What does artificial intelligence (AI) have to do with a 150-year-old lion statue in London’s Trafalgar Square? Good question. But Google’s never been one to let cognitive dissonance stand in its way. Today in partnership with British designer Es Devlin, the search giant took the wraps off “Please Feed the Lions,” an exhibit on display as part of the London Design Festival.
During the festival, curious onlookers can “feed” the fifth lion in Trafalgar Square by typing a single word into a Google Pixelbook, which will expand it into a line of poetry projected inside the lion’s mouth. As more folks contribute, the poem will “evolve” and (thanks to a clever bit of nighttime projection mapping) wrap around both Nelson’s Column and the lion’s lower body.
A neural network is responsible for the nouveau verses. Ross Goodwin, a self-described creative technologist who’s part of Google’s Artist and Machine Intelligence program at the search giant’s Arts and Culture division, trained it on 25 million words of 19th-century poetry and programmed it specifically never to repeat a line, so as to guarantee a degree of randomness.
Inspiration came from a conversation with London Design Festival Chairman Sir John Sorrell, Devlin said.
“[Sorrell] said: ‘[Sir Edwin Henry] Landseer never wanted those lions to look so passive; he proposed a much more animated stance, but Queen Victoria found it too shocking.’ The thought lodged in my mind. What if we could invest the lion with a diversely crowd-sourced collective poetic voice?”
Poems will be published daily on the exhibit website. Following the festival, which ends September 23, Google will digitize the sculpture as part of an online exhibition on the Google Arts & Culture app.
Google’s public experiments with art and AI date back to 2016, when it released DeepDream, a neural net that generates trippy canvases from images. April 2017 saw the launch of Google’s AI Quick Draw, which taps an algorithm to transform doodles into clip art. And in January, the Google Arts & Culture app gained a feature that scans selfies for resemblances in a 70,000-strong library of paintings and portraits.