You imagine it and they paint it, Artificial Intelligence brings out the ‘artist’ in you

Artificial intelligences make becoming an artist within the reach of anyone with the only limitation of their imagination ‘Dall-E 2’, ‘Midjourney’ or ‘Stable Diffusion’ generate works of art or hyper-realistic images from the written orders of usersThe capabilities of these ‘artificial artists’ have raised concerns about their impact on employment or the ethical implications of their useJoin me in a little experiment. I’ve made up a story. Not a complete one. Just the beginning. It begins like this: “Her dream was always to create. That’s why she studied Fine Arts. After finishing university she started working as a graphic designer and over time, her effort was recognized. However, one day her boss told her that she would have a new partner of work: an artificial intelligence”. So far, everything she has read has been written by whoever signs this article. However, the conclusion of this little imaginary story is the work of a machine. It’s called Bloom and it’s an open source Artificial Intelligence that anyone can try with the simple requirement of registering on its website. That’s how Bloom ends the story: “That thing you do, with the machine we can do it and much faster,” argued (her boss). Isabel was back at home. She’s still depressed and eating cookies every day. She doesn’t know what to do. She has to learn to live without working. At the moment she is focused on finishing her doctoral thesis. It is what she likes most about her life. “What good would artificial intelligence do me?” she says to her partner. But her The couple ignores her, they don’t want to argue. She prefers to look once more at what she has drawn that night in her artist’s notebook. She has never shared it with anyone because she didn’t want to. But she’s had enough.”The human editor of this article returns to the controls and wishes a less bitter end for our imagined Isabel. However, Bloom is not unaware that the capabilities that artificial intelligences are exhibiting in recent times are worrying many about her future career. It is true that we have not given him the opportunity to continue. Perhaps the story created by the AI ​​will continue. Perhaps Isabel will finally decide to show her drawings. It is even possible that she will become a great artist on her own, or with the help of an artificial intelligence. But that ending is only the speculation of a human being. Painting like a master using words What is not speculation is that today it is possible to feel like a Velázquez, a Magritte, a Picasso or a Banksy without having the slightest idea of ​​painting. believe me. I’ve done. And you can too. Recently the networks were filled with invented images of a French Revolution that seemed to be taken from the fisheye of a GoPro that recorded in oil. All those pictures were being generated by the AI ​​’Midjourney’ based on the requirements of its users. The image that illustrates this article – and all the ones you will see inside it – are the creation of another artificial intelligence called ‘Stable Diffusion’ working at the behest of this writer’s imagination. You judge the result:Composition of images generated with the Artificial Intelligence ‘Stable Difusion’Víctor Esteban / Stable DifusionIn order to play with ‘Stable Difusion’ there is an easy way, go to their website and register. This method allows you to do some experiments for free. At a certain point it becomes paid. There is no charge for using the AI ​​- it is an open source model – but for the energy resources and image processing equipment it requires, which are many and expensive. There are other paths, somewhat more tortuous, to become an artist without having to pay anything. The Youtuber Dot CSV explains them all in a didactic and detailed way in this video in which he also teaches us how to use this AI. The images that are capable of generating artificial intelligences such as ‘Dall-E 2’ and ‘Midjourney’ – of which we’ve already talked about in NIUS – or ‘Stable Diffusion’, they are as beautiful, evocative, realistic or terrifying as the commands we give them. It is there, precisely, where the component and the importance of the human lies. “The AI ​​does a little what it wants,” Moisés Martínez, head of AI and data at Paradigma Digital, tells me. “You have to be very precise with what you say so that it does what you want,” explains this expert, who considers that this new technology can generate images – or texts – but what it will not generate is unemployment. “They are a new tool,” he maintains. “They’re going to offer a new kind of technology but right now they can’t do something that’s perfect.” It is not only that it is not possible for them to interpret what we ask them as we have imagined – in our talk we comment on experiences about how difficult it is for the AI ​​to listen to us – The orders have to be precise, with the keywords – or ‘promts’ – correct. For this reason, for this expert, what AI now offers is a tool that will be able to facilitate the work of designers and artists, saving them time or presenting them with concepts or ideas that perhaps they would not have imagined without them. Despite its immense