Counter-Strike's Dust II runs purely within a neural network on an RTX 3090
The "DIAMOND" diffusion model simulates CS:GO on Dust II with an RTX 3090, achieving only 10 FPS and exhibiting glitches, highlighting challenges in AI game simulation. Counter-Strike 2 offers a better experience.
Read original article
A recent demonstration showcased the "DIAMOND" diffusion model, which simulates gameplay of Counter-Strike: Global Offensive (CS:GO) on the iconic map Dust II using a single NVIDIA RTX 3090 GPU. Despite the innovative approach, the performance is notably low, achieving only 10 frames per second (FPS), making it unsuitable for competitive play. The project, led by Eloi Alonso and Adam Jelley, involved training the model with 87 hours of gameplay footage, allowing it to mimic player actions. However, the simulation is plagued with glitches, such as players being able to jump infinitely due to the model's lack of gravity and collision detection, resulting in surreal gameplay experiences. While the technical achievement is commendable, the current limitations highlight the challenges of using generative AI for game simulation. The demonstration serves as a reminder of the potential and pitfalls of AI in gaming, raising concerns about ethical implications in the industry. For those seeking a more traditional gaming experience, Counter-Strike 2 is available for free on Steam, offering a smoother experience at higher frame rates.
- The "DIAMOND" model simulates CS:GO on Dust II using an RTX 3090.
- Performance is limited to 10 FPS, making it impractical for competitive gaming.
- The model exhibits various glitches, including infinite jumping and weapon morphing.
- The project involved training on 87 hours of gameplay footage.
- Counter-Strike 2 is available for free on Steam, providing a better gaming experience.
Related
GPUDrive: Data-driven, multi-agent driving simulation at 1M FPS
The paper introduces GPUDrive, a GPU-accelerated simulator that generates over a million experience steps per second, enhancing multi-agent planning and training reinforcement learning agents using the Waymo Motion dataset.
Alien – CUDA-powered artificial life simulation program
ALIEN is a CUDA-based artificial life simulation tool that evolves digital organisms, offering real-time interactions and GPU acceleration. It serves educational, entertainment, and research purposes, requiring specific Nvidia graphics cards.
Diffusion Models Are Real-Time Game Engines
GameNGen, developed by Google and Tel Aviv University, simulates DOOM in real-time at over 20 frames per second using a two-phase training process, highlighting the potential of neural models in gaming.
New AI model can hallucinate a game of 1993's Doom in real time
Researchers from Google and Tel Aviv University developed GameNGen, an AI model that simulates Doom in real time, generating over 20 frames per second, but faces challenges with graphical glitches and visual consistency.
Run Stable Diffusion 10x Faster on AMD GPUs
AMD GPUs now offer a competitive alternative to NVIDIA for AI image generation, achieving up to 10 times faster performance with Microsoft’s Olive tool, optimizing models for enhanced efficiency and accessibility.
1 commentsSeriously though, we can’t have good quality home recording devices thanks to the possibility of music and movies being ‘stolen’, why do we imagine once this is possible gameplay will be treated differently by our copyright enforcing classes?
Related
GPUDrive: Data-driven, multi-agent driving simulation at 1M FPS
The paper introduces GPUDrive, a GPU-accelerated simulator that generates over a million experience steps per second, enhancing multi-agent planning and training reinforcement learning agents using the Waymo Motion dataset.
Alien – CUDA-powered artificial life simulation program
ALIEN is a CUDA-based artificial life simulation tool that evolves digital organisms, offering real-time interactions and GPU acceleration. It serves educational, entertainment, and research purposes, requiring specific Nvidia graphics cards.
Diffusion Models Are Real-Time Game Engines
GameNGen, developed by Google and Tel Aviv University, simulates DOOM in real-time at over 20 frames per second using a two-phase training process, highlighting the potential of neural models in gaming.
New AI model can hallucinate a game of 1993's Doom in real time
Researchers from Google and Tel Aviv University developed GameNGen, an AI model that simulates Doom in real time, generating over 20 frames per second, but faces challenges with graphical glitches and visual consistency.
Run Stable Diffusion 10x Faster on AMD GPUs
AMD GPUs now offer a competitive alternative to NVIDIA for AI image generation, achieving up to 10 times faster performance with Microsoft’s Olive tool, optimizing models for enhanced efficiency and accessibility.