For the covers of issue 4 of Ecos de Soto, it was impossible for us to have distinguished architects design the cover, so we had to resort to generative Artificial Intelligence (AIg), specifically Midjourney, to design it according to their particular architectural styles. The results were very varied. Some were very curious, others were taken from fantasies, others were more macabre, but some closely resembled their possible authors.
The revolution of AI
One of the great advances in design has been the ability to incorporate AI into the creative process, which has generated much controversy in recent months. Many see it as a panacea, others as a chimera, and it is understandable because throughout history the same thing has happened before. It happened with the Industrial Revolution, where it was thought that machines would replace human labor (and it didn’t happen). It happened with television, where it was thought that it would kill the radio (and it didn’t happen).
The unknown, the new, generates a certain distrust and suspicion. There are many people who think that AI will end many jobs, and that will be true, but it will also create many other jobs. So we will have to do as always. Adapt or die. Know how to recycle and know how to coexist with a technology that has come to stay. And for this very reason, we decided to create the cover of our issue 4 with AI (and also because being inside a prison we could not contact architects to design the cover for us).
Imagine a prison
The first thing we do when we want to work with an AIg is to generate a prompt, which would be like the instructions we want to give the program to create what we have in mind. And to start, just use the command /imagine.
Chema Parsanz talks about AI: You learn from it at the same time that it learns from you. It is not simple, it is rebellious, it does what it wants, it is like a baby… you have to find the “recipe” for the machine to understand what you really want, but the baby will grow up.”
When you start imagining a prompt, you have to speak as if you were speaking to a young child who does not know the world around them. You can write that you want a prison with high red walls, but the AI won’t know if the red is for the walls or for the prison in general.
The prompt is a kind of collage of concepts.
Building a prison.
A fundamental step in understanding how AIg works is knowing that it is not really imagining, but rather studying and cross-referencing data that exists on the internet. So, when you want it to draw an apple, what it does is see what an apple is for us among all the photographs, what shape it has, what texture, what colors, and through that data it will make a reinterpretation that fits the parameters that we give it. What it does is process the sentence and compare it with the universe of concepts it has. In the end, when you want to write a prompt, it’s like a recipe. You have to indicate all the ingredients you want so that it shows you what you more or less have in mind. But there will also be cases – and this is very interesting – where we may not know what we want and we can let the AI have freedom to show us results that can inspire us or continue working.
Machines work well with numbers, but they work very poorly with words. So you have to start by first studying how the information is represented, then how you relate the data, and then how you learn, there is a temporal theme, quantity of data, but also the time in which you are extracting patterns and diverse information.
Reaching effective understanding with the machine is a matter of learning, trial and error, tricks, and your artistic and cultural background. Every little bit helps.
Getting to work.
Let’s now start with the process of creating the cover now that the concepts in which AIg operates have been explained a bit.
First, there was a brainstorming session about prisons. As mentioned before, for the AI to generate a prison, it has to compare it with photos of prisons. And that’s where the difficulty lay a bit, as there aren’t many photos of prisons, so to rely on design, according to the experience gained from the tests, it was primarily based on prisons from movies or video games, something that was quite far from a real prison.
We could find images where the emphasis was on walls, bars, and especially barbed wire. If we put barbed wire on a wall, it is very likely to pass as a prison. And then there is the other part, where, using fantasy, some models of utopian prisons were designed.
As seen, the results were far from reality. And since it would be very difficult for it to accurately represent a prison, it was decided that distinguished architects would design a prison. With that concept, it could be justified that it doesn’t look like a prison, but it could be one. Just like the City of Arts and Sciences in Valencia has many buildings that don’t really look like what they are.
Now we just had to come up with a very simple prompt, in which it imagines a prison in the style of Santiago Calatrava, Alejandro Arenava, and Tadao Ando.
The results were quite curious, although functionally I doubt that there will ever be a prison with those forms. But who knows… the future is ever-changing and uncertain.