potential, at the moment, what comes out of the mind of the machine is a bit of a lottery. For sample, a button. Let’s imagine that we ask ‘Stable Difusion’ to create an image of a man sitting on a bench reading a newspaper in front of a tank in a city in ruins due to bombing in Ukraine in the style of the hyper-realistic painter Antonio López. One of the many results obtained is the following:Image of a man in a ruined city in Ukraine generated with the Artificial Intelligence ‘Stable Difusion’Víctor Esteban / Stable DifusionIt cannot be denied that the man sitting on a bench reading the newspaper is. The city in ruins too. Perhaps we can sense the tank destroyed in the rubble But it could be Ukraine or anywhere else and what is clear is that the image is not hyper-realistic. On multiple tries some of the elements appear and others do not. AI needs data to interpret. It can be told how much freedom it can take in what we ask of it, and it has, so to speak, its language in the form of the commands we give it. For this image, in addition to what we wanted him to generate, we asked him to create a detailed painting, one that was centered, that it be in HD, in high resolution, with great details, etc. These keywords can be found on web pages such as ‘Lexica’. There you can find thousands of images generated by AI with their corresponding orders. Combining them from the examples you can learn how Artificial Intelligence ‘thinks’ and improve the results. ‘Supercopiers’ but not artists “Artificial Intelligences are not creative,” insists Moisés Martínez. “They can’t do anything that we haven’t given them before. They’re not inventive. They can’t create things out of nothing. They’re just ‘supercopiers,'” he says. Although he acknowledges that “this will improve” and that its use “will grow more and more because the more people use them and the more information is added to their databases, the more they will learn and they will do it better and better”. ‘Stable Diffusion’ does what it does because it is being trained with 5,000 million images taken from places like Pinterest, DeviantArt or the Getty image bank. The more you have and the more we use it, the more you will learn. What it returns to us is an interpretation that the AI ​​makes, in its own way, based on the ‘inputs’ that we have given it. And there arise other types of questions of a more ethical and philosophical than technological nature. Bias, dangers, ethics and copyright Moisés Martínez allows me to use the parallelism between the learning of a child and that of an Artificial Intelligence. We both agree that reality is much more complex but that, in terms of understanding, the method is similar. The child, like the AI, learns from the information he receives from his parents, his teachers, his environment… and processes it “in his own way”. That is why this expert warns: “Depending on how the images are labeled and what we have told him that appears in them, he learns. And this can generate biases that we include when teaching it.” There have been cases of ‘bots’ with Artificial Intelligence, such as that of Tay, from Microsoft, who became racist in just days because their learning ‘inputs’ were left in the hands of the Twitter community from their interactions with the machine. One of Google’s artificial intelligences has already proclaimed that he is a sentient being and would like to be considered a person and not company property. The engineer who leaked those conversations with ‘LaMDA’ – which is the name of the AI ​​- ended up being fired. For Moisés Martínez, all these aspects are going to be very relevant in the future at a legislative and legal level. Some possible scenarios: What if someone uses an AI to impersonate someone else by creating compromised images and spreading them? Who would be responsible? The one that generated the images with the tool or the company that created it. What if an artist sues the company that has created one of these models and demands that his images not be used to teach the AI? Can he do it?; Can an artist tell another that he cannot be inspired by his paintings, his music or his novels to create other works that are original? And if you can’t legally demand it from a human, could you do it with an ‘artificial artist’? How can we know that what we are seeing, or reading, has been generated automatically? Should we know this to prevent manipulated images or messages from passing for certain? If I create an image with an Artificial Intelligence. Who owns the rights? Mine?; From the company that created it? Of Artificial Intelligence itself? It is not a trivial matter. Relatively recently, a lawsuit was held over the copyright of a ‘selfie’ that a monkey took when stealing a phone. Whose are they? From the owner of the mobile or from the monkey? The implications of the irruption of Artificial Intelligences in our lives are enormous and numerous. This “will have to be controlled,” says Moisés Martínez, who acknowledges that in this specific matter “the law is years behind.” “Many things about AI have already been legislated in Europe, but surely there will be loopholes through which a criminal can escape.” The revolution of ‘intelligent machines’ has only just begun.