谷歌DeepMind在人工智能领域的探索再进一步，最近推出了一种名为SIMA（可扩展、可指导、多世界智能体）的新型人工智能Agent。SIMA的独特之处在于，它通过训练来学习游戏，展现出与人类玩家相似的游戏策略，而不仅仅是一个单一任务的高效执行者。
据《The Verge》报道，SIMA目前仍处于研究阶段，但其潜力已经初露端倪。在《Valheim：英灵神殿》、《无人深空》和《模拟山羊》等不同类型的游戏环境中，SIMA已经成功地进行了训练。它的目标不仅仅是掌握特定游戏的规则，而是要能够适应各种游戏，包括那些没有明确终点或开放世界设定的游戏。
这一进展标志着AI在游戏领域的学习能力达到了新的高度。通过模仿人类玩家的行为模式，SIMA有可能在游戏测试、设计优化，甚至是未来的娱乐互动中发挥重要作用。然而，如同所有人工智能项目，SIMA也面临着如何保持其学习的灵活性和适应性，同时避免过度依赖特定策略的挑战。
谷歌DeepMind的这一创新，再次展示了其在人工智能领域的领先地位，同时也预示着未来AI在游戏和其他复杂任务中的应用可能将更加广泛和深入。
英语如下：
**News Title:** “Google DeepMind Launches SIMA: A New AI Breakthrough, Training Intelligent Agents to Play Games Like Humans”
**Keywords:** Google DeepMind, SIMA Agent, Game Training
**News Content:**
Title: Google DeepMind Unveils SIMA Agent, Teaching Game Skills in a Human-Like Manner
Google DeepMind’s foray into artificial intelligence has taken another leap forward with the introduction of a novel AI Agent called SIMA (Scalable, Instructable, Multi-World Agent). The uniqueness of SIMA lies in its ability to learn to play games through training, exhibiting strategies akin to human players, rather than just being an efficient single-task executor.
According to a report by _The Verge_, SIMA is still in the research phase, but its potential is already evident. It has successfully undergone training in diverse gaming environments such as _Valheim: Valhalla_, _No Man’s Sky_, and _Goat Simulator_. Its objective goes beyond grasping the rules of specific games; it aims to adapt to a variety of games, including those with no clear end or open-world settings.
This advancement signals a new peak in AI’s learning capacity within the gaming realm. By mimicking human players’ behavioral patterns, SIMA could play a significant role in game testing, design optimization, and potentially future interactive entertainment. However, like all AI projects, SIMA faces the challenge of maintaining its learning flexibility and adaptability while avoiding over-reliance on particular strategies.
Google DeepMind’s innovation underscores its leading position in AI research and foreshadows a broader and more profound application of AI in gaming and other complex tasks in the future.
【来源】https://www.theverge.com/2024/3/13/24099024/google-deepmind-ai-agent-sima-video-games
Views: 1