Google is putting together a new group to work on AI models that can mimic the real world.
In a post on X, Tim Brooks said that he would be leading the new team. Brooks was one of the co-leads on OpenAI’s video creator, Sora, but left to work at Google’s AI research lab, Google DeepMind, in October. It will be a part of Google DeepMind.
Brooks wrote Monday morning, “DeepMind has big plans to make huge generative models that simulate the world.” “I’m hiring for a new team with this mission.”
Brooks’s post led to job postings that said the new modeling team would work with Google’s Gemini, Veo, and Genie teams and build on their work to solve “critical new problems” and take models “to the highest levels of compute.” Google’s most popular AI models are Gemini, which can do things like analyze pictures and write text, and Veo is Google’s own model for making videos.
For Genie, Google has made an AI that can mimic games and 3D environments in real time. It’s their take on a world model. Google’s newest Genie model, which was shown off in December, can make a huge number of 3D scenes that can be played in.
One of the job descriptions says, “We believe that scaling [AI training] on video and multimodal data is on the critical path to artificial general intelligence.” AGI stands for “artificial general intelligence,” which means AI that can do any job a person can. “World models will power numerous domains, such as visual reasoning and simulation, planning for embodied agents, and real-time interactive entertainment.”
The job description says that Brooks’s new team will work on adding “real-time interactive generation” tools to the models they create and will look into how to connect their models to multimodal models like Gemini that are already out there.
World Labs, an important AI researcher Fei-Fei Lee, an Israeli startup called Decart, and Odyssey are just a few of the small and large tech businesses that are looking for world models. Some people think that world models could one day be used to make interactive media like movies and video games, as well as run accurate simulations like robot training grounds.
But artists and writers have mixed views about the tech.
A recent study by Wired found that game studios like Activision Blizzard, which has fired a lot of employees, are using AI to save money, get more work done, and make up for the lost workers. And in 2024, the Animation Guild, a group for Hollywood animators and cartoonists, paid for a study that said AI would take away more than 100,000 film, TV, and animation jobs in the U.S. by 2026.
Odyssey is one of the new companies in the world modeling space that has promised to work with creative workers instead of replacing them. We have to wait and see if Google does the same.
Aside from that, there’s still the issue of copyright. Some world models seem to have been trained on video clips of people playing video games. If the videos used were not allowed, the companies that made those models could be sued.
Also Read: Google Releases Its Own Ai Model That Can “reason.”
In line with YouTube’s terms of service, Google, which owns the site, says it is allowed to train its models on YouTube movies. But the business hasn’t said which films it is using for training.
What do you say about this story? Visit Parhlo World For more.
