Quake II Returns with AI
- Microsoft researchers have unveiled an artificial intelligence model that allows users to interact with the Muse video game environment through keyboard and controller inputs.
- The AI generative model, designed for video games and dubbed Muse, was initially presented in Nature magazine on Feb.
- To showcase the model's capabilities, researchers trained it on a level from Quake II, a game Microsoft acquired through its purchase of Zenimax.
Microsoft’s AI Model Lets Users Play Inside Quake II, But Faces Challenges
Table of Contents
- Microsoft’s AI Model Lets Users Play Inside Quake II, But Faces Challenges
- Microsoft’s AI “Quake II” Experiment: Your Questions Answered
- What is Microsoft’s AI Model for Quake II?
- How Does the Muse AI Model Work?
- What Are the Limitations of This AI-Generated Quake II Demo?
- Why is Austin Walker Critical of the AI Demo?
- What Does This Mean for the Future of Game Development?
- How Does This AI Experiment compare to the Original Quake II?
- Why is This AI Experiment Still Interesting, Despite Its flaws?
- What is Quake, Exactly?
Microsoft researchers have unveiled an artificial intelligence model that allows users to interact with the Muse video game environment through keyboard and controller inputs. According to a recent blog post, players can now see the immediate effects of their actions, effectively enabling them to “play inside the model.”
The AI generative model, designed for video games and dubbed Muse, was initially presented in Nature magazine on Feb. 19. It is indeed designed to generate game visuals, character actions, or both.
To showcase the model’s capabilities, researchers trained it on a level from Quake II, a game Microsoft acquired through its purchase of Zenimax. TechCrunch reported on the release of the AI-generated Quake II demo, while also noting its limitations.
“We trained our model on the foundations of the well-known Quake II game,” the researchers stated. “To our initial joy, we were able to play inside the world that simulated the model.”
They added, “We could walk, move the camera, jump, crouch, shoot, and even throw the barrels into the air similarly to the original game. All this can now try.”
Playable Level exhibits Shortcomings
The creators acknowledged that the AI-generated demo has limitations, including blurred enemies, inaccurate damage and health counters, and object stability issues.
“It is an experimentation with our generative model. So we teach it the approximation of the real environment on whose data it was trained,” the researchers explained.
“Our intention is not to fully copy the real experience of playing the original game Quake II,” they continued. “It should be a research exploration of what we are able to create using current ML approaches. Imagine it as a model playing unlike playing the game.”
despite the imperfections, the experiment is intended as a source of entertainment, allowing users to engage enemies or teleport. It is not intended to be a polished, classic game experience.
AI Game Can Trap Players
Some users who have tested the demo have reported issues. Austin Walker,a writer and game designer,posted a video showing himself frequently trapped in a dark room.
“Microsoft is a real perpetrator,” Walker wrote on social media. “But anyone who writes a new post on the AI game demo of the Quake Copilot without a description of how it actually works (or fails) does the company unpaid PR work.”
Walker also responded to a recent statement by Microsoft Gaming CEO Phil Spencer regarding the potential of AI models to assist in porting games to different platforms.
“When I read such a statement, blood is cooked in me.I feel a fundamental misunderstanding not only of this technology, but also how games work,” Walker explained.
“The internal functioning of games like Quake (code, design, 3D art, sound) creates specific cases of gaming, including surprising marginal cases. This is a large part of what makes games good. If you are unable to rebuild key internal mechanisms,you lose access to these unpredictable marginal cases,” the designer added.
The Muse generative model and its experimental game require further progress to ensure proper functionality and prevent players from being consistently trapped in darkness.
Microsoft’s AI “Quake II” Experiment: Your Questions Answered
Microsoft is making waves with a new AI model that allows users to interact with video game environments – specifically, a demo within teh classic game Quake II. But how does it work, what are the issues, and what does it mean for the future of gaming? Let’s dive in.
What is Microsoft’s AI Model for Quake II?
Microsoft researchers have developed an AI model called “Muse” that allows players to interact with a simulated Quake II surroundings using keyboard and controller inputs.Essentially, you can “play inside the model,” seeing the effects of your actions in real-time within the game.
How Does the Muse AI Model Work?
The Muse AI model is a generative model designed to create in-game visuals and character actions. Researchers trained it on a level from Quake II. It can generate game assets and behaviors, enabling players to walk, move the camera, jump, crouch, shoot, and interact with the environment, mimicking the functions of the original game.
What Are the Limitations of This AI-Generated Quake II Demo?
the demo isn’t perfect. The creators themselves acknowledge several shortcomings. Here’s a breakdown of the key issues:
- Blurred Enemies: Enemies are not rendered with the same level of clarity as in the original.
- Inaccurate health and Damage Counters: The UI elements might not consistently display correct data.
- Object Stability Issues: Objects within the game, such as barrels, may behave inconsistently.They might not always interact with the environment as was to be expected.
these imperfections highlight that the demo is an experiment exploring what’s possible with current machine learning approaches, rather than aiming to perfectly replicate the original Quake II experience.
Why is Austin Walker Critical of the AI Demo?
Game designer and writer Austin Walker has raised concerns about the AI demo, particularly in a social media posts. He and other users have reported issues with the game demo, including a tendency for players to get trapped in dark rooms. Walker’s concerns stem from a deeper understanding of how games function internally. The AI’s potential failure to recreate these nuances could lead to an experience that lacks the emergent gameplay and unpredictability that make games enjoyable.
What Does This Mean for the Future of Game Development?
This technology has the potential to assist with game porting and level design. However, Austin Walker and others have highlighted the risks of prioritizing technological novelty over core game mechanics and design. The AI model needs further refinement. Its ultimate impact will depend on its ability to go beyond superficial imitation and address the underlying technical and design complexities.
How Does This AI Experiment compare to the Original Quake II?
Here’s a swift comparison:
| Feature | Original “Quake II” | AI-Generated Demo |
|---|---|---|
| Graphics | Cutting-edge 3D graphics for its time. | Generative: Can exhibit blurred textures and visual inaccuracies. |
| Gameplay | Fast-paced first-person shooter with a well-designed level structure. | Experimental; exhibits limitations like object stability and player trapping issues. |
| intended Purpose | To offer a fully polished, classic game experience. | To explore what can be created using current machine learning approaches. |
| Story | War between humanity and the Strogg. | Based on a level from the game. |
Why is This AI Experiment Still Interesting, Despite Its flaws?
The experiment is valuable because it is an early exploration of generative AI in gaming. It provides opportunities to engage enemies and test basic functionalities. It is not a polished game, but a demonstration of AI’s potential in developing games.
What is Quake, Exactly?