Tate Uses AI to Match Old British Art to New Photojournalism

Art history repeats itself in contemporary photojournalism. That’s the premise of Recognition, a new project that uses artificial intelligence to pair British artworks from the Tate’s collection with visually similar photographs from today’s 24/7 news cycle.

Recognition won the 2016 IK Prize, which challenged contestants to use artificial intelligence to explore Tate’s collection of British art. It’s a matching game that’s resulted in some unexpected comparisons. After scanning 30,000 digitized artworks, the program paired a recent Reuters photo of eunuchs applying makeup before Mumbai’s Raksha Bandhan with “Two Ladies of the Lake Family,” a 1660 painting by Sir Peter Lely, the chief painter of Charles II. Despite hailing from different centuries and hemispheres, both images feature two seated figures wearing elegant blue and greenish dresses against a background of deep red drapery.

What’s the point? Besides being a creative test of computer vision capabilities — like object recognition, facial recognition, and composition analysis — the program’s matches make the old seem new again, and vice versa. They draw connections between cultures and time periods that, on the surface, seem alien to one another. In doing so, the program provokes questions; the Mumbai photograph of eunuchs applying makeup paired with the 17th-century painting of royal British women might spark musings about gender identity, or cultural standards of beauty, and so on. In this mashup of old and new, the project is reminiscent of B4XVI, which pairs photographs of rappers with sculptures, paintings, and statues of similarly blinged-out historical figures from the Metropolitan Museum’s collection.

Developed by Angelo Semeraro, Coralie Gourguechon, and Monica Lanaro of the Italian communication research center Fabrica, in partnership with a team of AI specialists at the French company Jolibrain, Recognition is only the latest project to explore how artificial intelligence can change the way humans look at images. So far, the value of robots’ contributions to human understanding of art doesn’t often go beyond novelty — take Microsoft’s emotion-detecting app, which deemed the Mona Lisa 43% happy, for example. But with Recognition, a robot inadvertently creates subjective meaning with its pairings, and starts to seem like a shrewd commentator. A recent Reuters photograph of a Singapore airport’s control tower alongside a bustling construction site paired with “Industrial Landscape” (1955) by L.S. Lowry, for instance, seems like a comment on the relentlessness of industrial development. 

Just launched, Recognition will now spend three months analyzing the latest Reuters news photographs and comparing each to thousands of digitized paintings and sculptures. The best matches, as determined by the technology, will be entered into a searchable online gallery. Paired images are accompanied by explanations of how the program made the matches. The website indicates the strength of similarities between the paired images’ compositions, faces, objects, and metadata (i.e., titles, tags on the image, etc.). A corresponding exhibition at Tate Britain allows visitors to compare the machine’s matches with their own. 

See the results of Recognition online and at Tate Britain‘s Archive Gallery (Millbank, London, UK) through November 27. 

• ShareCopied to clipboard
• Mail
• Bluesky
• Threads
• LinkedIn