Could Google’s Veo 3 be the start of playable world models?
Demis Hassabis, CEO of Google’s AI research organization DeepMind, seemed to suggest on Tuesday evening that VEO 3, the latest video generating model from Google, could possibly be used for video games.
In response to a message about X that Google manages to “let me play a video game of my VEO 3 videos” and asked: “Playable World Models Wen?” Hagsabis responded“Now that wouldn’t be something.”
On Wednesday morning, Logan Kilpatrick, lead product for Google’s AI Studio and Gemini API, called in an answer: “🤐🤐🤐🤐”
Both messages from the Google managers are little more than playful suggestions, and a Google spokesperson told WAN that the company had nothing to share at the moment. But building playable world models is not outside the empire of possibilities for the technical giant.
World models differ from models for video store. The first simulates the dynamics of a real environment, so that agents can predict how the world will evolve in response to their actions. Video-gene models synthesize realistic video surveys.
Google has plans to convert its multimodal foundation model, Gemini 2.5 Pro, into a world model that simulates aspects of the human brain. In December, DeepMind Genie 2, a model that can generate an “endless” variety of playable worlds. The following month we reported that Google was a new team to work on AI models that can simulate the real world.
Others work on building world models-in particular AI-pioneer Fei-Fei Li. Li came from Stealth last year with World Labs, a startup that has built its own AI system that generates video game-like, 3D scenes from one image.
Veo 3, which is still in it Public PreviewCan make both video and audio to join Clips – everything from speech to soundtracks. Although VEO 3 creates realistic movements by simulating the physics of the real world, it is not really a world model yet. Instead, it can be used for cinematic stories in games, such as intermediate movies, trailers and narrative prototyping
The model is also still a generative model “passive output”, and it should switch (or a future VEO generation) to a simulator that is more active, more interactive and predictive.
But the real challenge with the production of video games is not only impressive visuals; It is real -time, consistent and controllable simulation. That is why it may be logical to see Google following a hybrid approach that uses VEO and Genie in the future, should it pursue a video game or playable world development.
Google could compete with MicrosoftScenario, catwalk, pika and, possiblyOpenAi’s video generating model Sora.
Given the planned movements of Google in the world models space and his reputation for the use of its deep pockets and distribution muscle to steam rivals, competitors in this space would be wise to keep an eye on.
