谷歌DeepMind最近推出了一种名为SIMA的人工智能代理，它能够通过训练学习各种游戏技能，使其在玩游戏时的表现更接近人类，而不是单纯追求高效率的传统AI。SIMA的全称是可扩展、可指导、多世界智能体（Scalable, Intelligent, Multi-world Agent），目前仍处于研究阶段。
DeepMind已经在多款游戏中对SIMA进行了测试，包括《Valheim: 英灵神殿》、《无人深空》和《模拟山羊》。该智能体的设计目标是学习如何玩任何类型的游戏，包括那些没有明确线性路径和结束的开放世界游戏。
SIMA的出现标志着游戏AI技术的一大进步，它不仅能够提升游戏体验，还有望在更广泛的领域中应用，如医疗、金融和自动化系统等领域。
英文标题：Google DeepMind Launches SIMA, a New Frontier in Game AI
英文关键词：Game AI, Machine Learning, Multi-world Agent
英文新闻内容：
Google DeepMind has unveiled a new artificial intelligence agent called SIMA, which can learn a variety of gaming skills through training, making its gameplay more human-like rather than purely focused on efficiency. Named Scalable, Intelligent, Multi-world Agent, SIMA is currently in the research phase.
DeepMind has tested SIMA in multiple games, including Valheim: 英灵神殿, No Man’s Sky, and Goat Simulator. The design goal for SIMA is to learn how to play any type of game, including open-world games with no clear linear paths or endings.
The emergence of SIMA represents a significant advance in game AI technology, which not only enhances the gaming experience but also has the potential for broader applications in fields such as healthcare, finance, and automated systems.
【来源】https://www.theverge.com/2024/3/13/24099024/google-deepmind-ai-agent-sima-video-games
Views: 1