Skip
Current Issue
This Month's Print Issue

Follow Fast Company

We’ll come to you.

2 minute read

How Art And Algorithm Came Together To Create "The Next Rembrandt"

Data science and art history joined forces on a convincing portrait in the style of the Dutch master.

  • 01 /09
  • 02 /09
  • 03 /09
  • 04 /09
  • 05 /09
  • 06 /09
  • 07 /09
  • 08 /09
  • 09 /09

A "new" Rembrandt has been unveiled in Amsterdam, but it’s not a long lost dusty relic that was found in someone’s loft—it was created by data analysts and computers.

The portrait of a man in a black hat is the result of 18 months work by art historians, data scientists, developers, 3-D print technicians, and organizations like Microsoft, Delft University of Technology, the Mauritshuis in The Hague, and Amsterdam's Rembrandt House Museum. It consists of more than 148 million pixels, based on 168,263 painting fragments from Rembrandt’s output.

The initiative is the brainchild of Bas Korsten, executive creative director at ad agency JWT Amsterdam, and was created for Dutch financial services giant ING. Korsten says ING approached the agency with a brief to "find a way to bring their innovative spirit to their sponsorship of Dutch art and culture in a way that would get people thinking."

Korsten says the actual idea came to him when he saw the face of Jesus."Not in a religious awakening, but in a magazine," he says. "It was an article about a group of archaeologists who had ‘recreated’ the face of Jesus out of three ancient skulls found in Jerusalem. I thought, ‘If they can make something new out of historical material, why can’t we distill the artistic DNA of Rembrandt and create a new work out of that?’"

Why not, indeed?

The assembled team then embarked on a journey, analyzing all 346 of Rembrandt’s paintings using high-resolution 3-D scans, which were upscaled by a deep learning algorithm. Then, working with Rembrandt experts, the subject was determined by demographic segmentation and study of all the artists’ previous subjects. A facial recognition algorithm learned from the geometric patterns Rembrandt typically used, and then replicated the style to generate new features. Finally, the 2-D version was given depth and texture by a computer mimicking the brushstrokes used by Rembrandt.

People can explore the process and interact with the content on The Next Rembrandt website.

"Our intention for The Next Rembrandt was not about trying to produce a 'new' Rembrandt," Korsten says. "Our idea was to create the 'next' Rembrandt. So, we chose to use new technology to actually get under the skin of Rembrandt. And then to predict, on the basis of analysis, what Rembrandt might have painted next. This isn’t so much about creating a piece of art, but rather about creating a context for a discussion on where technology can take us next."

Korsten believes the project raises some interesting questions on the nature of art, and to what extent technology could ever recreate the spark of creativity. Not to mention the ethics involved. "How does it relate to other paths where artificial intelligence is taking us?" he says.

loading