When you think of AI in gaming - you probably think of that dumb NPC character you had to escort across the map that kept running into walls and getting itself killed. You also probably think about those in-game dialog trees you'd navigate or the bosses you'd figure out how to defeat after your tenth attempt when you finally stopped being stubborn and went back to town to get that health potion.

Traditionally, this is what AI in games has represented. Pseudo-"intelligence" (if you can call it that) that made your world feel more alive and dynamic. And, while these systems might just be powered by hundreds of if/else statements written by a programmer on their 3rd Red Bull vs. the large language models (LLMs) of today - you'd still be right in thinking it hasn't changed.

For the most part, it's because it hasn't. Not yet. While many see their email clients, productivity tools, SEO-optimizing agencies, and pitch decks get the "✨" icon that has now universally been associated with AI, it feels like gaming has been idle to adapt many of these new tools. 

But it hasn't - and personally, it's one industry that I'm most excited to see the future of these LLMs truly excel in. 

There are a few ways AI-driven tools are starting to take root in the world of game development that I think are incredibly exciting and worth highlighting - as they benefit both game devs and players alike. 

When you think of random generation or procedural generation typically - many think of Minecraft, roguelike games, and any other game type that utilizes randomly changing elements to create unique and fresh play experiences for their players. Traditionally, this has been the best option for wanting dynamic content like this. You'd loosely set some rules/criteria that the generator would follow, and it would handle the rest. Those rooms get a treasure chest. That one some spikes. This one a door. That one a boss. Random, yes - but predictable. A dedicated player learns the patterns, the rules, and it begins to lose the "magic". 

Not with AI-driven procedural generation. When properly integrated, your random elements and the rules that define them blur their lines. Play experiences don't just feel unique, but they spark moments of true discovery, surprise, and joy - even for the game developer. And these rules can be intelligently swapped, on the fly, to make things more challenging, rewarding, forgiving, or straight up confusing. 

Speaking of dynamic content...

Leveraging on-the-fly content generation to create personalized assets and experiences for a player. The Finals, a new popular FPS game leverages this quite well. In this game-show-style battle arena, you have psychotic narrators that announce your team's every move. But, surprise - those narrators use AI voices. This gives them the ability to react to changes in gameplay that the developers may never have been able to predict with traditional cookie-cutter rules. Dynamic content generation in this case refers to sound, but it could just as easily apply to world elements, music, models, or many other game assets.

Dynamic Content can also refer to the game systems players interact with daily. Imagine if your difficulty curve or daily quest log adapted to how you play, what you like to do, and where you are in the game in ways developers didn't pre-program. Don't like to fight other players? Well your gameplay trends would report that - and maybe instead, you just get more collection-oriented quests.

Predictable logic-tree NPC dialog boxes are a thing of the past. Super Intelligent NPCs know your entire save file - what you've done, who you've helped, items you have - everything. They can be told to act in certain ways or react to events that otherwise would have taken tons of work to predict. And best of all, when combined with Dynamic Assets & Content - their voices could still sound true to life. No different then any other living player. Super Intelligent NPCs would bring a new layer of life to your game world and any player's experience.

The flip side of this is also creating better enemies then ever. If an enemy knows the weapons you have and your strategies, it can better understand how to counter you - give you a true challenge. That Orc you're coming up to may also choose to provide you with a much harder time because of all the others it knew you killed not long ago.

*Yawn* who cares? And I get it - if you don't always work with data, this one isn't exciting. But  - it very much is. AI-assisted analytic tools would help game designers come to conclusions, discover issues, and help drive better content faster. This is one of those where if it's working well, the player would never know it's being used, but game devs certainly will.

This can also stray into areas like automated testing and game balancing - which studios traditionally spend thousands of man-hours on to help build a better player experience. Being able to take a step back from these areas, even just a bit, would free up way more time for work on creativity and content.

Further, when an AI can fully understand not just what players are doing but how a game is meant to be played, it can start making judgments on outliers - such as those that are cheating, exploiting bugs, or doing other actions that go against the spirit of the game being built. All without relying on privacy-invading cheat detection tools that run on the device.

The best way to get more people playing your game? Make sure people can play it! Many games aren't localized; if they are, language availability will vary significantly in quality. Using LLMs, you can ensure that your game is available to as wide of an audience as possible, without having to do all the complex admin work every time you add a new UI screen or collectible item. 

The exciting part is almost all the technologies mentioned above are possible - today - with current LLM tech. Games and studios are already beginning to use AI in this way to their advantage. The accessibility of these features and how advanced they get will just become more accessible to reach as costs come down and more advanced models are developed. 

We're at the beginning of a genuinely new era. For players, it means games that are living, breathing worlds that react dynamically to how they play to create unique and personalized experiences that otherwise would never be possible. 

For developers, it means gaining access to a toolbelt that allows them to spend more time being creative and creating content players love and leveraging these tools to ship this content faster.

How will you utilize it?

- Devon