The topic of using artificial intelligence in the process of creating and improving video games is a topic of discussion. From time to time, large companies surprise us with new ideas in this field, thus showing us what possibilities the future may bring. Many of these ideas sounded like complete fantasy two or three years ago.
Relatively recently, Electronic Arts (FIFA, Star Wars Jedi, Battlefield) presented its idea for the use of AI in the game development process, presenting its new system for creating games and their rules using simple text commands, without using a line of programming code. Of course, this American company is not alone in its plans and aspirations, because other giants also have similar ambitions.
Players have also put their hands on technologies based on artificial intelligence and from time to time they can surprise us with their use. One example of such use appeared relatively recently on the Polish part of YouTube. We’re talking specifically about generating new graphics based on specific games.
Neural networks in games? What could their potential be?
The process of teaching an artificial intelligence to play Pokemon Fire Red
The topic of using neural networks in various fields of science and entertainment seems almost inexhaustible. For the purposes of this text, however, we will limit the use of this technology only to this second branch of human culture.
YouTube is full of all kinds of materials that show how the use of neural networks can help create games, design them from scratch or simply learn how to complete them by performing subsequent actions sequentially. At one time, materials that showed how such specially prepared programs passed titles such as Super Mario World, the first installments of Pokemon and more were a phenomenon.
All this showed how great the possibilities of such technologies are if you give them enough time to learn. After all, with the proper determination of the programmer, you can create a program that could break records without any major problems speedrunning (trying to complete) the most demanding games as quickly as possible, or develop new strategies for completing them.
What if we used the capabilities of artificial intelligence to “refresh” slightly old games?
Grand Theft Auto 4 is like real life
Material presenting the capabilities of neural networks in image processing
The DubStepZz YouTube channel published a material in mid-October that shows what Grand Theft Auto 4 (commonly known as GTA 4) looks like after minor corrections implemented by neural networks. JAt first glance it is clear that the changes and differences are huge and no one will probably have any doubts about it.
In some shots you can see that the graphics seem to be getting closer to what is called “photorealism”. However, you don’t need to have eagle eyes to see that this solution – although impressive – is still far from good.
In the video we see quite a lot of artifacts, strangely rendered objects or models that absolutely do not fit the surroundings. In addition, the issue of the face, proper body proportions or more dynamic scenes where artificial intelligence simply cannot cope yet remains problematic.
It is also worth taking into account that the graphics were not rendered in real time, but rather superimposed on the recording. This additionally makes us realize that there is still a long way to the real application of such technologies in this particular industry.
Neural networks are a chance to better refresh old games
AI creating video games and their remakes is not such an unrealistic and distant vision?
Even though this technology is still in its infancy, it does not prevent us from making certain visions of the future. After all, the use of refined and well-trained neural networks may be an opportunity for many old titles whose graphics have simply become outdated. It can also be a solid basis for the visualization of text-based games, which are now associated with the ancient times of gaming.
For now, however, this is just a curiosity and should be approached as such. It may also be a foretaste of what awaits us in the future, provided, of course, there are companies that are willing to be interested in this topic and develop such technology. Everything will probably depend on the money and potential profits that this solution will be able to generate.