Is this the future of video games?

On 5 August, Google DeepMind, a subsidiary of Google tasked with artificial intelligence research, launched the third generation of their generative AI model, Genie.
 
Google DeepMind has previously released video generation models such as Genie2, or the Veo models (latest version published in May 2025), that are able to produce a few seconds worth of pre-set video material in 360p and 720p to 4K respectively.

In contrast, Genie3 is a general purpose world model able to, for the first time, generate interactive (playable) environments.

From a single text prompt and with a resolution of 720p, this model is able to generate multiple minutes worth of consistent gameplay. This means that the model "remembers" itself as well as its prompt, and is retroactively malleable.

© Screenshot: Google DeepMind Genie3

Google Deepmind has also announced that Genie3 is their newest advancement regarding their research into Artificial General Intelligence (AGI).

AGI is a field of AI research focused on creating software that can teach and continuously develop itself, autonomously expanding its capabilities beyond its initial parameters. Unlike current AI models, which are limited to specific tasks like generating text, images, or audio, AGI would be able to teach itself to operate across different domains without human intervention.

In the same vein, Genie3 is capable of improving on the initial result to a prompt while still in the very same setting.

This makes Google DeepMind's latest project a literal game changer, as users can now navigate personally stylised 3D worlds in real time.

While opening doors towards AGI, Genie3 also opens up the possibility of fully personalised gaming experiences, catered to the individual preferences of each player. Until further developments one can only imagine the possibilities that may arise, like combining the general purpose world model with virtual reality goggles, for example.

Impact outside of the digital sphere

While many avid gamers might rejoice at the thought of being able to design and play the game of their very own dreams, this might not be such good news for others. Indeed, game developers may now find themselves at risk of redundancy

Although the rushing development of AI in the past few years has proven a continuous threat to artists and art itself, discourse about Genie3's impact, similarly to other AI models, on employment prospects for artists in the gaming world will surely emerge.

To go one step further, numerous studies, like the one conducted by the Massachusetts Institute of Technology (MIT), have also shown that generative AI already negatively impacts the environment. Findings show that generating a single image with generative AI tool uses as much energy as charging your phone, or that training a single AI model can emit as much carbon as five cars in their lifetimes.

Further reports have also uncovered the massive amounts of water that data centers already require to cool AI servers. Specific cases in Arizona and Oregon in the U.S., where data centers are being built in areas not fit for industrial development, show that communities struggling most under water shortages are also most affected by this expansion. Indeed, more than 50 million gallons of water are needed annually for the maintenance of servers and data centers.

These findings effectively put into perspective the drastic toll that AI's upkeep takes on our planet and the populations most directly affected by it.

At the moment, the impact Genie3 and its fellow state-of-the-art AI models will have, remains to be studied. However, considering the exorbitant amount of resources their sophisticated models require, it poses the question on how Google DeepMind plans to continue its research into a "safe and responsible" AI usage.

Until further notice, Genie3 remains unavailable to the general public, but Google DeepMind promises that exciting developments are on the horizon.