How Artificial Intelligence are used in Video Games
They Just Don’t Want to Think Like Us
What is the first thing you think of, when you hear the someone say Artificial Intelligence or the word AI? A group of robots taking over the world? A super intelligent robot or computer that is superior to mankind? The terminator? Well, it emphasizes the creation of machines that think and learn.
By definition, artificial Intelligence (AI) is the ability of a machine or program to think and learn. It’s idea revolves around building machines that act and learn from humans to be like humans.
The ability to learn and think should allow AI to mimic human thoughts and behaviors as well as to think rationally, ethically, and intelligently. That is the result of AI being capable of interacting with their environment and being able to act upon it in a human way.
With what the AI are capable of, would you think of the same think when you hear artificial intelligence used in video games? In video games, it’s used to determine the behavior of non-playable characters or NPCs. These guys are used for any reason at all, plot device, assistance, or just part of the game. The behavior of the NPC can range from simple actions to being an opponent in a chess program that are able to go up against human players.
The thing about how AI are used in video games is that the NPCs are not actually learning from the players who are playing. They are learning to act or react in a way depending on the scenario as compared to an AI that is learning to beat someone in a chess game. An example would be IBM’s Deep Blue supercomputer that was developed to play chess and defeated the grand champion Garry Kasparov in 1997.
So can a AI play a video game better than we can? Would they perform better than us in a way we expect or no? According to a bunch of researchers trying to teach an AI to play Sonic the Hedgehog as part of The OpenAI Retro Contest, the results are quite strange. The AI was told to increase its score through a gameplay. So that means collecting rings, defeating enemies while beating each level with the fastest time possible. While any other player would go through each level over and over to beat their time, the AI instead glitches through the game by going through the walls in one of the zones. In the end, the AI discovered shortcuts to reach its goal without understanding the concept of the game which was the opposite of what the researchers intended. This is why they can never think the same as we do.
