Researchers from the University of Pittsburgh have tested algorithms for artificial intelligence in video games.
The researchers made several observations about testing artificial intelligence in video games, and the best decision-making strategies in the Multiplayer Online Battle Arena games.
The significance of artificial intelligence
An analysis by Forbes projected that revenue from artificial intelligence will increase from $1.62 billion in 2018 to $31.2 billion in 2025. Dr. Daniel Jiang, an assistant professor of industrial engineer at the University of Pittsburgh’s Swanson School of Engineering, said: “It is exciting to see the tremendous successes and progress made in recent years. To continue this trend, we are looking to develop more sophisticated methods for algorithms to learn strategies for optimal decision making.”
The University of Pittsburgh explain that Jiang designs algorithms for machine learning. The algorithms learn decision strategies in uncertain and complicated environments.
Testing these algorithms in simulated environments allows the researchers to learn from their errors and reinforce successful strategies. The simulations need to mirror the real world to achieve effective testing.
Dr. Jiang explains the problem with data which is not simulated to mirror the real world. He said: “Historical data can be a problem because people’s actions fix the consequences and don’t present alternative possibilities. In other words, it is difficult for an algorithm to ask the question ‘how would things be different if I chose door B instead of door A?’ In historical data, all we can see are the consequences of door A.”
Artificial intelligence in video games
Video games allow testing environments for complex decision making without the risk of an immature A.I. being fully in charge and are therefore a safe way to explore the mistakes of an algorithm.
Jiang added: “Video game designers aren’t building games with the goal to test models or simulations. They’re often designing games with a two-fold mission: to create environments that mimic the real world and to challenge players to make difficult decisions. These goals happen to align with what we are looking for as well. Also, games are much faster. In a few hours of real time, we can evaluate the results of hundreds of thousands of gameplay decisions.”
What is the Multiplayer Online Battle Arena?
Multiplayer Online Battle Arena, or MOBA, is a gaming genre. Some examples of MOBAs are League of Legends and Heroes of the Storm, games where the player controls a character who destroys opponents’ bases and protects their own. Jiang used MOBAs to test his algorithm for artificial intelligence in video games.
According to the University of Pittsburgh, an algorithm for artificial intelligence in video games has to overcome several challenges, including:
- Real-time decision making; and
- Long decision horizons (where the consequences of decisions are unknown until later).
The winning algorithm
Jiang explains that a Monte Carlo tree search is a method to generate data and feed it into a neural network. The player moves randomly through a video game, then the algorithm analyses the game results to give weight to successful actions. After multiple iterations of the game, the more successful actions prevail and the player is more successful at the game. Jiang commented: “Our research also gave some theoretical results to show that Monte Carlo tree search is an effective strategy for training an agent to succeed at making difficult decisions in real-time, even when operating in an uncertain world.”