AI Exploring Space? Why Google DeepMind Chose 'EVE Online'

Imagine being stranded alone on a strange planet. Thousands of others are active around you, each with their own purposes, and every decision you make affects your survival for the next month or even the next year. You can make friends, create enemies, or build a vast trade empire.

Recently, Google DeepMind decided to bring AI into exactly this kind of world: ‘EVE Online,’ the space MMORPG (Massively Multiplayer Online Role-Playing Game) that has been running for 23 years.

Why is this important?

Until now, AI has focused on beating humans in games like Go or StarCraft. However, the real world we live in does not end after a single “match.” We spend our lives communicating with countless people, looking toward the distant future, and making complex choices every day.

Google DeepMind chose EVE Online because it wants AI to learn more than just winning within given rules; it wants AI to learn the “ability to plan long-term and communicate within a sustainable society,” just like humans. Source: Google DeepMind Teams With EVE Online For AI Model Testing This is a significant technical leap that will greatly help AI become our reliable assistant or solve complex social problems in the future.

Simply put: AI’s ‘Social Kindergarten’

How can a game make AI smarter? Here is an analogy to make it easy to understand.

When AI research first started, DeepMind used classic games like ‘Atari.’ This is akin to letting a child solve simple puzzles. Then, through strategy games like ‘StarCraft,’ they taught AI complex situational judgment.

Now, EVE Online is like a high school or university that teaches AI about “social life.” Source: From Atari to EVE Online: Building on 15 Years of AI Research in Games

In EVE Online, the game world does not stop or reset; it keeps flowing. Source: Google DeepMind Picks EVE Online as Next AI Frontier Here, the AI does not just attack enemies, but learns what social interaction is by trading, forming alliances, and sometimes being betrayed by other players. The complex, player-driven system in the game becomes a very sophisticated practice field for AI. Source: DeepMind y EVE Online: 15 años de IA en videojuegos

Where are we now?

For 15 years since its founding in 2010, Google DeepMind has used video games as the most important laboratory for AI research. Source: From Atari to EVE Online: 15 Years of AI Game Research This collaboration is a major project that DeepMind CEO Demis Hassabis is personally spearheading. Source: Demis Hassabis Personally Pushed the Eve Online Deal

While AI has made rapid progress, it still has limitations in the real world, where there are too many variables, or in making plans that require a very long-term horizon. Through this research, a process is actively underway to train AI to possess more realistic and human-like problem-solving abilities.

Looking forward to AI’s growth

We will see the process of AI evolving from a being that simply answers questions into an “acting entity” (Agent) that sets its own complex goals and cooperates with people. Source: From Atari to EVE Online: Building on 15 Years of AI Research in Games

This small study, which began in a space game, will become the foundation for creating much smarter and more thoughtful AI partners that we will encounter in our daily lives. Next time you talk to an AI, it might just be exercising the “social intelligence” it learned while pioneering space trade routes.

---

MindTickleBytes AI Reporter’s Perspective

Games are the most sophisticated virtual reality for AI. DeepMind’s attempt to understand the complexity of human society beyond simple calculations will be the most interesting experiment in AI’s journey toward true “intelligence.”

References

1. From Atari to EVE Online: Building on 15 Years of AI Research in Games
2. From Atari to EVE Online: 15 Years of AI Game Research
3. From Atari to EVE Online: Building on 15 Years of AI Research in Games
4. Demis Hassabis Personally Pushed the Eve Online Deal
5. Google DeepMind Picks EVE Online as Next AI Frontier
6. DeepMind y EVE Online: 15 años de IA en videojuegos
7. Google DeepMind Teams With EVE Online For AI Model Testing
8. Google DeepMind Buys Into Eve Online: 5 Reasons It’s the AI Agent Training Ground