Demis Hassabis, Managing Director of Google Deepmind’s AI Research Organization, appeared to suggest on Tuesday night that the latest VEO 3 model, the latest model to create Google videos, could possibly be used for video games.
Responding to a post on X Bseeeching Google to “allow to play a video game of my VEO 3 already”, and ask, “Playable World Models wen?” Dug replied“Now it wouldn’t be something.”
On Wednesday morning, Logan Kilpatrick, a lead product for AI Studio and Google’s Gemini API, fell with answer: “🤐🤐🤐🤐”
Both publications by Google executives are a little more than playful suggestions, and a Google spokesman told TechCrunch that the company had nothing to share right now. But building worldwide models is not out of the realm of the possibilities for the technological giant.
World models are different from video clips. The former simulates the dynamics of a real world environment, which allows agents to predict how people will evolve in response to their actions. Video-Gen models make up realistic video sequences.
Google plans to convert the Multimodal Foundation, the Gemini 2.5 Pro, into a global model that simulates aspects of the human brain. In December, Deepmind revealed Genie 2, a model that can create an “endless” variety of players that can be reproduced. Next month, we mentioned that Google formed a new team to work on AI models that can simulate the real world.
Others work to build global models-mainly, pioneer Fei-fei li. Li came out of Stealth last year with World Labs, a start that has created his own AI system that creates 3D -like scenes that look like a picture of a picture.
Veo 3, still in public previewIt can create videos as well as sound to proceed with the clips – anything from speech to soundtracks. While VEO 3 creates realistic moves simulating the real world’s physics, it is not yet a model of the world. Instead, it could be used for cinematic narrative in games such as cuts, trailers and narrative original
The model is still a “passive output” genetic model and computer science (or a future generation VEO) should be shifted to a simulator who is more active, interactive and predictor.
But the real challenge with producing video games is not just impressive graphics. It is in real time, consistent and controlled simulation. That is why it can make sense to see Google adopt a hybrid approach that exploits VEO and Genie in the future if a video game or global development follow.
Google could be found competing MicrosoftScenario, corridor, pika, and, finallyModel creating videos of Openai Sora.
Given Google’s scheduled movements in World Models Space and its reputation for the use of deep pockets and distribution muscles to Steamroll opponents, competitors in this area would be wise to watch carefully.
