As part of the latest batch of the so-called ‘disruptive technologies,’ machine learning (ML) and its parent, artificial intelligence (AI), are revolutionizing technologies in every major industry–from advertising and healthcare to agriculture as well as space exploration. AI’s contributions to them are well-documented as these industries affect more people around the world.
The gaming industry, however, is one area where AI and its subsets, machine learning and deep learning (DL), are proving to be an excellent match. AI researchers also find the industry interesting and challenging. In gaming, they get to implement ML components in creative and entertaining ways. Moreover, the gaming industry is getting bigger than ever. In 2020, the global gaming market was worth USD162.3 billion.
For game developers, technologies that help them execute their vision for the next blockbuster game are always worth it. With the possibilities that machine learning presents, this technology is the perfect tool to change the gaming landscape.
Here are a few examples why:
A Tool for Game Development
Machine learning, which is a subset of artificial intelligence, refers to the ability of a system to learn, adapt, and improve from experience, with no specific programming. In the gaming industry, machine learning in the past few years became widespread after significant advances in the processing speed of GPUs (graphical processing units).
Video games present players with many instances of decision-making. For instance, in-game opponents should be making smart choices and logically reacting to the player’s moves to be challenging, but not too challenging that players are turned off. To do this, AI collects a massive amount of data for processing, which is used to improve the gaming experience. Machine learning helps a lot in customizing user experience based on in-game decisions or choices by the user. You can learn more here.
This enabled game developers to develop games that feature complex interactions and a more realistic-looking environment. The huge amount of data needed to create games like these also needed appropriate hardware and software. With AI, its subsets–machine learning and deep learning–and advanced GPUs as well as powerful processors, games are more immersive and more challenging.
Improved GPUs also made it possible for game developers to make realistic and incredible game environments, like The Witcher 3 and the more recent Red Dead Redemption 2. Normally, creating such a lush and immersive game world needs scores of designers working hundreds of hours. With ML, the human designers’ load is lessened. In-game effects, like weather, can be taken care of by the AI, which translates into faster visual rendering.
Smarter NPCs (Non-Playable Characters)
Typically, NPCs and in-game opponents are predictable. Their actions depend more on scripts written specifically for them. With smarter and reactive NPCs, a player would be facing unpredictable enemies who’d be more reactive with the player’s actions. Depending on the player’s level, the difficulty is adjusted so that combat is more challenging. Mindless button-mashing is replaced with a more nuanced strategy and tactics.
Machine learning makes NPCs learn complicated actions and movements from human players. Single-player games, which imply AI opponents, are more compelling than before. As for online games, in 2019, OpenAI Five, a team of AI bots, beat a world champion human team in DOTA 2.
It was the first time in esports gaming history that an AI team won over a champion human team. It was significant because the match required teamwork and collaboration. For AI and ML, it was a huge deal. Even Bill Gates recognized it as an important milestone for artificial intelligence.
Makes Developing Games Faster
Writing behavior scripts for hundreds of NPCs needs hard coding (inserting data into the source code itself). But providing these NPCs with AI would considerably reduce the coders’ workload. Creating a realistic world for games is another incredibly intensive labor.
Games can take years to develop, especially those big-budgeted triple-A games. For instance, Cyberpunk 2077 took seven years before being released, bugs and all. It’s a great game, with memorable characters, a highly interactive game world, and hundreds of NPCs. But making such a game is extremely labor-intensive. Designing the game’s environment and writing AI algorithms alone must have taken hundreds, if not thousands, of hours.
Procedural generation, the method of designing game worlds, uses algorithms in making distinctive worlds. But this method isn’t as polished as designers want it to be. AI and ML, however, are transforming procedurally generated environments into something that could compare to AAA game worlds.
Testing newly minted games before releasing them is usually the purview of human game testers, who spend a significant amount of time on each level. AI and its child, machine learning, can run the tests on various game levels. Machine learning can run through game levels faster than any human game tester, too. Along the way, ML can also provide an analysis of each game level.
The billion-dollar gaming industry is starting to reap the benefits of ML, which is proving to be highly useful not only in game development but also in improving user experience. This was made possible when GPUs and processors improved tremendously, enabling the processing of a huge amount of data needed by AI and ML.
Machine learning made it possible to automate different processes that made game development faster and more efficient.