Interacting with AI
A conversation involving AI will likely bring to mind movies such as I, Robot or other futuristic narratives that center around human-like robots. While many now understand that AI has a largely utilitarian and business-based functionality, it’s still difficult to quantify what, exactly, AI does for us. The general public knows that AI gathers and analyzes incomprehensible amounts of data and that many emerging technologies utilize AI, though they may not realize that, in the realm of video games, AI programs are prolific through NPCs (non-player characters) that users play against and interact with.
In popular video games such as Counter Strike and Civilization, many times users won’t even realize that they’re intermingling with an AI program. These programs are powering characters that players don’t often pay too much attention to, like creeping enemies or supporting role characters such as merchants or animals in an RPG game. Given that a game like Counter Strike has almost one million monthly players on Steam, the number of users interacting with AI in video games on a daily basis is staggering.
AI has also crept into other sectors of video game entertainment. In 2015, the most popular US casino games were slot machines, which 48% of visitors played, followed by blackjack, poker, and roulette respectively. However, this data accounts for live, in-person play. Today, it’s much more common for someone to play roulette, poker, blackjack, and slots online directly from their mobile device. Deals are competitive, and apps are constantly improving play for users based on technology like ‘reinforcement learning models’.
But what does the general public think about these interfaces with AI through NPCs? After all, these aren’t the types of ‘robots’ that are commonly associated with everyday life like those in the aforementioned movies like I, Robot. Instead, the AI programs discussed above are designed to utilize data to make decisions based on sets of patterns, much of which is predetermined by game developers. A hesitant answer is that most of the general public may not have a big problem interacting with AI on a daily basis through some of their favorite games because there isn’t a physical component to this communication—as in, there isn’t a robot sitting behind a screen at their side.
AI vs. Human Players
While some AI programs are redefining what we thought AI could do, game-based AI are still leading the way in terms of popularity, even perfecting some of humanity’s most ancient games. Chess, for instance, began in the 7th century in India and is one of the most prolific board games in the world. With 64 squares composing a square board and 16 pieces that perform various functions and are bound by various rules, human players will take a lifetime to accrue insights into patterns and strategies to employ, as well as knowing what to expect from their opponents.
However, an AI program can handle massive inputs of data and, through the reinforcement learning models mentioned above, can utilize millions of segments of data to predict an opponent’s strategy and respond within a number of seconds. There’s no sitting around idly in a park for these bots. AI programs will learn to play games like chess by playing themselves millions of times with little more than basic game instructions at the start. As aforementioned, a human counterpart will spend a lifetime attempting to accrue such critical insights that can only be gained from experiential play.
Returning to the realm of video games, one of the world’s most popular games, StarCraft, which sees about 2 million users per month, is now interacting directly with an AI program on more than an NPC basis. In fact, the program plays openly as a character under its own account and is better than 99.8% of players (round of applause for the human grandmaster players in that .2% bracket). Unlike the programs developed to play chess, the StarCraft AI program had a bit of help by human opponents who agreed to play against the bot in order to help it develop its skills. Still, the feat is impressive, and indicates that reinforcement learning processes can be applied to programs outside the realm of video and puzzle games.
