Take a fresh look at your lifestyle.

An AI analyzed works from the National Gallery in London, suggests that a millionaire painting bought in 1980 is fake

The National Gallery in London has been home to one of the museum’s proudest paintings since 1980: ‘Samson and Delilah’ by Peter Paul Rubens. A painting that he acquired for 2.5 million pounds at the time, worth almost 8 million euros in current values. But there is a problem, an artificial intelligence ensures that the work has a 91% probability of being false.


For a long time critics have suggested the play could be fake. It does not coincide entirely in time and especially in style with the works of the Flemish artist Rubens. Now there is one more proof in favor of the falsehood of the work: the study made by an artificial intelligence that has analyzed in great detail this ‘Samson and Delilah’ and hundreds of other works by the artist. Artificial intelligence artists and now also art critics.

91.78% probability that the painting is fake

Art Recognition is one Swiss company that uses artificial intelligence to recognize and authenticate the veracity of works of art. Since it was created, it has analyzed more than 400 works and works permanently with some institutions such as the University of Tilburg in the Netherlands. She was recently put to analyze the controversial painting by Rubens (or not his).

Screenshot 2

Example of a certificate of authenticity generated by Art Recognition.

When analyzing a work to determine if it is true or false the neural network first learns the characteristics of the artist. To do this, he trains with numerous real works by the artist, which he divides into smaller pieces and analyzes in detail. Once the training is completed, the image to be analyzed is offered, based on the characteristics that the artist knows, will assign a percentage of similarity to the painting.

The report on this occasion was conclusive: ‘Samson and Delilah’, with a probability of 91.78%, is not an original work by Rubens. In order to draw such a conclusion, artificial intelligence analyzed 148 other paintings by Peter Paul Rubens to have a database where it could be compared. Determined that the color style used does not match the artist’s own palette. The owners of the artificial intelligence decided to analyze the painting on multiple occasions, in all of them the AI ​​assured with more than 90% of probability that it was not a painting by Rubens.

What art hides and technology uncovers: science is allowing us to discover secrets of great works (and rescue them)

But the work It is not the first time that you have received these criticisms. Over the years numerous art critics have shown that there are serious doubts about its authenticity. The toes of Samson’s right foot are clipped for example, whereas they are shown in an original engraving by the artist. Likewise, different in composition from two other contemporary copies made by Rubens himself. The National Gallery painting was authenticated by Ludwig Burchard, an expert who, after his death in 1960, was found to have false certificates of authenticity to obtain commissions.

Using artificial intelligence to determine the veracity of a work of art has its positive and negative points. On the one hand, it can be argued that you will not be able to appreciate sublime details of a work of art, details typical of being appreciated by a human eye. On the other hand, however, the AI ​​can analyze thousands of factors in moments and compare with a huge database without missing any pixels. All of it without any subjectivity or bias in favor or oppossing.

Via | The Guardian