The Revolution of Generative AI Based Games
Unleashing Limitless Creativity and Redefining Game Design
The biggest media and entertainment transformations come from new technologies enabling new formats that are matched to new distribution channels. Generative AI (GenAI), GenAI-powered games, and UGC gaming platforms represent just such a transformative opportunity. One in which players, not studios, and not publishers, control the future.
Over the past few years, the gaming industry has undergone a remarkable transformation propelled by advancements in networking, computing power, device capability, and of course digital distribution. Today, GenAI, represents perhaps the biggest step function innovation since the advent of mobile. GenAI is poised to radically alter the process and cost of creating game assets, avatars, environments and even code. However, while much of the hype around the technology has focused on its ability to change how, how quickly, how cheaply and how easily games are created, its most transformative impact will likely be in how it enables new gaming formats i.e. new types of games entirely.
Generative AI empowers players to craft novel characters, stories, settings, and more by taking seeds of inspiration and crafting entire products, be that characters, environments, or even code and game mechanics. This capability has the potential to radically alter much of the gaming landscape. At first order, its ability to quickly and cheaply create content from ideas lowers the barrier to execution enabling more customization and likely increasing the total amount of output. More broadly, GenAI has the potential to radically alter game loops. As GenAI technology improves, game content will become more and more expressive, individual and disposable. In doing so, GenAI will transform game loops, extending them beyond gameplay into content creation and social sharing.
Just as critically, this tech is emerging into the modern world of UGC gaming platforms and social platforms more broadly. UGC gaming platforms like Roblox and UEFN, not to mention social platforms like TikTok and Discord, are well suited to serve as the channels for GenAI games. Social and algorithmic distribution is perfectly matched for gaming content that is by its very nature designed to be shared, consumed, remixed and redistributed. Already gaming content serves as one of the most popular substrates for UGC content consumption across non-gaming platforms i.e. serving as the building blocks for Twitch streams, YouTube videos and TikTok clips. But as GenAI games turn games themselves into expressive interactive content generators, these platforms along with UGC gaming platforms will hyperscale these games as users adopt them not just for gameplay but for social content purposes. GenAI games are far better than normal games at allowing people to express themselves and as a result, user adoption for social purposes creates a powerful positive reinforcing loop between GenAI games and UGC platforms.
This positive reinforcing loop will radically alter the gaming landscape and empower players to create the future.
New Tech
GenAI is effectively a series of reinforcement learning mechanisms that can generate novel hybrid creations and extrapolate what the image, text, or more generally content should be based on a a smaller set of inputs. By leveraging these machine learning techniques, AI algorithms can create or more typically aid in the creation of characters, environments, narratives, and game levels. To date, much of the early hype has been around text-based game generation and object/asset generation, particularly 2D asset generation given the technology’s current limitations and human preference for visual gameplay. While some of the hype has run ahead of current capability, there is limited doubt that further improvements will yield increasing breadth and fidelity across both 2D and 3D gameplay writ large. We may disagree on timelines and quality, but it’s clear that the technology will eventually reach its intended destination: the ability to create high-fidelity gameplay from minimal inputs, perhaps even a few words or drawings.
New Formats
Most of the hype around the technology, at least when it comes to games, has focused on how it will upend the game production pipeline i.e. how it will reduce cost or increase speed for game asset or even game generation. Indeed most of the debate around the technology has centered on questions of how it will impact game production and game developers, designers and artists. These are important questions and worthwhile debates, but lost in these discussions is the fact that GenAI isn’t just going to change how games are made, how long they take to get made and who makes them, but the nature of the games themselves. CGI didn’t just change how movies were made, it changed which movies and genres were made.
Traditional game design heavily relies on scarcity as a means to create value for digital goods and mechanics. GACHA games entirely rely on scarcity to drive gameplay. Even less directly scarcity-driven games ultimately revolve around user actions and choices being bound to limited experiences and optionality. However, generative AI-based games challenge these established norms. The ability to generate an infinite amount of content renders scarcity nearly obsolete. The capability to tailor experiences perfectly upends the notion of directed gameplay. Instead, the value of these games lies in the creative input and individual experiences that players bring to the generated content.
All this suggests that game designers will need to think critically about GenAI games and design GenAI games from the ground up as opposed to simply bolting on the technology to existing game loops. A GACHA game with GenAI bolted on is simply a bad GACHA game. A role-playing game with GenAI bolted on is simply a never ending RPG.
At Infinite Canvas, the game studio and publisher I founded, we experienced this firsthand. When initially designing CreatureCraft, a discord-based, GenAI-powered game, we initially attempted to take an existing GACHA-based gameloop based on a popular format called Karuta and simply add GenAI to it. The result was initially less than stellar. By adding GenAI we transformed a game based on scarcity and social status into something else entirely, but in doing so we eliminated the various game loops that made the initial game interesting. Users could create infinite versions of characters, rendering any sense of collecting moot. Our game loop became less driven by coveting and then collecting cards created by others and more about creating one’s own. But at the same time, the social value of creating a card paled in comparison to the value of showing off a rare card desired by others. We ended up with a game that neither fully embraced the creative impulses unlocked by GenAI nor the collecting urges inherent to traditional games.
