The creativity of machines is in question. Is an artificial intelligence capable of creating art? Could synthetic artists exist in the future? Arthur I. Miller, emeritus professor of history and philosophy of science at University College London (UCL) and author of 'The Artist in the Machine: The World of AI-Powered Creativity' has participated in the Future Intelligence Fest, in Fundación Telefónica, to talk about these issues.
Nowadays, there are movie trailers being generated with AI, there are algorithms that shape our tastes, and we see more and more tools being used to create art. “There are several algorithms that show interesting trends. One of them, for example, is DeepDream, which allows an artificial neural network to create images hitherto unimaginable”, he explains. An artificial neural network is a machine whose internal structure is wired in the same way as the human brain.
Thus, a current problem of artificial intelligence is to ensure that the neural layers inside the machine, which are called “hidden layers”, know how to reason, because these types of networks form the core of the Internet of Things and the Internet of Things. driving autonomous cars. “DeepDream – he continued – was invented to investigate how these hidden layers work. It works like this: you take an artificial neural network, you show it millions of images, from ImageNet, for example, and then you give it a JPEG. The machine will try to recognize the image of that JPEG”.
So, what you do in this case is interrupt the analysis and ask the machine what it sees. That's where DeepDream comes in. “It turns out that what the machine sees has, curiously, little to do with the original image. It's surreal. It is the world seen with the eyes of a machine”.
Networks that “dream”
Arthur I. Miller has also indicated that antagonistic generative networks, also called GANs, allow an artificial neural network to dream and imagine, allowing it to create a life of its own . “A GAN works as follows: it consists of two networks, a generative network that generates noise, that is, images out of nothing, and a discriminatory network. The generative network sends the images to the discriminatory network. The discriminatory network evaluates whether they are real images or not, guided by a database of thousands of faces taken from the Internet.
Soon, the generative network starts generating images: not out of nothing, but from the images returned by the discriminatory network, and ends up generating faces that have nothing to do with any person on this planet. “That is why I say that GANs allow machines to dream, imagine, and begin to generate their own internal life: the internal life of the images returned by the discriminatory network”, he highlighted.
How to connect a “Magic Mouse” to a mac: Hey youtube so in this video i will show you how to connect a “ Magic Mo... http://bit.ly/huS7E1
— Mac Blog Fri Apr 22 10:12:49 +0000 2011
Arthur I. Miller has given other examples: in music there are many devices that allow musicians and AI to collaborate and tap into each other's creativity. And in literature there is the famous GPT3, Generative Pre-trained Transformer. “The 3 indicates that it is the third in a series of devices. The GPT3 is the most powerful word processor that exists today. It generates almost human texts. He makes mistakes that need to be corrected, but he does it quite well. Collaboration is the present”.
An Einstein of Artificial Intelligence
Expert Arthur I. Miller believes we are at a turning point in the creative industry. And he believes that it is the result of combining artificial neural networks with machine learning, with machines that learn without being programmed for it. "Who knows what the future holds? Perhaps an Einstein of artificial intelligence will come along and come up with a hitherto unthinkable structure that will allow machines to work faster than now and be truly creative."
But what he finds most fascinating of all is “that the machines show flashes of creativity. I find it very interesting. They do it by using algorithms like DeepDream, or a GAN, or with AlphaGo."
AlphaGo is a program that works like an artificial neural network to play go. In 2016 he defeated a very prestigious go master, and other go masters. That meant that a machine had deciphered go as a game. "It was fantastic, it was something huge."
“When we create something that goes beyond the material we have, we call it creativity. Why not equally recognize the creativity of machines? Why that rejection? Is it scary that machines are creative? Yes that's how it is. I have seen it. The machines are not going to replace Bach or Mozart, just as Beethoven did not replace them. The machines go further. They create their own music, they create music that we are not yet capable of imagining”, has asserted Arthur I. Miller.
The future of creative machines
And what will the future look like? “We are already seeing synthetic artists with autonomous production. You can put a webcam on a painting bot, for example, and send it out into the street. You will look around and see something you want to paint, because it reminds you of an image you have in your database. It is something that has already been done. Something like this supposes, in a very primitive way, a certain level of will and free will”.
Even so, Arthur I. Miller has insisted that a machine will not be fully creative until it has feelings. “There is talk of “synthetic artists” in reference to machines or robots. And it is that, if that "thing" is capable of producing art, why not call it an artist? Although I already say that he will not be an artist until he has no feelings and until not a single human intervenes in the process. It is somewhat the same with artificial intelligence, which some consider an oxymoron. There is consciousness and intelligence. Both can occur in other life forms, and are not called "artificial" simply because they are not human. With strong artificial intelligence there will be only one intelligence, not one artificial and one natural."
In addition, he has stated that it is said that machines cannot feel because they do not “live” in the world. “They don't experience real feelings, like falling in love with someone or feeling one with nature. But someday, in the near future, machines will master a language, let's say English, and they will be able to read the net thoroughly and acquire more knowledge than we can in a lifetime. And so they will convince themselves, and they will convince us, that they have lived essential experiences for creativity: being inspired, loving, hating and so on. We then wire them up with complex sensor systems, relative mechanisms, and communication pathways that will allow them to develop a set of feelings that mimic ours."