谷歌DeepMind近日宣布推出了一款名为SIMA（可扩展、可指导、多世界智能体）的新型人工智能系统，该系统专为学习和掌握游戏技能而设计。SIMA的独特之处在于，它旨在模拟人类玩家的游戏行为，而非展现出纯粹的机器式效率。
据《The Verge》报道，SIMA目前仍处于研究阶段，但已经在《Valheim：英灵神殿》、《无人深空》和《模拟山羊》等不同类型的游戏上进行了训练。通过这些训练，SIMA不仅学习了如何在游戏中生存和进步，而且能够适应各种环境和挑战，展示出类似人类的适应性和策略性思考。
谷歌DeepMind的最终目标是让SIMA具备玩任何类型游戏的能力，包括那些没有明确结束点的开放世界游戏。这一突破性的进展可能对人工智能研究产生深远影响，不仅可能革新电子游戏的设计和AI在游戏中的应用，还可能为其他领域如机器人学习、决策制定等带来灵感和新技术。
SIMA的出现再次证明了人工智能在模仿和理解复杂行为方面的潜力，同时也引发了关于人工智能与人类智能相似性的讨论。随着技术的不断发展，我们有理由期待SIMA在未来能实现更广泛、更深层次的学习，进一步拉近人工智能与人类智能之间的距离。
英语如下：
**News Title:** “Google DeepMind Unveils SIMA: An AI Gaming Prodigy with Human-like Learning Capabilities”
**Keywords:** Google DeepMind, SIMA Agent, Game Training
**News Content:** Google DeepMind recently announced the development of a novel artificial intelligence system called SIMA (Scalable, Instructable, Multi-World Agent), specifically designed to learn and master gaming skills. The uniqueness of SIMA lies in its intention to emulate human players’ behavior rather than exhibiting sheer machine efficiency.
As reported by _The Verge_, SIMA is still in the research phase but has been trained on a variety of games, including _Valheim:英灵神殿_, _No Man’s Sky_, and _ Goat Simulator_. Through these training sessions, SIMA not only learns how to survive and progress in games but also adapts to diverse environments and challenges, demonstrating human-like adaptability and strategic thinking.
Google DeepMind’s ultimate objective is to equip SIMA with the ability to play any genre of game, including open-world games without a clear endpoint. This breakthrough could have a profound impact on AI research, potentially revolutionizing game design and AI applications in gaming, as well as inspiring new techniques and innovations in fields like robotic learning and decision-making.
SIMA’s emergence underscores the potential of AI in emulating and understanding complex behaviors and sparks discussions on the similarity between artificial and human intelligence. With ongoing technological advancements, it is reasonable to anticipate that SIMA will achieve broader and deeper learning in the future, further narrowing the gap between AI and human cognition.
【来源】https://www.theverge.com/2024/3/13/24099024/google-deepmind-ai-agent-sima-video-games
Views: 1