We have seen many applications of AI in industry sectors but in regards to the entertainment niche, Artificial Intelligence continues to diversify day by day operations. Video Games are an essential part of almost every person out there, after a tiring day of work, we all relax with a video game whether it is on our handheld device or dedicated gaming rig.

The goal of Artificial Intelligence in video games is not limited to competitive gameplay but rather it intends to create the most enjoyable AI for players to compete with. The inclusion of AI in games dates back to the early 90s when Wolfstein 3D came out but the AI non-player character (NPC) was at the beginning level at that time. It implemented simple concepts such as to evade the player’s attacks and attack when the appropriate time came, all in an enclosed finite state machine. 

 

So far the AI in games we play nowadays can be anticipated. If you keep the above state machine in view, you can understand that there are only a few states where the AI will switch between. And these states essentially can be memorized. However, that is not always the case and an example is Dark Souls 3 in which the creators have programmed the AI to anticipate human mistakes. Essentially, this makes the game very unforgiving.

 

So essentially, there are two basic parts of an AI in games as per Julian Togelius, an associate professor at New York University which are pathfinding and finite state machines. All NPCs are in particular states and they navigate through them as you progress in the game, hence making decisions and introducing some sort of automation in the game. 

 

With that being said, we should also understand that there is a difference between AI designed to interact with human players and the one which aims to play a game with human-level intellect. The most popular is the IBM Deepblue which bested world champion and grandmaster Kasparov in a game of Chess:

https://youtu.be/NJarxpYyoFI

While the human mind can remember a lot of patterns, it is no match for the memory of a machine and in case of that match above, the DeepBlue was capable of calculating millions of possible combinations in a second and memorizing the techniques of former opponents. Essentially, that served as a basis for the computer to get smarter and smarter until the time came that it was able to beat the grandmaster.

 

AI is also capable of learning real-world skills through video games. For instance, a team deathmatch where there are an AI commander and AI group members, the AI commander can offer an analysis of different combinations if a certain AI member had gone through a different route or chosen a different method of defeating the opponent. This is reinforcement learning and forms the basis of improvement for the NPCs.

 

There is a lot that can be said about AI in video games but right now they are infancy. Due to one reason that the human mind can’t keep up but they certainly add a whole new level to video gaming which has never been seen before, provided we implement it right.