- Google Deepmind has just unveiled Genie 3, its latest world model
- Unlike Genie 2, this model allows real -time interaction and provides all this in 720p
- This means that you can generate an environment, explore it and change it on the fly
Google’s AI World model has just received a significant upgrade, because the technology giant, in particular Google Deepmind, introduces Genie 3. It is the latest world model of AI, and it launches things in the proverbial higher speed by allowing the user to generate a 3D world with real quality.
It’s really good, and I strongly recommend that you watch DeepMind’s ad video which is integrated below. Genie 3 is also very different from, let’s say, the Veo 3 still impressive, because it offers a video with the audio which goes far beyond the limit of 8 seconds. Genie 3 offers several minutes of what Google calls “the interaction horizon”, allowing you to interact with the environment in real time and make adjustments if necessary.
It is a bit as if AI and VR merge; It allows you to build a world on an prompt, add new articles and explore all of this. GENIE 3 seems to be an improvement compared to GENIE 2, which was introduced at the end of 2024. In a graphic shared in the post Deepmind of Google, you can see the progression of Gengengen to Genie 2 to Genie 3, and even a comparison with Veo.
Google has also shared a number of demos, some of which you can try in the blog post, and this gives us vibrations to choose your adventure. There are a few different scenes that you can try on a snowy hill or even a goal you want AI to reach a museum environment.
To watch
Google sums up as, “Genie 3 is our first world model to allow real -time interaction, while improving consistency and realism compared to GENIE 2.” And while my mind, and my colleague launches Ulanoff, went to interact in this environment in a VR helmet to explore a new place or even as a big boon for game developers to test the environments and perhaps even the characters, Google considers it – not a surprise – a step towards act. It is the general artificial intelligence, and the point of view here of Deepmind is that it can train various agents of AI in an unlimited number of deeply immersive environments within genius 3.
Another key improvement with Genie 3 is his ability to persist objects in the world – for example, we have observed a set of weapons and hands using a paint roll to apply blue paint on a wall. In the clip, we saw some large stripes of blue paint rolled on the wall, then we diverted ourselves and looked back to see the paint marks always in the right places.
It is neat and similar to part of the permanence of objects that Apple is achieving with Visionos 26 – of course, which covers your real environment, so perhaps not as impressive.
Deepmind exposes the limits of Genie 3, noting that in its current version, the world model cannot “simulate real world locations with perfect geographic precision” and that it only takes care of a few minutes of interaction. Genie 3’s capacity minutes are always a significant jump compared to Genie 2, but this does not allow hours of use.
You cannot jump into the world of Genie 3 either at the moment. It is available for a small set of testers. Google notes that he hopes to make Genie 3 available to other testers, but that is to find the best way to do so. We do not know what the interface looks like to interact with Genie 3 at this stage, but according to shared demos, it is quite clear that it is a convincing technology.
Whether Google restricts its use in research and training of AI, or that it explores the generation of media, I have no doubt of genius 4 here in a short time … or at least an expansion of Genie 3. For the moment, I will come back to play with Veo 3.
