The games industry is an innovation driver from which many economic and social areas benefit. The interaction between man and machine is the basis of every game, and developers of computer and video games try to get the best gaming experience out of the available hardware.
As a result, they lead the development and application of virtual reality and realistic physics and graphics simulations and thus regularly push the boundaries of what is technically feasible. The components that have already proven themselves in games are in high demand from other industries. Game mechanisms and technologies serve as a basis for developing complex simulations that are used for training and experimental purposes in industry or medicine. The development and application of artificial intelligence (AI) also have a long game industry tradition. Computer and video games pose interesting and complex problems for the AI to play a game successfully. AI has also been used in games for decades and is increasingly used in game development.
AI In Games
AI has always played an important role in computer and video games and their development. Increasingly complex rule sets are used in games so that characters and the game world react credibly to the player’s behavior. Mostly intelligent behavior is imitated and simulated. Above all, AI serves to present the game world credibly: For example, the computer opponent adapts his strategy according to the playing style and the skills of the human player, thereby simulating reactions that would be expected from a human player. Game worlds are getting bigger and more complex, which further increases the demands on AI.
However, games can also promote difficult thinking within certain parameters and are a valuable test platform for AI systems. IBM’s Deep Blue chess computer, which defeated world champion chess player Gary Kasparov over 20 years ago, is perhaps the most famous example. Since then, there have been many other applications: In 2012, two AI-controlled game bots managed to pass the “Games Turing test” in the game Unreal Tournament 2004. The “Games Turing test” is a variant of the Turing test in which game viewers must correctly guess whether an observed game behavior in a game corresponds to that of a human or an AI-controlled bot. E-Sports professionals recently competed against an AI from the Google subsidiary Deep Mind in the real-time strategy game StarCraft. This was trained in cooperation with the game manufacturer Blizzard based on a database of games played by humans. In the game, the human was still the winner, unlike in a bot from open-AI, a research organization backed by entrepreneur Elon Musk, among others. The bot uses machine learning to play Dote 2. The bot can play games against itself every day for 180 years, learning successful ways of playing, and has already been able to defeat professionals. E-Sports professionals can also benefit by adopting the bot’s successful strategies by using the technologies in training. The findings from the games are also used in research to improve the algorithm, which can also be used in other areas. The fictional state of San Andreas from the game Grand Theft Auto V offers the test field for autonomous driving in another application. The self-driving neural network learns how to deal with traffic in the game world.
AI Technologies In Games Development
However, games are used as a training object in developing AI and the games themselves. AI is also playing an increasingly important role in the development process. A procedural content generation has been used since the 1980s to replace units and items on the playing surface every time the game starts. Methods of automated generation are used, which are intended to make content appear more organic and “handmade” through machine learning and AI. This should streamline the development process and improve the credibility and immersion of the game world.
However, neural networks and machine learning are being developed to display game content such as emotional expressions and animations even more realistically. The goal of the technology initiative SEED(Search for Extraordinary Experiences Division) of the game manufacturer Electronic Arts (EA), for example, is to develop an interactive virtual human. The games company Ubisoft also uses AI to process movement data captured using motion capture methods to clean up the raw data and look natural. It takes a human about four hours to process a given chunk of data, while an AI can achieve almost the same result in about four minutes.
The Next Steps Of Games & AI
In the future, the use of AI in connection with games will increase in all areas mentioned. Other technologies will also play a greater role, which can enable major leaps in development. For example, there are first applications to outsource the amounts of data generated in the game via a cloud connection and evaluate the huge amounts of information in high-performance data centers. This allows customized game content to be generated with the help of AI. Games will also continue to drive hardware development forward. Graphics cards, which are primarily developed for the most realistic possible representation of game worlds, offer the best conditions for applying deep learning and AI thanks to the speed of their chips.
All in all, games will continue to drive AI research forward and play a key role in further development. To be successful, research funding in the games sector must be expanded. Games technologies are an important building block to becoming one of the world’s top locations for AI. It is also imperative to expand the supply of high-speed Internet. Low-latency gigabit networks must finally be available across the board so that the resulting applications can also be used and all citizens can benefit from new application options.