Google DeepMind Trains Its AIs in Eve Online

⚡

Key Takeaways

1Google DeepMind will use the MMORPG Eve Online to train its AI models, leveraging its complex and interactive universe.

2DeepMind has invested several million dollars in Fenris Creations, the new name of CCP Games, to strengthen this collaboration.

3The emergent narratives and massive battles of Eve Online provide an ideal training ground for DeepMind's AI algorithms.

💡Why it matters — This initiative could revolutionize the way AIs are trained, using rich and dynamic virtual environments.

Google DeepMind and Eve Online: An Unprecedented Collaboration for AI

In a bold move, Google DeepMind has chosen to partner with the massively multiplayer online role-playing game (MMORPG) Eve Online to train its future artificial intelligence models. This game, which combines space combat, piracy, and virtual commerce, is developed by CCP Games, an Icelandic company. For over twenty years, Eve Online has provided players with a vast universe where they can explore, form alliances, and engage in various activities ranging from mining to space battles.

A Strategic Investment in Fenris Creations

DeepMind has also announced that it has taken a minority stake in Fenris Creations, the new name for CCP Games. Although minority, this investment represents several million dollars, according to Hilmar Veigar Pétursson, CEO of Fenris Creations. This strategic decision allows the company to maintain its independence with its own board of directors, thereby facilitating crucial decision-making for the future of Eve Online. Previously, CCP Games was owned by Pearl Abyss, a South Korean publisher known for the game Crimson Desert, before being acquired by its management.

Eve Online: An Ideal Training Ground for AI

Launched 23 years ago, Eve Online has captivated a wide audience, making it a logical choice for DeepMind. The game is renowned for its emergent narratives and large-scale battles, some lasting several hours and involving high-value virtual assets. These elements, influenced by stakes perceived as real, are of particular interest to Google, which seeks to analyze player data to train its AI models.

Video Games: A Cornerstone of Innovation at DeepMind

Demis Hassabis, CEO of Google DeepMind, emphasized the importance of video games in his journey and in the company's advancements. Since childhood, Hassabis has been passionate about games, having started his career designing complex AI simulation games. He recalls that flagship projects of DeepMind, such as Atari DQN, AlphaGo, and AlphaStar, were developed through gaming environments, thus demonstrating their crucial role in the development and testing of AI algorithms.

Secure and Promising AI Research

The goal of this collaboration is to explore new gaming experiences while advancing AI research securely within the complex universe of Eve Online. DeepMind plans to study player behavior on isolated servers to avoid disrupting live gameplay. Ultimately, this research could not only enrich the understanding of game dynamics but also contribute to the improvement of Eve Online itself.