AI Generates New Doom Levels That Are Interesting to Play

Looking at how games can be engaging and complex, it’s evident that game makers spend quite sometimes creating the software. That’s why for example a company like Nintendo will announce, say, an oncoming level in one of their games six months before launch, like the iPhone’s Nintendo Mario Kart Tour. As in, brains have to go to work to ensure acceptability; however, the AI technology has crept in to help in the process of video-game making.

Just to inform you, content creation in the gaming industry has never been a walk in the park, which makes it extremely expensive to manage projects or even keep already developed games afloat for long. However, even without AI’s assistance, some video game franchises have managed to survive the odds for decades, like the Doom series. The series hit screens back in the 1990s, has sold 10 million plus copies and the game has not stopped gaining popularity, even with the current Facebook, Tweeter, Pinterest and etc. generation.

A Glance at the Doom Series

Source: i.imgur

The game puts you, the player in a marine spaceship and gives you a gun to defend yourself. So you are a first-person shooter battling zombies and demons of different strengths and destructive power. If you delay the adversaries have various options to end your life, including eating you alive. In the overview the game is among the first ones to run on 3-D graphics, making it possible to accommodate multiple players in machines operated on MS-DOS.

Ideally, we can accredit this game’s resilience (how it has survived this long) to the fact that it allows users to create their custom game levels, which in a way empowered research. That’s why we are here: as in, with that provision experts have managed to develop a deep-learning software that is capable to create levels in the Doom series, which gamers can find interesting and engaging. Just to make it clear, the algorithm, in this case, is designed to create extensions to the game, not really to create codes for new games –but of cause, if this is possible we can conclude that we are headed to a place where machine intelligence will be the new game-maker.

The New AI-Driven Level-Building Model 

Source: cdn

The new page we are entering has been pioneered by Edoardo Giacomello and his friends, who are based in Italy, at Politecnico di Milano. These experts dreamed of ways they can automate the process of building compelling Doom levels and as they put it, “AI provides an awesome technique to do that, and this could redefine how game content is generated.

What is amazing is that the researchers did deploy a direct approach as opposed to most machine-learning methods. Training their system, they harvested around 1000 doom stages from Video Game Level Corpus, a source platform that harbors Doom and Doom 2 official levels, and other repositories containing 8000 plus levels build by the gaming community.

The procedure also involved helping the AI system to master important features in the images of all levels, which included key objects, wall heights, obstacles, walkable areas and so on. Mind you that all these features needed to be identified by their actual size, the space they occupy and so on.

The information was then used on a deep-learning tech know as the generative adversarial network. And after 36,000 rehearsals the AI system managed to produce impressive levels that people would be interested to explore. “Well the levels produced by our algorithm may not be as perfect as you’d expect but the point is, the approach is highly promising,” said Giacomello.

A Feel On the Newly Developed Doom Levels

Source: technologyreview

In relation to their good experience in the game, the team took rides and confirmed that the levels are entertaining, and would capture any gamer because they bring out Doom features (the constricted tunnels, walls) and the rest in a super amazing version.

As Gicomello explains, levels are significant to all games because they are what makes the player want to explore more in any particular platform. As in ideally, people easily ditch off games that hit the end after they manage to conquer the last adventure or test.

That is why game developers have no otherwise but to ensure their game remains relevant with new interesting levels, but you know what, that can be super expensive if you have to retain experts to keep updating the game – and the reason why automation is seen as a rescuer by game developers. Well, that sounds a bit hash mentioning because it seems to confirm AI as a threat to jobs in the industry.

But here comes the question, automating the process of game development might soon become a reality, so what happens to the experts who’ve been doing that? No worries, still, humans will be key in designing games because as it is, machine learning or artificial intelligence has proved weak in handling creative tasks, and there is no debating that game making is highly dependent on creativity.