What is DALL-E 2, and why an AI could forever revolutionize the way video games are developed

What is DALL-E 2

Addressing the topic of artificial intelligences, British physicist Stephen Hawking once went so far as to say that "the success in creating advanced AI would be the greatest event in the history of mankind. Unfortunately, however, it could also be the last. ", referring to the fact that the greatest danger to the survival of our species on the planet would not be to be found in global warming or a nuclear holocaust, but rather in the invention of an artificial intelligence so advanced that it rebels against our control, just like in the films with Schwarzenegger, Terminator.

Nowadays the possibility of being wiped out by a computer may seem particularly remote, after all PCs still go into crisis when you try to install a new printer and robots of Boston Dynamics are famous more for their disastrous falls than for their practical applications. To tell the truth, the technology surrounding artificial intelligences has instead made a simply extraordinary evolutionary leap in the last year and a half, and if on the one hand the neural networks of natural language reworking are now indistinguishable from a human interlocutor, some organizations Researchers such as OpenAI have gone even further, developing AI capable of assigning meaning to images as well as words.

We are talking about DALL-E 2, an artificial intelligence that has become popular in recent months for his extraordinary ability to elaborate an image, be it realistic or purely artistic, starting from a simple textual description. An anime dedicated to Sonic doing financial trading? Or Mario engaged in the compilation of taxes? No matter how absurd or unlikely the input we feed it, the neural network metabolizes it to create a graphical representation of what we have requested out of thin air.

In this article we will explore the possible applications of DALL-E 2, especially in the field of video games, and we will try to understand what the role of a new generation of artificial intelligences, with almost unlimited potential, could be in the future , in the development of those playful experiences that our grandchildren will play. Provided they won't be slaves to the machines yet, of course.

From text to image, in a fraction of a second

"A frame from Sonic's brokerage anime" e " Mario who deals with taxes, realistic style, photo from 1998 ", from DALL-E 2 In reality, there are several artificial intelligences in addition to DALL-E 2 that offer the possibility of obtaining images from a text prompt, each with the own strengths and weaknesses. The one developed by OpenAI is currently in beta and can only be accessed by registering for a very long waiting list, but among all it is the most effective in processing realistic images inspired by real photographs and has no equal in the representation of faces. humans.

Then there is Midjourney, which is already available to anyone with a Discord account: you enter the AI ​​channel and you can immediately start experimenting, even if the attempts are limited and after having them out of stock, you are asked to subscribe. Midjourney is very skilled in creating elaborations with a strong artistic impact, has a predilection for theatricality and for this reason it would be perfect to help developers generate spectacular concept art from which to take inspiration.

"Last Pope on earth", Midjourney What makes artificial intelligences like DALL-E 2 and Midjourney extraordinary, however, is their ability to paint something that does not exist in the world, providing a personal, specific creative vision of an idea, a concept, just like a human artist would do. . There is nothing exceptional about processing a photo of a dog with a top hat and a cigar, but that is when the AI ​​is asked to derive a representation of hell, heaven, the boundaries of the universe or the center of a black hole, which the situation changes drastically. In that case the neural network is absorbing the text to give its own free interpretation and this is as amazing as it is vaguely disturbing.

Applications

Gordon Freeman, Mario and Spyro in alternative universes, Midjourney Obviously, DALL-E 2 and other neural networks similar to that of OpenAI lend themselves to an almost infinite number of applications, especially in the field of design and graphics, but also in that of marketing and in all those sectors where images are used to convey a message. Of course it is not mandatory to use them only for purely professional purposes, for example we can't wait to use them to create illustrations of the characters of a Dungeons & Dragons campaign or to embellish the desktop with unique wallpapers of their kind.

As far as video games are concerned, the applications of neural networks and artificial intelligence would be just as numerous and no, they would not suppress the creativity of the artists as many of you will be quick to state in the comments section below the article. Think for example of building a game world, and the amount of textures and other assets that are constantly repeated in an open world. An AI could process in a few moments the 2D surfaces of streets and buildings, or the details of the roofs, or even the facades of the shops, making each element different from the previous one without an artist spending hundreds of hours dealing with it in person. Starfield with over 1,000 celestial bodies featuring unique planetary surfaces? A productive nightmare for a software house, but a walk for artificial intelligence.

Set of settings created by Midjourney DALL-E 2 would also be able to integrate into the creative process that precedes the actual development, providing ideas for character design, or by providing concept art to the team that builds the levels. By supporting the developers in their various needs from time to time, an artificial intelligence could considerably streamline the production processes surrounding the creation of a video game, make the crunch less frequent and give a better work experience to all employees of the studio, without replacing none of them.

What awaits us

"Arthur Morgan in Night City", Midjourney Thinking about the future of the application of AI in video game development is quite simple, since as said by the Hawking himself in the speech on artificial intelligence quoted at the beginning, there are no real limits that can clip its evolution. If today they are already able to interpret a text to translate it into an image, there is no reason to doubt that in the near future they will move on to develop complete 3D models, building entire sections of the game map by themselves starting from the instructions of the developers. Something like this is already present today regarding the composition of virtual forests, but in some time we will be able to see the further diffusion of these systems, with the result that larger experiences will gradually become easier to develop despite their size.

If we were to look even further, by collecting Hawking's indications we could not skimp with the imagination and trace the contours of a distant but not very distant future, in which our grandchildren will be able to sit at the PC and ask their AI to realize them at the moment a video game that does not exist, mixing the typical elements of a souls-like vintage experience a la Miyazaki with the unforgettable aesthetics of Night City, featuring a very young Arthur Morgan with teardrop glasses and a robotic arm.

What if the Starfield planets were all the result of an AI? The British physicist was convinced that it was not a question of himself, but only a question of when. And we really like to believe him.

What do you think? What would you ask DALL-E 2 to accomplish? And what do you think will be the horizons reached by artificial intelligences in the next 50 years?

Have you noticed any errors?