Have you ever imagined stepping into a world crafted entirely by artificial intelligence, where you can interact with both virtual beings and the environment around you? Google DeepMind has just unveiled Genie 3, a groundbreaking AI model that transforms this dream into reality!

This cutting-edge “world” model is not just a step forward; it’s a giant leap into the future of digital environments. Genie 3 is designed to create immersive 3D spaces where humans and AI agents can move and interact in real time, paving the way for longer and more realistic experiences. Imagine returning to a virtual chalkboard or a beautifully painted wall, and finding every detail exactly as you left it – that’s the magic Genie 3 brings to the table.

World models like Genie are revolutionizing how digital environments are generated. Unlike video games that rely on painstakingly crafted 3D assets, Genie’s AI constructs everything based on simple prompts, allowing for a dynamic and engaging experience. This technology has vast applications, from educational tools to entertainment and even training autonomous robots or AI systems.

What’s more, Google has reported a dramatic increase in interaction time with Genie 3. While its predecessor, Genie 2, could only manage a mere 10-20 seconds of interaction, the new model boasts an impressive few minutes. This upgrade means that the system can retain visual details in memory for around a minute, allowing users to step away and return without a hitch.

With a new feature called “promptable world events,” users can manipulate conditions inside the world just by typing commands. Want to change the weather or add a new character? Just type it out! This interactivity adds a whole new dimension to the experience, making it not only more engaging but also more customizable.

While Genie 3 is a marvel, it’s important to note that it won’t be available to the general public just yet. DeepMind is initially rolling it out as a limited research preview, allowing a select group of academics and creators to explore its capabilities while studying potential risks and safety measures. For now, interaction features will be somewhat restricted, and any readable text in the generated worlds will primarily rely on what users include in their initial prompts. Google hints at expanding access in the future, promising even more exciting developments.

With the world models team spearheaded by a former co-lead of OpenAI’s Sora video generation project, the innovation behind Genie 3 is undeniable. The future of AI-generated environments is here, and it’s nothing short of extraordinary.