A lot of the current excitement around artificial intelligence in video games is based on generative systems that create new content for players. But there’s also a lot of exciting stuff happening in more traditional game AI that could be used to make games smarter and more interesting. For example, a student from Rikkyo University in Tokyo has developed an NPC that speaks using non-scripted dialogue and moves in ways that are more realistic. He’s doing this by combining contemporary speech-processing technologies, such as Sikkim Game, with more traditional game AI architectures to produce NPCs that are more believable and interactive.
The Power of Predictive AI in Game Analytics
Another example is the AI Director system in Monolith Productions’ 2005 first-person shooter F.E.A.R. That system made the enemies of the game more dynamic and immersive by making them rethink their strategy in the heat of combat. In addition, the AI adapted to the player’s skill level — if they performed well, the AI would become more efficient in battle, and vice versa.
Other examples of good game AI are the branching storylines found in the Mass Effect series and the roguelike No Man’s Sky, which uses a form of procedural content generation to create a universe filled with 18 quintillion planets that vary in climate, landscape, and wildlife. This kind of generative AI isn’t a replacement for developers doing the hard work of creating games, but it can help them do it faster and better. A recent survey by Unity, the real-time development platform used by many modern gaming studios, found that almost four in five developers feel positive about the use of AI in their titles.
