies such as Activision Blizzard are using artificial intelligence to boost productivity, leading to numerous layoffs. Additionally, a 2024 study indicated that more than 100,000 jobs in film, television, and animation in the United States could be affected by AI by 2026.
Some startups, such as Odyssey, have affirmed their commitment to collaborating with creative professionals rather than replacing them. However, the question remains: will Google adopt a similar approach?
Furthermore, copyright remains an unresolved issue. Some models appear to be trained using clips from video games, which could expose their developers to lawsuits if the materials used are unlicensed.
Does Google meet the conditions?
For its part, Google—which owns YouTube—maintains that it has the right to train its models using YouTube videos in accordance with the platform’s terms of service. Nevertheless, the company has not yet specified the exact types of clips it uses for training.
