Google wants to create AI capable of “simulating the real world”
At the end of last year, Google revealed Genie 2its artificial intelligence capable of simulating fully playable virtual worlds. Today, the firm intends to take the next step: leaving the virtual world to simulate a physical world. This is Tim Brooks, one of the former chefs behind the creation of Sora, OpenAI's video generation modelwho announces it on X (formerly Twitter). After leaving OpenAI for Google, Tim Brooks announced: “DeepMind has the ambitious idea of creating generative artificial intelligence models capable of simulating the real world: I am recruiting a new team for this mission“.
Many positions to fill
According to the job descriptions posted online by Brooks, the team will aim to collaborate and work with those responsible for Gemini, Veo and Genie, then tackle entirely new problems by creating artificial intelligence models of a completely new power. “We believe that the scaling up of [l’entraînement de l’IA] on video and multimodal data is a crucial step towards artificial general intelligence” can we read on the ad.
“World models will power many areas, such as visual reasoning and simulation, planning for embodied agents, and real-time interactive entertainment“, adds the official announcement of the search for collaborators at Google. Note that the American giant is not the first to embark on such an enterprise: firms like World Labs, Decart or Odyssey all consider that these models of generated worlds by AI could offer new multimedia experiences (close to video games or cinema) capable of running precise and realistic simulations of physical worlds.
