Key Takeaways
- Modern AI is a pivotal tool for game devs, aiding in inventive storytelling and unique gameplay mechanics.
- AI-native games showcase the extraordinary possibilities AI brings to the gaming world by incorporating LLMs into games.
- While AI is revolutionizing gaming, it should be viewed as a tool complementing—not replacing—traditional game development methods.
There are plenty of concerns around the use of generative AI in modern video games. While some developers are looking into incorporating AI-generated assets and narratives into their games, others are using AI to bring seemingly impossible ideas to life.
AI has been a part of gaming since its inception, but recent advancements in AI technology aren’t as widely used by game developers. Generative AI and similar tools may seem useful for developers—especially since AI can be used to quickly create new assets and narratives—but with the numerous ethical concerns and quality issues surrounding these technologies, it’s usually better to leave these responsibilities in the hands of human professionals.
That doesn’t mean game developers should ignore all forms of modern AI. Instead of using AI to perform tasks that humans could complete with better results, some recent games are using AI to deliver experiences that can’t be produced with traditional development tools. These are known as AI-native games, meaning they incorporate AI programs into their core gameplay mechanics. While most of these titles are still in early access, they perfectly demonstrate why modern AI is becoming a game-changing tool for developers.
Before discussing some examples, it’s important to understand what a large language model (LLM) is, and why it’s the most important part of many AI-native games. An LLM is a framework that enables AI-powered programs like chatbots, image generators, and certain search engines to process and respond to custom inputs.
But how can LLMs support video games? The answer is found in many AI-native games, which incorporate LLMs to give players a seemingly infinite number of ways to interact with their virtual worlds and characters. Whereas most games feature limited dialogue options and often force you to complete objectives in specific ways, AI-native games provide a level of freedom that can’t be found in traditional video games.
1001 Nights
Released a few years before ChatGPT, 1001 Nights provided an early glimpse at the creative potential of LLMs in entertainment. The game puts you in the role of Shahrzad, a storyteller with the power to bring fiction to life, who must use her magical abilities to defeat the evil King Shahryar.
You and the king (who is controlled by the AI) take turns telling the story, and your goal is to summon four weapons by weaving them into your tale. However, you can’t outright mention weapons without angering the king, so you must trick him into mentioning swords, spears, or other deadly armaments.
This inventive premise turns the storytelling of 1001 Nights into a complex puzzle, with the AI adding an extra layer of unpredictability to the mix. You never know if you’ll successfully manipulate the king into summoning a weapon, or if he’ll wreck your plans by introducing a surprise twist into your story.
That’s only half of the challenge. Collecting all four weapons will initiate a turn-based battle against the king, during which you will attack with the summoned weapons, but there’s more. The strength of each weapon is determined by its context in the story. An ordinary knife won’t do much damage, but a legendary sword that’s strengthened by an elaborate backstory can quickly turn the tide during the final confrontation.
1001 Nights is a short yet stellar showcase of the creative ideas that modern AI could bring to gaming. The game places just enough restrictions on you to transform storytelling into a challenging puzzle, while also providing enough freedom for you to be creative with your stories.
Suck Up!
One of the most notable gaming innovations introduced by modern AI is fully immersive dialogue. Plenty of modern games aim to make conversations with non-player characters (NPCs) seem as realistic as possible, but there’s no way for developers to program every NPC to have a response to every question you could possibly imagine. Even the dialogue systems in older role-playing games like the Ultima series and early Elder Scrolls entries—which allow you to question characters using specific keywords—leave a lot to be desired.
Thankfully, modern AI is making limitless dialogue a reality through games like Suck Up!. You play as a bloodthirsty vampire and must convince the eccentric residents of a small neighborhood to invite you into their homes (You can probably guess what happens afterward). However, knocking on their doors and politely asking to suck their blood won’t get you anywhere, which is why you must employ creative lies to charm your way into each house.
