AI In Games Services: Art Production

It seems 2023 is the year that we all start fully paying attention to AI. It’s been an important, incoming topic in games services for years. Now we’re seeing commercial AI implementations that are getting people truly excited.

What we're still to see, to an extent, is where AI will slot-in within the games production environment. Some services naturally lend themselves to AI solutions. Others exist somewhere in-between, always requiring some human involvement.

But what does a games industry powered by AI look like? How does that affect the way we create games and where does AI fit in? How does that change how studios build their teams? And ultimately, how do these changes affect players?

AI in Art Production

Art Production in games is one of the most discussed areas regarding AI enhancement. Developers are already using AI art within games in certain countries. Recent releases within the Chinese games market have been using AI. These games generally use AI to generate character and environmental art. The technology can produce art 40 times faster than a human. It also enables those without art skills to generate art using keywords. Thereby reducing the art skills required on an art team.

But who owns that art? This question is obviously important in the games industry. The question has caused some major games companies to internally ban the use of AI art, due to the murky situation around its ownership.

Copyright

The question of ownership relies on numerous factors. Typically, copyright law looks at who created the copyrighted content. But in this case that would be an AI. While the user may have prompted the result, it was the AI that created it. There’s then a follow-up question based on a core issue with the legality, and the morals, of AI art. AI neural networks currently do not create from nothing. AI art is the result of these AI systems consuming the work of previous artists. This allowed “The Next Rembrandt”, a project creating a new Rembrandt art piece, which used machine learning to study his previous work. It absorbed and detailed his use of light, color, brush strokes, and patterns. This was all then used to create a completely new portrait. This raised the question; does the copyright belong to Rembrandt, the AI, or the team that generated it? This issue alone has caused many companies to write-off its use, at least in a current context. These issues compound when considering the number of artists whose art may have contributed as data to create a game asset.

The UK is one of the few countries that doesn’t have a huge issue with the copyright of AI generated content. It’s covered in the Copyright, Designs and Patents Act 1988. It assigns ownership to the person "by whom the arrangements necessary for the creation of the work are undertaken”. Unfortunately for AI developers, the UK is one of few countries that has this law. Taking Europe as an example, countries define the treatment of computer-generated work differently. Few align closely with the UK’s law. Spanish copyright law specifically designates the “natural person” who creates it. Germany defines the owner as the author, but when defining author, it heavily implies it must be a human. The lack of consistency is one of the major hurdles of utilizing AI generated art. In different countries assets would have different copyright ownership. This uncertainty is always to be avoided.

But even if a game releases only in regions where asset copyrights are clear, what effect does that have on the game itself? And what about the players who play it?

Player Enjoyment

In games we have something that’s somewhat comparable in concept to AI generated assets: procedural generation.

Procedural generation takes rules and pieces or other assets and creates things from them. We can all think of examples like levels, worlds, maps, etc. An easy example to reference is No Man’s Sky. It uses procedural generation to build its universe with new planets for each player. While it’s now well regarded, on initial release players complained about a lack of consistent quality and distinct variety.

But like procedural generation, quality can be an issue with AI generation. Recently players in China have bemoaned the use of AI art. Largely due to issues with its quality and lack of consistency. Players pointed to generated player models and character art as an example of quality issues. They often break immersion, where misaligned facial features are immediately recognizable. Breaking player immersion is a notable problem in any game, especially when it’s something graphical and unavoidable. There are also players commenting about how similar some games look that feature AI generated art. Games can end up with similar results, especially if generated from the same AI systems or using similar keywords.

Many of these complaints have been from Mobile titles released in the region. Constant updates and events lend themselves to the speed of generating AI art. Due to that speed and efficiency, AI art has also been having a negative effect on artists in the industry.

Illustrators

Discussion around the effects of machines on different industries has continued for centuries. From the printing press affecting bookmakers to robotics in the automotive industry. Technology almost always impacts job opportunities and roles within the related industry.

Games haven’t had this same issue so far…. Almost all parts of creating a game have traditionally required humans to do the work. We’ve had tool and technology evolutions that have simplified jobs and altered specializations, but none of these are quite on this level of disruption. AI art creation is one of the first areas we’re seeing a technological change, reducing the amount of human work required.

Reports suggest that over the last 12 months 70% of illustrator jobs in the Chinese games industry have disappeared. Recruiters noticed this last year and have said that so far the positions have not reemerged. 70% is a huge percentage of positions to lose. But it’s no surprise with the 40-fold increase in output when comparing AI generation to human. Teams now can include a few artists using their understanding and skills to generate AI art endlessly. In an interview with Rest of World, Xu Yingying, an illustrator at an independent game art studio, said that one third of their character art illustrators were let go in the last year, recounting:

“Two people could potentially do the work that used to be done by 10”

The Impasse of Human and AI Art

This change has been a long time coming and it isn’t going to be solved quickly. AI generated art is an incredible tool but one that has far reaching effects. Whether it’s the morals of AI generation based on human artists work, the effects on employment, or the cold world of copyright law. It’s something that will change the industry but not all at once. Due to different laws, and lack-there-of in some places, we’ll likely see an adapting rollout as we go forward.

The capabilities of AI are endless, and art is one of the first disciplines where this conversation needs to happen. Currently we can say that utilizing AI art will increase initial profit but is likely to disappoint players, at least to an extent, which could lead to long term issues. Dissatisfaction builds as more games use it. This could create a future problem for companies using it habitually. Currently, and for the foreseeable future, human-made assets will continue to be the industry standard. That seems to stay true generally for both Western and Eastern games markets. How long this rule will hold true is still yet to be seen.

Next, we’ll look at AI in localization, an area it’s already being used, and how human involvement is still utilized.