Google is forming a brand new team to work on AI models that may simulate the physical world.
Tim Brooks – considered one of the co-leaders of OpenAI's video generator Sora, who moved to Google's AI research lab, Google DeepMind, in October – will lead the brand new team, he said announced in a post on X. It will likely be a part of Google DeepMind.
“DeepMind has ambitious plans to create massive generative models that simulate the world,” Brooks wrote Monday morning. “With this mission I’m recruiting a brand new team.”
Accordingly Work Listings Brooks noted in his post that the brand new modeling team will work with and construct on Google's Gemini, Veo and Genie teams to handle “critical recent problems” and scale models “to the best levels of computing power.” Gemini is Google's flagship line of AI models for tasks like analyzing images and generating text, while Veo is Google's own video generation model.
Genie is Google's tackle a world model – an AI that may simulate games and 3D environments in real time. Google's latest Genie model, unveiled in December, can create an enormous number of playable 3D worlds.
“We consider that scaling (AI training) on video and multimodal data is the critical path to artificial general intelligence,” considered one of the job descriptions says. Artificial general intelligence, or AGI, generally refers to AI that may perform any task that a human can. “World models will power quite a few areas, resembling visual pondering and simulation, planning for embodied agents, and real-time interactive entertainment.”
According to the outline, Brooks' recent team will seek to develop “real-time interactive generation” tools based on the models they create and explore how their models integrate with existing multimodal models resembling Gemini.
Various startups and large tech corporations are chasing world models, including influential AI researcher Fei-Fei Lee's World Labs, Israeli upstart Decart and Odyssey. They consider that world models could at some point be used to create interactive media resembling video games and movies, and to conduct realistic simulations resembling training environments for robots.
Work with Tim and the Deepmind team on huge world simulation models 🙂
On the critical path to AGI. https://t.co/4Zuju5eMHb
– Logan Kilpatrick (@OfficialLoganK) January 6, 2025
However, creatives have mixed feelings in regards to the technology.
A youngest Wired research found that game studios like Activision Blizzard, which have laid off large numbers of employees, are using AI to chop corners, increase productivity and compensate for turnover. And a 2024 study Commissioned by the Animation Guild, a union representing Hollywood animators and cartoonists, it was estimated that over 100,000 U.S. film, television and animation jobs will likely be impacted by AI by 2026.
Some startups within the emerging world of modeling, like Odyssey, have committed to working with creative professionals — not replacing them. We'll need to see if Google follows suit.
There can also be the unresolved query of copyright. Some world models look like trained on clips of video game playthroughs, which could make the businesses that develop these models the goal of lawsuits if the videos were unlicensed.
Google, which owns YouTube, claims that it’s allowed to coach its models on YouTube videos under the platform's terms of service. However, the corporate has not said what specific videos it’s sourcing for training.
