Almost at the same time that Openai has presented its free and open source models GPT-Oss, Google DeepMind Genie 3 premieres, a generative AI of dynamic worlds that exceeds its predecessor, as well as Gamengen and I see.
Genie 3 It differs from a video generative AI such as Sora or V1, for example, in which it does not generate video that you only look at. Create authentic interactive worlds With real image For which you can walk, moving with the cursors of the keyboard as in a video game, or by by building with the objects of that world.
It is easy to imagine the immense possibilities of this artificial intelligence in aspects such as games, interactive education, virtual tourism, etc.
This is how Genie 3 works, Google Deepmind’s new AI
As you can see in the previous video, Genie 3 can create completely real worldsfrom one where you are riding on horseback, to another where you visit an Egyptian temple, or walks through Venice.
We move inside the video in real time. As the trailer looks, it is possible to do things like painting a wall, or writing on a board.
The big problem that these are what is called «Memory of the world«. They are able to create real -time scenarios from a prompt, but do not remember the changes that have occurred after generating it.
For example, imagine that you have created a world that simulates a school of the beginning of the century. If you write a word with a chalk on the board, tours, sight, and look again, the word is gone.
While other similar ones only «remember» what you do for 10 or 20 seconds, Genie 3 Google keeps him more than a minute. It is an important achievement.
In front of its predecessor, which created these scenarios at a 360p resolution, Genie 3 does 720p and 24 fps.
Applications of dynamic worlds created with AI
Another important improvement in this new language model is what Google calls «events through Prompts.» Once the world is generated, while you explore it, It is possible to add elements in real timeas new characters, objects, actions, etc.
These types of AI models are still little more than curiosities, because they can only maintain those scenarios one or two minutes. But it is easy to see the immense applications they will provide in a few years.
From video games created in real time to interactive teaching projects, virtual tourism, training in situations of emegence, robots training, and much more.
Unfortunately, for the moment Genie 3 Google Deepmind is only available for researchers and specialists that Google has invited to try it. But they are looking for how more people can try it.
Know How we work in NoticiasVE.
Tags: Artificial intelligence, video, video games, viral, curiosities