The integration of generative AI into PC gaming has been met with widespread player apprehension and developer caution. Amidst this controversy, EVE Online developer Fenris Creations, formerly CCP Games, has announced a significant research partnership with Google DeepMind. This collaboration will see DeepMind acquire a minority stake in Fenris while training its AI agents on a dedicated, offline instance of the expansive space MMO.
Fenris CEO Hilmar Veigar Pétursson and DeepMind co-founder Adrian Bolton discussed the partnership at the recent EVE FanFest. While specific plans for EVE Online remain largely undefined, the announcement did not elicit the significant backlash one might expect from the game's notoriously vocal player base, despite broader industry trends. DeepMind's history with AI in gaming, including its Grandmaster-level StarCraft 2 bot and its use of Atari games for agent grading, was highlighted. Bolton noted potential applications of simulated 3D movement in areas like self-driving car technology, though this is unlikely to be a focus for EVE's complex ship combat.
Pétursson addressed concerns about AI 'players' diminishing multiplayer engagement by referencing chess's continued relevance after Deep Blue's victory over Garry Kasparov, suggesting AI does not inherently make human competition obsolete.
Addressing EVE's Legacy Codebase
During an audience Q&A, potential applications of AI for EVE Online began to emerge. One key area identified is the use of AI 'coaching agents' to help manage and repair EVE's decades-old, complex codebase, often referred to as "spaghetti code." This legacy code, accumulated over 30 years of development and staff turnover, contains lost context and poor documentation, making it difficult to maintain and update. Pétursson explained that these coaching agents have already proven valuable in identifying memory and security issues within the codebase, with humans performing the actual code replacement.
Enhancing In-Game NPCs and Narrative
Another potential application involves enhancing EVE's existing procedural generation systems by integrating generative AI features into non-player characters (NPCs) that provide missions, known as Agents. Pétursson believes this could make these characters feel more dynamic and "alive." He noted that the system managing most Agent missions is 24 years old and suggested that AI could breathe new life into these interactions, fitting well with the game's existing narrative structure.
Despite the generally positive reception at FanFest, concerns about AI's potential downsides, such as accidental data deletion or the creation of unnerving NPC interactions, persist. However, the EVE Online community, known for its history of vocal dissent against controversial monetization or game changes, has shown a surprising lack of hostility towards this AI partnership. Players actively engaged in discussions about the AI's potential to fix long-standing issues like the problematic "spaghetti code" that they feel hinders the game's progress.
Community figures, including alliance leaders and veteran players, have expressed a range of reactions from excitement to cautious curiosity. Some, like Dylana Aivo, a 10-year veteran, acknowledge a lack of clarity on the partnership's specifics but are willing to observe its development, believing AI can be integrated without compromising the game's core spirit of player-driven economy and PvP. Others, like shutupandshave, a former leader of the Black Omega Security corporation, voiced humorous anxieties about AI capabilities but ultimately expressed faith in Fenris's track record of successful partnerships with major technology companies, citing a past collaboration with IBM to improve server infrastructure.
Fenris is also experimenting with a GPT-style tutorial chatbot called Aura Guidance, currently in a limited trial. While this predates the DeepMind deal, Fenris executive producer Snorri Árnason differentiates it from generative AI, stating it aggregates existing tutorial information rather than generating new content. This initiative has reportedly been well-received.
Player Trust and Transparency
Fenris community manager Peter 'FC Swift' Farrell emphasized the EVE player base's discerning nature, calling them "the biggest bullshit detectors on the planet." He believes Google DeepMind has proactively and transparently communicated the nature of the research partnership to players. Farrell stated that DeepMind has a proven, respectful approach to game development, referencing their work on StarCraft and Go, which has helped allay player fears about potential AI overreach.
While the exact benefits for Fenris and the financial details of the deal, reportedly in the millions, remain undisclosed, the partnership seems poised to address some of the EVE Online community's most persistent technical complaints. The collaboration between Fenris Creations and Google DeepMind marks a significant step, and for now, EVE's dedicated players appear willing to cautiously explore the possibilities AI may bring to New Eden.