The neural network was “copied” from Velasquez – Kommersant FM

The neural network was “copied” from Velasquez – Kommersant FM

[ad_1]

Artificial intelligence has restored a lost painting by Diego Velazquez. Spanish restoration artist Fernando Sanchez Castillo was able to recreate the painting “The Expulsion of the Moriscos” using a neural network, El Pais newspaper reports. The canvas was painted in 1627 under the law of King Philip IV. The plot was supposed to reflect an important event for the authorities: the expulsion from Spain of 300 thousand Muslims who converted to Catholicism, who were considered possible traitors. It was this painting that secured Velázquez’s place as the main court artist, but in 1734 it was destroyed in a fire.

To restore the canvas, the surviving sketch and detailed descriptions of contemporaries, which were processed by artificial intelligence, were used. However, the artistic value of such a work is low, says Kommersant FM art columnist Dmitry Butkevich: “Since this painting is considered to have disappeared since the first third of the 18th century, its value for the history of world art is practically insignificant. Since it was not physically there, how could it be inserted into some anthologies? It is as if it did not exist, since in the 18th century there was no recording of works of art, and we cannot understand what we have lost.

Of course, any discovery or resumption of works by such a great artist as Diego Velazquez is a sensation. But here, of course, there is a big question about the authenticity of this work.

Indeed, with regard to paintings that, as they say in professional language, have been restored, experts already have doubts about whether they can be considered original and authentic. And here, in this case, it’s just a completely restored work, apparently, starting from the canvas and ending with the paints.

Probably, artificial intelligence can completely copy the artist’s work style, but the question is, who will be the author of this work? Will it say “Diego Velazquez” or “Artificial Intelligence Robot #276” at the bottom of the canvas? We still have to investigate this topic, since there have been no such precedents yet.

At the same time, there is no doubt that such technologies will develop. They are only to help restorers and art historians, since they can restore once lost works. But, from my point of view, this work in particular is more professional, or something, that is, something that is necessary for studying the work of Velazquez. But, in my opinion, it still does not have independent artistic value.”

Services for generating images using AI are also available to ordinary Internet users. Among the most popular are DALL-E, Midjourney and Kandinsky. At the same time, the method of operation of these services is almost no different from the technology that was used to recreate Velasquez’s painting, explained AI specialist Igor Shcherbakov:

“Conceptually, this is the technology that is now used in all diffusion models. In fact, the sketch and text description are vectorized, and based on the resulting vector, which contains information about what was in the sketch and what was reflected in the text description, the model reconstructs the image itself. This, in fact, is no different from what you and I can do by going to some Midjourney, uploading your photo and asking us to complete something there. The mechanism is exactly the same.

At the same time, I don’t think that this work is really what it was originally. It will be the same as its contemporaries saw it: whatever details they noted on it will be reproduced.

If the author depicted something additional that was not reflected in the description and on the sketch, that did not leave traces, then it will, accordingly, be lost.”

The artist Paula Garcia also took part in the process of restoring Velazquez’s painting. According to media reports, she is exploring the possibilities of using artificial intelligence in modern sculpture.


Everything is clear with us – Telegram channel “Kommersant FM”.

Ulyana Gorelova

[ad_2]

Source link