Despite the simple concept, Suck Up! succeeds at making every conversation feel unique. Each character has a distinctive personality that determines their likes and dislikes, so you can’t always rely on the same tricks to get you through every door. You might find yourself bonding over shared interests with one character, and trying to sell an imaginary product to the next. You also have a wide variety of disguises that help you assume different roles, as characters will acknowledge your current clothing (or lack thereof).
Vaudeville
Much like Suck Up!, Vaudeville uses AI to create an open-ended dialogue system that allows you to ask characters anything that comes to mind. However, Vaudeville builds upon this concept with an elaborate murder mystery. Stepping into the shoes of Detective Martini, you’re tasked with uncovering the identity of a mysterious killer involved in a string of recent deaths. You’ll interrogate a large cast of suspects, each of whom possesses unique personalities, backstories, and relationships that shape their interactions with you.
Vaudeville is one of the most ambitious AI-native games out there, and the game uses its technology to create an unforgettable experience (for better or worse). Conversations with Vaudeville‘s AI can get surprisingly in-depth, as characters will remember topics from previous discussions and challenge your accusations with their own alibis. Furthermore, even when presented with irrefutable evidence, they may still deny your claims if you haven’t fully convinced them to confess.
Unfortunately, Vaudeville‘s inconsistent AI struggles to support the weight of its ambitious premise. Despite the game’s attempts to make its characters seem more complex than those found in other AI-native games, most of the cast blends together with similar personalities and speaking patterns. Conversations are also bogged down by characters’ stilted AI voices, as well as their tendency to go off on random tangents and give contradictory answers to simple questions.
Vaudeville is an entertaining showcase of conversational AI, but it sadly isn’t a satisfying murder mystery. The AI’s inconsistent quality and confusing responses can be frustrating when you’re trying to solve the case. However, it’s ability to hold extensive (and often bizarre) conversations makes it an entertaining—albeit deeply flawed—demonstration of AI’s potential contributions to the future of gaming.
Browser Games
Although most AI-native games use LLMs for dialogue, many browser-based games are using the technology in other inventive ways. The most famous example is Infinite Craft, a game centered around combining objects, elements, and intangible concepts to create new items. Although you start with only a few basic elements, you’ll gradually build up a massive list of the items you’ve crafted, which can be used to discover even more combinations.
As the name suggests, Infinite Craft supports a seemingly infinite number of items, as the game uses an LLM to allow for nearly any combination of items you can imagine. The results could be as simple as water and clouds combining to form rain, or something as bizarre as fusing Dracula and a dinosaur to form “Dracorex”.
What Beats Rock uses AI in a similar way to create a Rock, Paper, Scissors-inspired game that allows you to use any object you can think of. However, each round has you face off against the item you last used without repeating answers.
For example, you might start by using paper to beat rock. Now you need something that can beat paper, so you might type “scissors”. The rest of your answers will require you to think more creatively, since “rock”, “paper”, and “scissors” have already been used. Your goal is to see how many times you can beat your answers before you eventually invent one that can’t be topped.
Despite the numerous innovations that modern AI can bring to gaming, it would be a mistake to treat AI as a complete replacement for conventional development tools.
AI-native games overcome the limited dialogue and storytelling of other titles, but these aspects aren’t always flaws. In role-playing games like Disco Elysium and Baldur’s Gate 3, your limited dialogue choices lead to well-written and carefully crafted conversations that serve a clear purpose, whether it’s to inform you about the world, provide insight into a specific character’s motivations, or bring levity to an intense scene. This deliberate pacing and purposeful storytelling elevated these games to critically acclaimed masterpieces.
There is still value in AI-native games, but they can’t take the place of story-driven titles or other conventional experiences. Instead, AI-native games are deserving of their own genre. While some of these titles take inspiration from genres, they’re uniquely defined by the limitless gameplay that other types of video games can’t replicate.
AI-native games can be a benefit to the overall gaming industry if more developers learn the right lessons from them. As games like Suck Up! and 1001 Nights prove, LLMs and other forms of modern AI work best when they’re used as a framework for innovative new ideas.
Source link