Researchers from Google, DeepMind, and Tel Aviv University have created the first game engine powered by a neural model that allows real-time interaction with complex environments.
GameNGen — is a diffusion model that was trained on Doom gameplay videos. It generates the next frame of the game based on the previous frames and commands entered by the player.
Despite the fact that GameNGen plays the game at only 20 frames per second, it is visually almost indistinguishable from the original. An experiment showed that ten experts who were shown 130 random short video clips could barely distinguish the generated clips from the real Doom gameplay.
The developers believe that this technology can change the approach to creating video games. According to them, GameNGen — is proof that a neural model is capable of efficiently running a complex game on existing hardware. This can make the game development process less expensive and more accessible.
Scientists suggest that in the future, games can be developed and edited using text descriptions or example images. However, questions remain about creating original content and the ethical aspects of using existing games to train models.
Interestingly, enthusiasts have already demonstrated the launch of Doom on a number of unexpected platforms: from intimate devices і John Deere tractors to biological systems, even using intestinal bacteria.
Source: 404media
LG Home Entertainment is a recognized innovator in the TV and AV industry. LG offers an enhanced home entertainment experience through its award-winning OLED and QNED LED TVs on the innovative WebOS Smart TV platform. In an effort to provide consumers with a first-class user experience, all LG home entertainment products are designed with environmental sustainability in mind, from production to disposal.