Google is building its own 'world modeling' AI team for games and robot training

Google DeepMind is establishing a new team focused on developing "world models," which are AI systems designed to simulate physical environments. This initiative, led by Tim Brooks, a former co-lead of OpenAI’s Sora project, aims to create AI models that can enhance video games, movies, and robot training scenarios. The team will work on scaling pretraining using video and multimodal data, which DeepMind believes is essential for achieving artificial general intelligence (AGI). The project is part of a broader competitive landscape in AI, where companies like OpenAI and Nvidia are also advancing their own technologies. DeepMind's world models will integrate with existing projects, including the Gemini AI models and the Veo video generator, to support applications in visual reasoning, planning for robots, and real-time interactive entertainment. As the race for AGI intensifies, Google’s focus on world modeling reflects its ambition to lead in this emerging field.

- Google DeepMind is forming a team to develop AI "world models" for simulating physical environments.

- The initiative is led by Tim Brooks, previously of OpenAI, and aims to enhance video games and robot training.

- DeepMind believes scaling pretraining on multimodal data is crucial for achieving artificial general intelligence (AGI).

- The project will complement existing Google AI efforts, including Gemini and Veo.

- The competition for AGI is intensifying, with other companies like OpenAI and Nvidia also making significant advancements.