In 1945, a fire He also mentioned three controversial paintings by Gustav Klimt. Sent in 1894 to the University of Vienna, the “Faculty Paintings” – as it became known – were no different from any previous work of the Austrian embassy. As soon as he showed them, the critics were in an uproar because of their impressive departure from the beauty of the moment. Teachers at the university immediately rejected the offer, and Klimt resigned. Soon, the works were organized into groups. During World War II, he was kept in a palace in northern Vienna for safekeeping, but the house was burned to the ground, and the paintings were probably taken away. All that is left today are black and white pictures and writing from that time on. Still looking at them.
Well, not just art. Franz Smola, a Klimt specialist, and Emil Wallner, a machine learning researcher, spent six months combining their expertise to revive the work that Klimt lost. It has been a daunting task, one that began with black and white photographs and then combined with the artist’s ingenuity and ingenuity, attempting to recreate the lost images. The result is what Smola and Wallner are showing me — and they are amazed by the amazing technicolor graphics that AI has created.
Let’s be clear on one thing: No one is saying that AI is bringing back Klimt’s original functions. “It’s not a way to re-create real colors, and to re-create images,” Smola quickly realizes. “The process of photography is just a myth from real works.” What the machine learning is doing is giving a glimpse of something that is believed to have been lost for years.
Smola and Wallner find this to be fun, but not everyone helps AI fill these voids. The idea of machine learning to repair lost or damaged works, such as Faculty Paintings, is controversial. Ben Fino-Radin, an art specialist, says: “My main concern is to learn how to use a machine learning system in place,” says Ben Fino-Radin, an art specialist, because of the proliferation of moral and ethical issues that have led to people should not suffer too much. sick machine learning section. ”
To be sure, the use of technology to revive the works of human art brings with it many questions. Even if there is a perfect AI that can detect the colors or brushstrokes Klimt can use, no algorithm can make a valid target. Controversy over this has been going on for centuries. Back in 1936, before Klimt’s photographs were destroyed, author Walter Benjamin criticized the duplication of machines, even in photography, stating that “even the most masterpiece of the art is one thing: its existence in space and in space, its uniqueness. the existence of the living. ” This, Benjamin wrote in Art Work in the Generation of Machine Manufacturing, and what he called work “aura. ” For many art lovers, the idea that computers reproduce the invisible thing is impossible, if not impossible.
And yet, there is much to learn from what AI can do. Faculty paintings were crucial to Klimt’s growth as an artist, an important bridge between his oldest paintings and, later, the most successful works. But the appearance of these colors is still unknown. That’s what Smola and Wellner are trying to solve. Its project, created through Google Arts and Culture, was not about the perfect release; it was about giving an idea of what was missing.
To do this, Wallner designed and taught the three-part process. First, the algorithm was fed thousands of artistic images from the Google Arts and Culture database. This enabled him to understand the material, the artwork, and its design. Later, it was taught mainly in Klimt art. “This creates a bias for his colors and goals at the moment,” explains Wallner. And in the end, AI was fed color light to other areas of the picture. But without the symbols, where did the symbols come from? Even scholar Klimt Smola was surprised by the detail at which the writings of the time were revealed. Because the paintings were considered obscene and strange, critics often commented on them at length, down to the artist’s choice of genres, he says. Simon Rein, the project manager for the project, said: “You could say it was a wonderful experience. “The fact that the artwork caused the insult and rejection makes it possible for us to return it because there was so much writing. And those types of data points, if fed into an algorithm, create the most accurate type of how the diagrams look at the time.”
The key to this accuracy lies in combining Smola algorithms with Smola technology. His research showed that Klimt’s work at the time was characterized by strong and consistent performance. Reading the paintings that existed from before and after Faculty Paintings provided insight into the paintings and paintings that recurred in his work at the time. Even the amazing experiences of Smola and Wallner are confirmed by history. When Klimt first exhibited his paintings, critics noticed his use of red paint which, at the time, was in short supply in the artist’s palette. Koma The Three Ages of the Woman, the artist soon after Faculty Art, boldly uses red, one Smola believes it was the same color that caused the uproar when he first appeared in Faculty Art. The record of the time also enhances the color and cries of the wonderfully green sky in another Faculty Painting. Combining these notes with Smola’s knowledge of the Klimt leaf palette, fed into the algorithm, is what made it one of the first amazing images from AI.