- Google Deepmind has just presented Genie 3, its latest world model
- Unlike Genie 2, this model allows real -time interaction and delivery everything in 720p
- This means that it can generate an environment, explore it and change it on the fly
Google’s AI World Model has just received a significant update, since the technology giant, specifically Google Deepmind, is presenting Genie 3. This is the latest model AI World World and kicks things to the high progress to allow the user to generate a 3D world with quality of 720p, explore and feed the new indications for interacting or changing the environment In real time.
It is really good, and I recommend that you watch the Deepmind announcement video that is embedded below. Genie 3 is also very different from, for example, the still impressive I see 3, since it offers an audio video that goes far beyond the limit of 8 seconds. Genie 3 offers multiple minutes of what Google calls the ‘interaction horizon’, which allows you to interact with the environment in real time and make adjustments as necessary.
It is as if AI and VR merge; It allows you to build a world of a warning, add new elements and explore everything. Genie 3 seems to be an improvement on Genie 2, which was introduced at the end of 2024. In a shared box in Google’s Deepmind publication, you can see Gamengen’s progression to Genie 2 A Genie 3, and even a comparison with I see.
Google also shared a series of demonstrations, including some that you can try in the blog post, and is giving us vibrations to choose your adventure. There are some different scenes that you can try in a snowy hill or even an objective that wants the AI to achieve inside a museum environment.
Attend
Google summarizes it as, “Genie 3 is our first world model to allow real -time interaction, while improving consistency and realism compared to the genie 2”. And while my mind, and my colleague Lance Ulanoff’s, went to interact in this environment in a virtual reality headset to explore a new place or even as a great blessing so that games developers try environments and maybe the characters, Google sees this as a surprise, a step towards Agi. That is artificial general intelligence, and the opinion here since Deepmind is that it can train several AI agents in an unlimited number of deeply immersive environments within the genie 3.
Another key improvement with Genie 3 is its ability to persist objects within the world; For example, we observe a set of arms and hands using a paint roller to apply blue paint to a wall. In the clip, we saw some wide blue paint stripes wounded on the wall, then we move away and look back to see the paint marks still at the right points.
It is ordered, and similar to some of the permanence of the objects that Apple establishes with visions 26, of course, that overlaps its real world environment, so it may not be so impressive.
Deepmind establishes the limitations of Genie 3, noting that in its current version, the world model cannot “simulate real world locations with perfect geographical precision” and that only admits a few minutes of interaction. Genie 3’s capacity minutes remain a significant leap on Genie 2, but does not allow hours of use.
Nor can you jump into the world of Genie 3 right now. It is available for a small set of testers. Google note that he expects Genie 3 to be available for other evaluators, but is discovering the best way to do it. It is not clear how the interface is seen to interact with Genie 3 at this stage, but from the shared demonstrations, it is quite clear that this is a convincing technology.
Whether Google restricts its use for the research and training of AI, or explores the generation of media, I have no doubt that we will see Genie 4 here in a short time … or at least in an expansion of the genie 3. For now, I will play again with I see 3.
