In 2016, ING Bank, advertising agency J. Walter Thompson Amsterdam, and Microsoft unveiled “The Next Rembrandt,” a brand-new portrait designed to look as if Rembrandt van Rijn had painted it, even though the artist died in 1669. The work was a 3D-printed canvas built entirely from a computational analysis of Rembrandt’s existing output.
According to Microsoft’s own account, the team analyzed all 346 of Rembrandt’s known paintings using high-resolution scans and machine learning. They built an algorithm that detects over 60 facial features to determine typical proportions, then assembled a composite subject - a 30-to-40-year-old white man with facial hair, wearing dark clothing and a hat, facing right - matching the demographics most common in Rembrandt’s portraits. The project ran for 18 months, consumed 150 gigabytes of rendered graphics, and used Microsoft Azure to run the analysis on Linux virtual machines “up to 1,000 times in parallel.” The final image was 3D-printed with layered ink to mimic real brushstroke height and texture.
The stated goal was not to fool anyone but to “stimulate the discussion on how data and the use of data could lead to innovation.” The campaign went on to win dozens of advertising awards.
Why business readers should care: years before generative art tools went mainstream, The Next Rembrandt showed corporations using machine learning on cultural heritage as both a technical demonstration and a marketing vehicle, previewing the credit, authenticity, and “is this really art” debates that followed.