THE VIDEO GAME INDUSTRY'S IMPACT ON SOCIETY
GAMES AS AN ENVIRONMENT
It comes to no one’s surprise that video games’ graphics have come a long way. Now more than ever, with the new generation of consoles on the way and the announcement of a fifth version of the famous game engine Unreal Engine, the environments built for video games can rival even the most beautiful picture or movie. Nevertheless, while for some AI researchers it is necessary that the game contains “life-like” graphics and physics (gravity, wind etc.), for other researchers the fact that video games can provide a malleable virtual environment is very useful for the development of Artificial Intelligence.
If we refer back to the aforementioned Julian Togelius, Associate Professor at the New York University Tandon School of Engineering, Video games have become the best manner of testing the intelligence of Artificial Intelligence. Indeed, he describes: “Video games are a great way of training AI algorithms because they are designed to give human minds gradual progression into harder and harder challenges.” (How video games, 2018)
Credits: Wallpaper Abyss
In fact, Adrien Gaidon, a computer scientist at Xerox Research Center Europe, and his team used Unity, a widely used game development engine for 3D video games, to create scenes to help train deep-learning algorithms. His inspiration for this decision came after he was tricked to believe that a trailer for the graphically-impressive Assassin’s Creed game was for a movie (see picture above), while in reality the environments were all digitally rendered. Thus, he understood that if a human can think that these environments were real, so would an Artificial Intelligence. At the moment, the testing is ongoing to compare the effectiveness of training algorithms with virtual environments against those trained by real ones. This is an important project to prove new ways Artificial Intelligence for any purpose in society can be trained using solely video games.
Credits: Brambrai
Similarly, Artur Filipowicz and his team at Princeton University have trained their AI algorithm to recognize stop signs we can find on roads by using the virtual environment built in the latest Grand Theft Auto. Indeed, this game was built with real situations and real driving laws in mind. The game has a plethora of stop signs in a variety of situations that are ideal for training AI. Filipowicz had to tweak the game so it could be played by another computer program instead of only by a human. The goal is to use virtual gaming structures to improve AI’s ability to self-teach, which could then be applied to applications in the real world, such as advancements in GPS development and progress in smart and sustainable cities. (Marr, 2018)
Another project on the development of Artificial Intelligence which saw the day solely thanks to the unique possibilities offered by videogames is Project Malmo, by Microsoft. This project is an open-source platform that allows AI experiments within the world of Minecraft; it aims at supporting fundamental research into AI. Minecraft offers its users endless opportunities for simple and more complex tasks, it is thusly a very useful platform on which to work for developers. This project could also help researchers understand the building blocks of intelligence and how AI interacts with its world, learn how it makes sense of its surroundings and develops an internal representation of an environment. While Project Malmo sets its expectations relatively high regarding the understanding of intelligence itself, the unique possibilities offered by the game Minecraft are undeniably appealing for AI researchers.
Credits: Grand Theft Auto Fandom
According to a paper published in 2018 by machine learning researchers from the University of Freiburg in Germany, an algorithm using evolutionary strategies exploited an undiscovered bug in the video game Q*bert (1982) to rack up almost infinite points. The evolutionary strategy method of developing AI consists of a competition of different AIs toward a similar goal in which only the most successful algorithm survives and becomes the new building block toward other AIs for the next round of the competition. The most surprising fact about that discovery is that no one, including the original game developer, had spotted the bug in the 36 years the game has been out. The algorithm’s exploration behavior and its willingness to try new things to maximize points in the long term helped it find the glitch. Its advantage, compared to a human or the other competing algorithms, is that it was programmed to maximize the points gained in the game, on the long term. It shows that Artificial Intelligence developed in order to achieve a goal and not simply follow a pre-defined pattern can act in ways not previously conceived by the developers. This advancement in the understanding of Artificial Intelligence’s behavior could not have been achieved without the use of video games. This proves once again the value of video games for the general development and research of Artificial Intelligence. (How video games, 2018)
Credits: EA
On a different yet quite similar note, video games have played an important role in developing Artificial Intelligence as it helped analyze human emotions and relations which are much more difficult than simply analyzing and using data. In fact, the gaming industry is now trying to tackle how to create emotional AI that would resemble actual human relationships. A great example could be the character Elizabeth in the game Bioshock Infinite (see aside) which is a great example of a smart, helpful and emotional partner which adapts to situations and has compelling writing and mannerism. Emotional AI would provide a better player experience, but these advancements would be beneficial outside the gaming industry as well. In this instance, video games can provide the “testing ground” to vet more elaborate AI and see if emotional and relatable AI can be developed and please the general public. These advancements would have further applications potentially to other industries from banking, scientific research and more. It would take normally for other industries a lot of time to acquire the necessary data, develop, and test this relatable Artificial Intelligence. (Marr, 2019) However, the gaming industry has the edge as the players are the testers which give off a lot of data and many people in this industry are working toward this same goal with different methods.
The previously mentioned project developed by Electronic Arts to perfect Reinforcement Learning on Artificial Intelligence is also another project on developing Artificial Intelligence by using video games. Indeed, researchers from EA’s Search for Extraordinary Experiences Division (SEED) set out to expand on previous work using games to train neural networks (how the machine “think” and decide to act). In order to test their novel training system, they built a custom 3D environment for a machine to navigate. Their goal is to improve mastering imperfect information games in which human ingenuity and unique thought patterns make each play session unique. “Playing a game like Titanfall 2 taps into a skill set that’s closer to playing poker than a board game where all players are given the same information”. Their goal on the long term is to make the Artificial Intelligence be able to work with undefined variables to achieve a pre-defined goal consistently by exploiting their 3D environment to their advantage and adapt to uncounted for situations, as would a human player would do in games such as Battlefield 1. (Greene, 2018)
Credits: Games&Geeks
As we can thusly see, video games constitute an opportunity to develop and test Artificial Intelligence techniques, and an opportunity to transfer the technology to other industries; an opportunity apparently similar to that of the symbolic games for the archaic forms of Artificial Intelligence of in the 1980s and 1990s. (Miikkulainen, 2006)