Three months into the job, entrusted with responsibility.

Before Sora was released and crashed, its team leader Tim Brooks announced on social media that he jumped the car and went to rival Google DeepMind.

The news caused an immediate uproar. According to The Information, Tim Brooks left because of serious problems with Sora technology: Not only the generation speed is slow, but also the performance is difficult to compete with Luma, Stability, Runway and other rivals, not to mention the recent hot Pika and domestic video generation models.

While Sora is still behind closed doors, competitors have been gaining ground in the market and playing hot.

At the time, some industry insiders predicted that Tim Brooks was likely to play a big role in its video generation model Veo project after joining Google.

But what makes Sora amazing is not just the effect of video generation, but its ability to simulate real-world physics. The wind of world modeling is blowing in the AI circle: Google's GameNGen can directly generate games without a game engine, and the recently popular Oasis can directly create games that can be played using video models. Video generation technology is surging into the world of simulation games.

Tim Brooks' new task after joining Google DeepMind is in this direction. DeepMind co-founder Demis Hassabis revealed this in his welcome speech to bring "the long-unfulfilled dream of creating a world simulator to life";.

"DeepMind has ambitious plans to build large-scale generative models that can simulate the world," Tim Brooks announced at X on Monday. "I'm forming a new team for this mission."

The new team will work with Google's Gemini, Veo and Genie teams and build on their work to solve "critical new problems" and scale the model to "the highest level of computing," according to the job Posting linked to the post.

Gemini is Google's flagship family of AI models that can be used for tasks such as analyzing images and generating text. Veo is Google's own video generation model, and the latest version of Veo 2 can already generate beauty blogger unbox videos or vlogs in one step, which is realistic enough to be real.

Google Veo Vlog generated 2 video, photo source: https://x.com/jerrod_lew/status/1872673797939044487

As for Genie, it's Google's attempt at a world model - an artificial intelligence that can simulate games and 3D environments in real time. Just last month, Google announced their next-generation world model, Genie 2, which allows an interactive and playable 3D world to be created directly from a single image.

Genie 2 generates an interactive virtual world

"We believe that expanding AI training on video and multimodal data is a critical path to general AI," reads one job description. "World models will advance numerous areas such as visual reasoning and simulation, the planning of embodied agents, and real-time interactive entertainment."

Job requirements for Tim Brooks' team

From the above JD, the new Tim Brooks team will focus on developing "real-time interactive generation" tools on top of the models it builds and how to integrate its models with existing multimodal models, such as Gemini.

There are two main positions: research scientist and research engineer, with annual salaries ranging from $136,000 to $245,000.

The comments section was also very enthusiastic, and your resume may have already been submitted.

The World model is the focus of many startups and large tech companies, such as Li Feifei's World Labs, startups Decart and Odyssey. They believe that world models could one day be used to create media that can interact with viewers in real time, such as exclusive game stories that only belong to you. At the same time, the world model can also better simulate the world and solve the problem of lack of data in the robot training environment.

But our friends, who rely on creativity for a living, may not be so sanguine about models of the world.

Recently, an investigation by Wired magazine found that game studios like Activision Blizzard, which have laid off large numbers of employees, are using AI to cut corners, increase productivity, and make up for attitution. In 2024, a study commissioned by the Animators Guild, which represents Hollywood animators and cartoonists, estimated that more than 100,000 jobs in the film, television and animation industries in the United States will be impacted by AI by 2026.

But AI startups like Odyssey have made it clear that their goal is to work with creatives, not replace them. As for whether Google can use the world model to create a new era of AI and human creative symbiosis, let's wait and see.

Reference link:

https://techcrunch.com/2025/01/06/google-is-forming-a-new-team-to-build-ai-that-can-simulate-the-physical-world/.

https://x.com/_tim_brooks/status/1876327325916447140.