Artificial intelligence is currently capable of many things, from carrying out complex calculations to automating tasks like stock orders. However, AI can also be used creatively to generate new content - including art, music and text.

This has opened up all kinds of new opportunities for video game developers. Using this tech, it could be possible to speed up the production of video games and add new limitless features that could make games even more immersive and realistic than before. This post delves further into detail as to how generative AI may transform video games in the future.

Where are we currently with generative AI?

ChatGPT is the most famous AI tool right now. It takes the form of a chatbot in which users enter a question or demand. By accessing huge amounts of data collected across the internet, ChatGPT is able to answer pretty much any question and convey this answer in a human manner. Not only that, but ChatGPT can be asked to carry out various writing tasks, including writing short stories, poems, news articles, blog posts, job applications and even essays.

Generative AI is also being used to create art. The likes of Midjourney, Nightcafe and Dall-E are able to produce realistic photos or images in the style of famous artists using text prompts. Within photo editing, it’s possible to create an entirely new AI background or even add new details such as hats and new hairstyles - all with a simple text prompt.

We’ve also seen generative AI being used to impressively (and problematically) recreate the voices of famous people, from singers like Drake to presidents like Obama. Users can enter text or record their own voice, which can then be used to provide the impression that someone is speaking those exact words.
 

How will generative AI transform the future of video games?

All of this may sound fascinating - but how does it apply to video games? Well, there are many possible uses for this generative AI in video games. In fact, some of these applications have already been explored. Some of the different possible applications for generative AI in video games are listed below.
 

Detailed randomized worlds and levels

Creating an expansive universe within a game used to be challenging and time-consuming work. The Elder Scrolls IV: Oblivion is known for its many dungeons and Oblivion gates. Although there is a certain degree of randomization as to where the gates appear, each one of these dungeons and gates was handcrafted. That’s a huge amount of manually created content that is purely dedicated to side quests.

AI is certain to reduce the time spent creating such expansive universes by randomly generating worlds and levels. This is a feature that has been used in many games before, however the randomization is often fairly simple and based on a limited amount of map seeds. AI is likely to take this randomization to the next level by creating an infinite number of highly detailed worlds and levels.

Spelunky 2 is a game that already incorporated this AI into its gameplay. Although based on a certain number of rules and parameters to make them ‘playable’, every single one of its levels is randomly generated - making it new and exciting for players whenever they play the game.
 

Life-like limitless conversations with NPCs

Conversations with NPCs in games are typically scripted. There are usually a limited number of questions and responses you can give, and each one of these is recorded by a voice actor. Star Wars: The Old Republic was famously the biggest voiceover project for a video game - containing over 200,000 lines of dialogue played by 321 actors.

Using the limitless conversation scripting of generative AI tools like ChatGPT combined with AI voice replication, conversations with NPCs in the future could be much more varied and much cheaper to engineer. It may soon be possible to ask characters endless questions with all kinds of unique answers. And hours of dialogue won’t need to be recorded - instead only a few recordings of an actor’s voice will be needed and the rest will be able to be automatically generated in the same tone of voice.

NVIDEO ACE has already been producing this technology for video games developers to use, so it may not even be too long before we see it start being applied to games. The hard part will be making sure that each NPC stays within character and that no potentially offensive or politically biased responses are given (unless that’s part of the character).
 

Endless character customization

Many games already offer incredibly in-depth character customization. However, AI could step this up a gear - providing endless options for characters.

Using text prompts, it may be possible to describe how you want your character to look, allowing more rapid character customization. Already the technology for this exists via AI art generators, and it seems that new games will likely implement this pretty soon.

The technology even exists to create a video game character using a selfie. Combined with realistic graphics, it may be possible to play using an avatar that truly looks like you.

Of course, there are many issues that this technology could present such as being able to bully people by playing as them as well as playing as copyrighted characters from other games. Video games will have to address this when implementing AI.
 

Unique enemies with unique reactions

Current enemies in games are hand designed and are programmed to follow specific behaviours. This could be something that also changes in the future as AI develops.

Firstly, we could start to see enemies that are completely randomly generated in terms of what they look like. Space exploration game, No Man’s Sky, has already done this - introducing AI-generated creatures. Combined with the game’s AI-generated worlds, it allows for each planet and its inhabitants to be totally unique (and often more alien-looking as a result). More games may eventually adopt this technology with even more detailed graphics. Just imagine what this could mean for horror games!

The idea of creating AI enemies with unique reactions also has a lot of potential. Already, bots are being trained on the gameplay styles of online players. Using AI, it may be possible to create completely randomized NPC enemies that react just like online players - some better than others and some favouring different styles of attack and defence.
 

Personalized stories and cutscenes

AI will eventually be able to create detailed and unique videos based on text prompts. Combined with AI’s ability to write stories, this could result in the ability to create entire custom games complete with unique cutscenes using a simple text prompt.

Such games are likely to be a long way off. However, text-based games like AI Dungeon can offer a small taste of what is to come - such a game generates unique text-based adventure games using text prompts entered by the player. There are obvious copyright concerns that these types of games could produce.

Such technology also poses some difficult questions about the future of video game design - if an entire video game can be produced using a text prompt, will video game developers still be necessary? Already games like Roblox have proven the popularity of games built by players. If anyone can produce a detailed video game, studios will have to get a lot more creative to set themselves apart.