Artificial intelligence-based facial recognition technology has been making waves in the art world, particularly when it comes to the authentication of artworks. However, a recent case involving an AI model’s identification of a previously unattributed painting as the work of Raphael has raised questions about the reliability and limitations of this technology.
The painting in question, known as the Brécy Tondo, was identified by an AI model developed by Hassan Ugail of the University of Bradford as being from the hand of Raphael. This discovery has led to its display at the Cartwright Hall art gallery in the UK. However, another AI model created by Art Recognition found with an 85 percent probability that the Brécy Tondo was not painted by Raphael.
Carina Popovici, executive director of Art Recognition, expressed surprise at the contradictory results and voiced concerns about how this could impact the progress made in establishing AI as a primary method of authenticating art. She emphasized the importance of adhering to rigorous scientific standards to avoid criticism and potential consequences for the entire field.
Experts have weighed in on this debate, highlighting that while AI can be a valuable tool in art authentication, it will never fully replace traditional methods that rely on human judgment. Art historian Martin Kemp stated that there will always be a place for human interpretation and expertise when it comes to analyzing artworks. He also pointed out that AI may struggle with analyzing old masterpieces like Salvador Mundi due to their extensive damage or artists with varying styles throughout their careers.
Larry Silver, an art historian who worked on authenticating Virgin Flaget, acknowledged that AI can be useful but stressed that human intervention is necessary. He believes that AI will likely play a role in auction houses, museums, and galleries but should not be solely relied upon.
Popovici defended Art Recognition’s approach by stating that their network is trained on images of an artist’s artwork to recognize distinctive characteristics. However, she acknowledged that expert opinions are essential, especially when dealing with artists who worked collaboratively or had diverse styles.
The rise of AI-generated images also poses a challenge to art authentication. While AI can help identify stolen paintings, it can also be fooled by AI-generated fakes. The use of AI in identification and verification is still in its early stages, and claims about its effectiveness should be approached with caution.
The final outcome, while AI has the potential to aid in art authentication, it cannot replace human judgment and expertise. The limitations of this technology, such as difficulty analyzing damaged or varied artworks, mean that traditional methods will continue to play a crucial role in the field. As advancements in AI continue, it is important to maintain rigorous scientific standards and involve human interpreters to ensure accurate and reliable authentication processes.
