The Role of Generative AI in Game Development
The Expanse: Osiris Reborn has recently joined the ranks of games openly acknowledging the use of generative AI in their development process. Developer Owlcat has taken a transparent approach, even as the gaming community expresses its discontent with this controversial technology. This week, Pearl Abyss, the studio behind Crimson Desert, issued an apology for AI-generated assets that were “unintentionally included” in the final version of their game. This scenario is reminiscent of past incidents where titles like The Alters and Clair Obscur: Expedition 33 have similarly claimed that AI elements were not meant to be part of the shipped product. Such occurrences highlight the importance of clearly distinguishing placeholder assets from final content.
Despite the views of industry figures like former Blizzard president Mike Ybarra, there is a palpable resistance among players against the integration of AI in gaming. Developers seem to be taking note, with many opting to either steer clear of AI or conceal its usage. While it is likely that generative AI is still employed to optimize certain processes, its presence in the final product remains minimal. For instance, Arc Raiders is actively moving away from using AI-generated voice actors, a decision that reflects a broader trend within the industry.
When AI-generated art does surface in games, developers often resort to the explanation that such assets were merely placeholders not intended for the final release. This rationale raises eyebrows, as placeholder assets are typically designed to stand out, often marked with clear indicators such as “CHANGE ME” to prevent their unintentional inclusion. A case in point is Slay the Spire 2, currently in early access, which features placeholder assets that are intentionally crude and mismatched with the overall art style, signaling their temporary nature.
Josh Sawyer, known for his work on classics like Fallout: New Vegas, recently shared examples of placeholder assets from his project Pentiment, further emphasizing the need for clear differentiation between temporary and final content. From simplistic stick figures to jpegs of recognizable characters, these placeholders are crafted to be conspicuous, ensuring they do not mistakenly make it into the finished game.
Even if one were to argue that generative AI could serve as a substitute for traditional concept art, this notion falters under scrutiny. The creative process thrives on the unique vision and artistic freedom of human creators, elements that generative AI cannot replicate. The act of creation is inherently challenging, and replacing the imaginative efforts of artists and writers with an AI model undermines the very essence of game development.
Moreover, the environmental impact of utilizing AI technology cannot be overlooked. The energy demands of data centers, which support AI operations, pose health risks to surrounding communities, adding another layer of concern to the conversation.
Assuming developers are sincere in their claims that AI assets were unintended, one might still question whether this is a mere oversight or a calculated risk taken by studios hoping players won’t notice. The absence of Steam disclosures until after the fact suggests a potential lack of transparency. In contrast, Owlcat’s candid acknowledgment of its AI usage sets a noteworthy precedent in an industry often characterized by ambiguity.
Ultimately, while placeholder assets may occasionally slip through the cracks—such as the infamous ‘Lorem ipsum’ text in Pentiment—players tend to react humorously to minor blunders. However, the reception of AI-generated artwork is markedly different; it often faces immediate backlash, leading to swift apologies from developers. This disparity raises an important question: when the use of AI stifles artistic creativity and invites ridicule, is it truly a worthwhile endeavor?
