Barthélé M. Jobert is obsessed with the 19th century, so he has a vast view of it. For him, it began intellectually in the 1760s and ran in the 1920s. A leading art historian in Paris and former president of the present Sorbonne University, he is particularly specialized in the works of Eugène Delacroix, the French romantic artist most famous for his 1830 painting, “Freedom to Lead People.” Now, Jobert greatly boosts his ability to use artificial intelligence and other 21st century technologies in his long-standing quest to explore Delacroix's art and solve the mysteries of its belongings.
This week, Schmidt Science, a nonprofit founded by former Google CEO Eric Schmidt and his wife Wendy Schmidt, is set to unveil a new grantmaking programme that will undertake the project for Jobert, known as Digital Delacroix. Jobert aims to digitally analyze much of Delacroix. His letters and magazines, murals he painted later in his career, modern newspaper descriptions of men and his works – and aims to cross-reference them to communicate them online for others to explore. A grant from Schmidt will help him gain more computing power, and even in France, he can increase his current team of six by hiring several researchers trained in both the unusual varieties of art history and AI.
For Schmidt Sciences, Digital Delacroix is the first of a grant of 10-15 people who will receive a total of $10 million to apply AI to humanities research. Expenses are expected to range from under $100,000 to $1.5 million. (Schmidt Sciences does not provide accurate numbers for the support of the Digital Delac Roy.) The University of Sorbonne briefly announced the organization's involvement in February shortly after the international AI summit was held in Paris, but its role has not been detailed until now.
For Jobert, it is the pinnacle of the passion he had for almost 40 years, as he was a young teaching fellow at Harvard. He stood at the Museum in Boston, before Delacroix's “Lament” on a canvas from 1848, in mourning surrounding the body of Christ after the cross, and when struck by John the Baptist, often covered in red cloaks, symbolizing the challenge. “I can't explain why,” Jobert said in a video interview. “But for me, this red cape is an image of the soul.”
At this point, Jobert has gathered an informal consortium of French institutions, including the Ministry of Culture and the National Centre for Science and Research, as well as the Center for Humanities, the AI and other institutional centre at the University of Sorbonne. As work on the digitization of the text is on track, Jobert's attention now focuses on murals of murals painted for the grand buildings occupied by the French Parliament.
His focus is on the Parliament, the House of Representatives, which occupy the 18th century Palais Bourbon. “We have two projects,” he said. “The first thing is to make it accessible on the website” – the vast room that Dela Croix has worked for nine years, zooming in on what they want, to allow people to take a virtual tour of the Congressional library. The second goal is to analyze these murals to solve the problem of belonging. What did Delacroix paint himself, and what did he leave behind for his assistant? “This is the part where AI plays a major role,” he said.
It is also the part where Schmidt science intervene. “This issue for multiple authors is really difficult,” said Brent Shields, an American computer scientist who leads the organization's humanities department. It's difficult to solve that using AI, he added, “This is one of the reasons why I love it.”
Seal has encountered a tough problem before. A few years ago, he and his team at the University of Kentucky had invented the process of using AI, among other technologies, to decipher the contents of carbonized papyrus scrolls excavated from Roman villas buried in eruptions that destroyed the Dead Sea and Pompeii.
“As philanthropists, we have the ability to take risks that governments and businesses can't or don't,” Wendy Schmidt said in an email. Bloomberg currently estimates Schmidt's wealth at $32.5 billion.
One reason why attribution efforts are expected to be difficult is that they rely on analytical AI that is completely different from the current frenzy-making AI that will cause the frenzy in 2022 with the release of tools like ChatGPT.
To get a grasp of who painted what, under Jober's direction, researchers created closed, high-resolution photos of the mural and reconstructed the work in digital 3D using photogrammetry. Technical data on Delacroix and other painters is provided by the research department within the Ministry of Culture. All of this is fed into a computer vision system that is trained to recognize Delacroix's brush strokes and his studio assistant's brush strokes. “I think it's likely to work,” said Xavier Fresquet, assistant director of the Sorbonne Artificial Intelligence Center.
Jobert wants to do the same with his murals in an even more spectacular room in the 17th-century Palais du Luxembourg, home to the Senate. But his ultimate goal is far more ambitious than that. Delacroix's virtual reconstruction using generative AI of all-talented murals once adorned the Salon de la Pike at Hotel des Villes in Paris. Their central element was “Peace descends to Earth,” a ceiling panel depicted in the words of 19th-century writer and critic Théophile Gautier.
Earth's prayers, if not art, will not be answered in life. In 1871, eight years after Delacroix's death, his murals came to flames along with the rest of the buildings as Parisian revolutionaries closed the place when they were being crushed by government forces. What remains in the archive are sketches of Delacroix, some etchings and two watercolors, presented to Queen Victoria by Emperor Napoleon III in 1855. Nevertheless, Jobert hopes he can come up with a reasonable fax machine for the Hotel Des Ville mural.
“We don't give you an exact replica of the room. That's not possible,” he said. “But we will give you what it may have been” – and it will still be the case if peace actually came down.