Artificial Intelligence, Video Games, and the Future of Both
Artificial intelligence has recently become a hot topic in the tech world. People are finding new and exciting ways AI can improve tons of industries, ranging from healthcare to how you search the internet.
Gaming is one industry that is incredibly interested in AI, both to help with playing a game and to create new possibilities for us to play games. While the technology is still in its infancy, AI is progressing very quickly. Already, it?s having an impact on video games, and it could go on to do some really amazing things to make our games more immersive and fun.
What is AI?
The term ?artificial intelligence? is already a common term used in gaming to reference how the gaming programming acts and reacts to the player. This programming is really just a complex set of algorithms and programmed behaviors of how the game is to behave with the player, especially with enemies and NPCs. The game doesn?t change its programming to perform better, it just reacts in a specific way to what the player does.
True AI is when a machine learns how to perform a task and changes its behaviors based on the data it is presented with and by making mistakes. The program develops and changes over time, constantly improving.
For the purpose of this article, when I reference ?AI,? I mean true machine learning ? not the behaviors and patterns of a program told to act a specific way, but an actual machine that learns and develops on its own.
The History of AI Playing Video Games
No longer are video games just for humans. Many businesses are using video games as a means to test their AI programs and see what kind of things they can accomplish. It all started with simpler games, like old Atari and Nintendo games. For example, an AI program from the tech company Deepmind began training in old Atari games like Space Invaders and Breakout. The first couple of games, the AI would fail spectacularly, but from there, it learned very quickly the rules of the game and created its own strategies. This system played each game available for two weeks, and after that period of time, could beat 75 percent of scores achieved by an average human on the same game.
Another AI program called MarI/O, learned how to play an entire level of Super Mario World with absolutely no prior information or knowledge. It didn?t even know that it needed to go right in order to win. Then, it only took the program 34 tries to learn the level and win.
The true potential of AI though was shown when an AI machine defeated a pro DOTA 2 player. For the AI, named OpenAI, learning and mastering a game like DOTA 2 is huge. AIs have beaten masters in other board games like chess and Go, but strategy games like DOTA are much more complex. While the game was a bit simplified, with it being a 1v1 match instead of the standard 5v5, it?s still a pretty complicated game. The craziest part about this: it only took the AI two weeks to reach this level of skill.
AI as Gameplay Elements
As a person plays a game, they begin to notice patterns in the gameplay that they can capitalize on. Enemies move in a certain pattern, items spawn in specific locations, and other elements in the game can all be exploited for the player?s benefit.
If a game were to introduce AI as a part of design, this could open up an all new level of game design. Imagine playing Pokemon where your rival is controlled by an AI instead of just being a set NPC. You might abuse a specific system or utilize a specific strategy, and in the next fight, the AI would be prepared to stop that. This, in turn would push the player to come up with a new solution. The AI would become a real rival and not just be another obstacle in the game.
AI could also be used to create games catered to individual people. One rising genre in gaming is roguelike/rogue-lite games, which involve a randomized level generation element to the game. Each time you play the game, it?s a unique experience. AI could still keep that fresh and random element, but cater the game design to either challenge or favor the player?s strategies. It would be like having a dungeon master guiding your game instead of leaving everything up to chance.
Playing Against AI
Thanks to the AI beating out a professional competitive DOTA 2 player, it?s clear that AI can become skilled enough to go toe to toe with the best of gamers. This could have severe implications for multiplayer games in every genre.
One major problem in gaming is the skill difference between playing single player and online multiplayer. Many people feel confident after playing a single player campaign, go to online modes, and get completely destroyed. AI could help ease this transition by providing multiplayer-styled games but with AI-controlled opponents. The AI could progress and learn with the player and then, once the game figured they are good enough, put them into live matches with other humans.
AI controlled opponents could also be used when matchmaking for a game fails. One frustrating part about online gaming is that people could drop out of a match at any time. If the game is team based, this could put a team at a large disadvantage until another player joins. Having an AI sub in or just take the place of the abandoning player could prevent any major disadvantage and try to keep the match running.
One major concern people have though is whether people could use AI to cheat. The more an AI plays, the better it gets. There is really no skill ceiling for them, as they can continually improve and play the game constantly. At a certain point, an AI could beat any human opponent easily. If somebody could employ an AI to play their game for them, would this qualify as cheating? If so, what are game companies going to do to prevent this?
Another concern is for games with in-game currencies and economies, like World of Warcraft. People could create AI bots capable of farming that in-game currency and then sell it to players, or use it for themselves.
Tech in Gaming Helps the World
A big reason why video games is being used to test AI is because it acts as a simulation that humans can also do. We can directly measure how well people perform in a game versus the machine. There is no risk for an AI to go wild and cause problems, like when Microsoft released that AI program onto Twitter.
It?s actually pretty common for technology to get its start in video gaming, then evolve into other fields. Like how VR and AR are being used to train football athletes. Originally, people thought that VR would only be good for video games, but now the potential is endless.
AI is capable of doing many amazing things that humans can?t. AI can crunch massive amounts of information fast, letting them find patterns and solutions. We live in an age of data, but interpreting it is very difficult ? something AI can help us with. For example, AI could analyze big data to help prevent and control global epidemics.
It?s very important, though, that we test out these functions thoroughly before we put our trust in AI. It?s still a new technology, and we don?t know what could happen.That?s why we need to be sure to test out the capabilities of AI in simulations like video games and see what happens. Just like in the movie War Games, where the computer had to learn a lesson through a game, we can similarly teach AI moral lessons.