As such, we had to go back and redesign the entire game to take advantage of GenAI, rethinking what it means to collect, and reimagining what drives social value and status in a world where every card is bespoke. The result was a blend between a GACHA and an RPG that leans into the notion of creation, making other resources scarce, while emphasizing the social value not of collecting but of creation. We emphasized the importance of crafting, and crucially of modifying and continually improving one’s creations. We deemphasized the notion of collecting for its own sake. We reimagined progression and had to redesign the gameplay to encourage constant experimentation as opposed to static hoarding.
While we were having success with the game, what we learned was perhaps more interesting and important. First, players are far more attached to content, characters and situations they crafted themselves. We believe this to be a pretty key insight -- fundamentally there's a human attachment to self-created objects and narratives. In a GenAI future, there is an opportunity to create IP lock-in and scale effects based on IP as opposed to mechanics, gameplay, or even social networks. Users grew intensely attached to their creations, even eschewing more optimal gameplay and progression choices to nurture and maintain their own. Second, social status became far more distributed around a player’s imagination and the quality of their adventures, characters and interactions than pure progression level and “skill.” Third, users were far more interested in horizontal gameplay, utilizing their creations in new ways and new mechanics, than necessarily going deeper in 1 or 2 game loops. Depth and richness were provided by mastery of creation.
New Gameloops
Perhaps most of all, we found that Generative AI-based games shake the very notion of what constitutes gameplay. In modern games, the primary focus revolves around mechanics and progression. Users of course use games to generate content, either as substrate for Twitch streams or even as engines for creation via gameplay videos or in some cases as production engines for Machinima. But for most players, especially those not especially skilled, the gameplay ends with the game itself.
However, with generative AI allowing for far more expressive and instant content creation, inside of the game itself, accessible to not just skilled players or producers but every player, the game loop expands beyond gameplay alone. Players become active content creators at all times, engaging in storytelling, character customization, and world-building as they play. The game loop doesn't end with the "You Win" screen. It extends to sharing your wins on social media, and more critically, how you won, and what you created to win.
The most human reaction to creating something awesome is to share it, and GenAI games tap into this human reaction like no other. This has the upside of fostering a profound sense of community and collaboration among players. It also fundamentally changes how designers need to think about their games. Already games like Fortnite and Roblox function as social networks. But with GenAI, every game either becomes a social network itself or becomes legible in part only in the context of its users seeing the game as an extension of other social networks, to be used as a content creation engine for those very distribution channels. As such, with our GenAI games, we found ourselves designing content creation, sharing, and networking features not as afterthoughts on top of the game, but as primary gameplay features. We also found ourselves not solving for the average user experience but rather for tail-end outcomes that would generate the kind of content that users could make viral. It didn’t matter if the median experience was great, what mattered to users was that some meaningful percentage of the time they could create something that had the potential to go viral. Winning no longer meant earning a high score, it meant garnering the most views.
New Distribution
Given this emphasis on content creation and social sharing, generative AI-based games are perhaps best suited for distribution through existing UGC gaming platforms like Roblox or social platforms like Discord and TikTok than they are traditional platforms like Steam or even the Apple App Store. These platforms provide a robust infrastructure for players to share their creations and connect with fellow gamers on an enormous scale. The algorithmic nature of these platforms ensures the curation and visibility of the most captivating and relevant generative AI-based games and GenAI-based content. While this is also true for modern games, with GenAI games, on average more users create more content, moreover, a larger number of users want to share their creations and therefore built-in social and algorithmic distribution matters even more for GenAI games. In a world of content abundance, like that enabled by GenAI games, distribution becomes the scarce resource. Studios and publishers that adapt will thrive, but those that eschew these social and algorithmic channels will find themselves increasingly competing against an ever-rising tide. While not a game itself, the enormous success of Midjourney’s image generation bot on Discord highlights the symbiotic relationship between GenAI applications and social/algorithmic distribution mechanisms and how important distribution is in an age of abundance.
Conclusion
Generative AI-based games mark the advent of a new era in game design, enabling players to actively shape their gaming experiences and create not just customize. As the technology matures, developers will empower users to delve into character creation, story development, and world-building. Scarcity will be fundamentally redefined. Games will be less about the progression and vision of developers and more driven by the creative contributions and unique experiences players bring to the game. More importantly, as GenAI games proliferate, we will see a shift from gameplay-focused loops to content creation and social sharing-focused gameplay. At an extreme, we may even see games themselves become disposable memes. Regardless, the future won’t be controlled by studios or publishers or game engines. It will be built by players and distributed by communities.