Home » Here comes Genie, the AI ​​that creates video games starting from an image

Here comes Genie, the AI ​​that creates video games starting from an image

by admin
Here comes Genie, the AI ​​that creates video games starting from an image

Listen to the audio version of the article

Google’s new project is another significant step in the revolution of video game entertainment thanks to increasingly complex generative models
After producing texts, photos and videos, video games also join the list of things that an AI can generate from a text prompt. After Sora, OpenAI’s artificial intelligence, which was able to create a video game similar to the famous Minecraft, including a physics simulation and the possibility of controlling the player’s character, Google has also decided to show the first fruits of experimentation in the gaming sector, following the already announced Gemini model. This is Genie, a new model of generative artificial intelligence that can create two-dimensional video games starting from a textual “prompt”, but also a photo or a simple sketch on a sheet of paper.

How Genie works and why it is “brilliant”

Based on the instructions provided to Genie, the artificial intelligence can create a “playable image” with different styles, even on a realistic background, distinguishing static and dynamic elements, creating a user interface with which to interact and managing the physics of movement in a credible way. Not only that, Google’s new Ai also offers the possibility of feeding digital artwork or the result of a two-dimensional world created by other generative Ai, and letting Genie create the video game based on the instructions and materials provided. In practice, Genie not only creates graphic assets from the input provided by the user, but also compiles the necessary code, although obviously this can be very limited and also contain possible errors. A short demonstration of how it works was shared on X by Tim Rocktäschel, leader of the Genie project.

See also  How AI works to avoid power outages from heat waves


Google DeepMind researchers trained the model on more than 200,000 hours of publicly available 2D platformer video game footage to teach the AI ​​the basics of design for the 2D platformer genre. The learning by the AI ​​occurred unsupervised and without the content being manually labeled by the developers. This was a challenge for developers because videos on the Internet typically don’t have labels relating to the action being performed. “Surprisingly, Genie not only learns which parts of an observation are generally controllable, but also infers different actions that are consistent in the generated environments,” the researchers explain on the page dedicated to the Genie project.

Robotics and video games the future of Genie

Although still an experiment (available in beta phase for research), the potential shown by this new AI, as underlined by the developers, goes beyond the generation of simple video games. Thanks to Genie’s ability to understand real-world physics, Google’s model has also been successfully used to train a robotic arm to move and interact more efficiently. A solution that in the future could lead Genie to be used in robotics, but also in the simulation of virtual environments for scientific and medical research. In its use in the gaming sector, Google’s new Ai will certainly be able to evolve and improve, continuing its ” training” also on 3D models of modern video games. The ambition is to become a concrete tool in the hands of independent developers and startups, who in this way will be able to bring their ideas to life by experimenting with new approaches in the creation process, saving time, human resources and therefore money. A question remains open and represents the other side of the generative AI coin. If machines like Genie become increasingly capable of creating video games, even complex ones, the impact on the work of programmers and game designers will certainly not be negligible.

See also  all May 2023 releases...

You may also like

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.

This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More

Privacy & Cookies Policy