In news your 8-year-old kid probably knew about weeks ago somehow, a new Minecraft rip-off is available to play online. In a surprising twist on the genre, every frame in this one is entirely generated by artificial intelligence (AI).
The game, Oasis, lets players explore a 3D world filled with square blocks, mine resources, and craft items, just like the ridiculously popular Minecraft. It’s a little surreal to play, with distant landscapes morphing into other shapes and sizes as you approach.
But underneath the hood, the game is quite different from any other you have played. It lets you choose from a host of starting environments, as well as the option to upload your own image to be used as a starting scene. However, the player’s actions and movements soon change the environment around you.
“You’re about to enter a first-of-its-kind video model, a game engine trained by millions of gameplay hours,” the game’s creators, Decart and Etched, explain as you enter. “Every step you take will shape the environment around you in real time.”
“The model learned to allow users to move around, jump, pick up items, break blocks, and more, all by watching gameplay directly,” the team explained in a statement. “We view Oasis as the first step in our research towards foundational models that simulate more complex interactive worlds, thereby replacing the classic game engine for a future driven by AI.”
There are all sorts of weird quirks to the game, a result of everything being generated frame-by-frame, and being trained on real gameplay. Items will randomly pop up in your inventory, as the game tries to match what Minecraft players would have in their inventory in similar environments. The landscape warps and changes as you look away from it, with the entire world feeling like it is simulating what it’s like not to have object permanence yet.
For a game concept that revolves around building permanent structures, this isn’t ideal, and we doubt anyone would play it for fun rather than out of curiosity. AI is not replacing video game developers any time soon.
However, as a curiosity, it is pretty interesting.
“In essence, diffusion models learn to reverse the iterative process which adds Gaussian noise to the input and thereby enable generating new samples given noise,” the team explained. “This approach can be extended towards video generation by adding additional temporal layers to the model architecture that receive a context consisting of the previous frames that were generated (e.g., in an autoregressive fashion).”
The team has high hopes that AI generation could “revolutionize” games, enabling more interaction between gamer and the game they are playing.
“Simply imagine a world where this integration is so tight that foundation models may even augment modern entertainment platforms by generating content on the fly according to the user preferences,” the team says, “or perhaps a gaming experience that provides new possibilities for the user interaction such as textual and audio prompts guiding the gameplay (e.g., ‘imagine that there is a pink elephant chasing me down’).”
Whether gamers want this or well-made, human-designed video games, remains to be seen. You can play the game for yourself here, though be warned, you may have to join the queue.
