Advanced artificial intelligence (AI) systems possess the capacity to discern subtle visual intricacies within imagery that often elude human observation. In the year 2023, a sophisticated AI neural network detected an anomaly pertaining to a facial depiction within a renowned Raphael masterpiece, ultimately indicating it was not the original artist’s work.
The specific visage under scrutiny is that of Saint Joseph, situated in the upper-left quadrant of the artwork commonly referred to as the Madonna della Rosa, which translates to Madonna of the Rose.
Art historians have, in fact, long engaged in scholarly discourse concerning the painting’s authenticity as a genuine Raphael creation. Historical records from 1667 indicate the piece was housed in a Spanish monastery, where it remained until its acquisition by the Spanish national art museum in 1857.
However, the documented history, or provenance, of the Madonna of the Rose prior to 1667 remains entirely undocumented.
Although the painting has consistently been attributed to Raphael in its entirety for the duration of its presence in Spain, connoisseurs have observed that the color palette and the technical execution of Saint Joseph’s portrayal distinguish it from the other figures, suggesting the potential involvement of other artists within Raphael’s studio.
“The attribution to Raphael’s workshop gradually gained acceptance later, with specific designation to his pupil Giulio Romano and possibly Gianfrancesco Penni,” commented Howell Edwards, a chemist affiliated with the University of Bradford in the UK, during a 2023 interview. he stated.
“Within Spain, the original attribution has never been called into question.”
While the definitive establishment of an artwork’s provenance necessitates a comprehensive array of evidence, a novel analytical methodology employing an AI algorithm has aligned with the perspective of those who contend that at least certain brushstrokes were executed by an artist other than Raphael.
A brief video below provides a synopsis of the investigative team’s findings:
Researchers from both the United Kingdom and the United States have engineered a bespoke analytical algorithm, meticulously trained on extant works definitively authenticated as originating from the Italian master’s hand.
“Through the application of deep feature analysis, we utilized images of verified Raphael paintings to instruct the computer in recognizing his distinctive style with exceptional granularity, encompassing brushwork, color choices, shading techniques, and every facet of artistic rendition,” explained Hassan Ugail, a mathematician and computer scientist from the University of Bradford, in a 2023 statement. he elaborated.
“The computational system perceives visual data at a significantly deeper level than the human eye, extending to a microscopic resolution.”
Machine learning processes conventionally require training on an extensive dataset of examples, a condition that is not always met when considering the complete oeuvre of a single artist.
In this particular instance, the research group adapted a pre-existing framework developed by Microsoft, designated as ResNet50, and integrated it with a conventional machine learning technique known as a Support Vector Machine.
This analytical approach has previously demonstrated a remarkable 98 percent accuracy rate in the identification of Raphael’s artworks. Typically, the system is trained on complete compositions; however, in this investigation, the team also directed it to analyze individual facial elements.
While the figures of the Madonna, the Child, and Saint John were consistently identified as creations attributable to Raphael, this was not the case for Saint Joseph. The researchers highlight that in prior discussions regarding the painting’s authenticity, Saint Joseph’s face was frequently considered to be rendered with less skill than the other figures present in the composition.
“When we analyzed the della Rosa painting as a unified whole, the outcomes were inconclusive,” stated Ugail.
“Consequently, we proceeded to examine the individual components, and while the remainder of the artwork was confirmed as Raphael’s, Joseph’s facial representation emerged as most probably not his creation.”
It is plausible that Giulio Romano, one of Raphael’s esteemed pupils, may have been responsible for the execution of the fourth facial representation, though this remains a speculative attribution. This development represents another instance where contemporary technological advancements are instrumental in uncovering the mysteries embedded within classical artworks, with AI playing a pivotal role in this specific case.
Art historians estimate that the Madonna della Rosa was rendered on canvas between the years 1518 and 1520. It was during the mid-19th century that art critics began to entertain the possibility that Raphael might not have been solely responsible for the entire composition.
These conjectures have now been substantiated with a high degree of certainty; however, the research team behind the study emphasizes that this AI technology is intended to augment the capabilities of art experts in the future, rather than supersede their roles.
“This is not a scenario where AI is displacing human employment,” asserted Ugail.
“The process of authenticating a piece of art involves consideration of numerous factors, including its historical provenance, the composition of pigments used, the physical condition of the work, and other related elements.
“Nevertheless, computational tools of this nature can serve as a valuable auxiliary instrument within this broader process.”
