News

They manage to play CS:GO on a map generated with generative AI

One of the most cutting-edge technologies that we can currently find is artificial intelligence, and it is everywhere, it is only a matter of time before we begin to see how it is implemented in video games, and this time, we can see a very current game generated entirely by AI. This time we can see how one of the most popular eSports titles, CS:GO runs directly on a AI generative, powered only by an RTX 3090, which manages to generate this title in real time, although with terrible performance.

The arrival of artificial intelligence has managed to change a lot how we see some things, and it has managed to become one of the most important, and in some cases dangerous, technologies that we can find on the internet and which practically anyone can access. And as we can imagine, it is advancing faster and faster, since many users and developers are collaborating to try to create a intelligence artificial that allows you to do anything, and this time we can find one that can be downloaded for free on Github that allows you to generate the video game CS:GO in real time.

A completely AI generated game, what could go wrong?

It is normal for us to wonder what the current limits of generative artificial intelligence are, and on many occasions we can come to think that the models that seek to offer these capabilities have remained stuck in generating images or videos. But this is very far from reality, and there are some that can run games, that is, they allow you to interact directly with the computer. AI while it is creating the stage, and the most current example that we can find is how the DIAMOND model has been applied to generate the mythical CS:GO.

Although the performance is not what we might expect since it only works at 10 FPS, the game runs entirely on a neural network powered solely by an RTX 3090, which has quite a lot of power, but even in these cases It shows how demanding AIs are. Eloi Alonso, one of the collaborators working on this model, has presented a series of videos of how the title works, all of them showing real people playing one of the most important maps in this game, Dust 2.

Ever wanted to play Counter-Strike in a neural network?

These videos show people playing (with keyboard & mouse) in ๐Ÿ’Ž DIAMOND’s diffusion world model, trained to simulate the game Counter-Strike: Global Offensive.

๐Ÿ’ป Download and play it yourself โ†’ https://t.co/vLmGsPlaJp

๐Ÿงต https://t.co/8MsXbOppQK

October 11, 2024 โ€ข 20:13

Compared to the original title, there are a series of moments in which it is practically impossible to distinguish it (if we do not take into account that it goes at 10 FPS), and it offers an almost perfect copy of what the game is, but generating everything at the moment as the user performs actions. This causes quite a few bugs and problems, of which Eloi has shown some of the funniest ones they have had in a thread on Twitter (X), such as hallucinations caused by moving the mouse too quickly, the character teleporting when hitting a wall or weapons transforming when the lighting changes.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *