Art of Neural nets challenges Man

In the I, robot movie Will Smith’s character, a Detective, asks an android if he could splatter paint across a piece of fabric to produce a masterpiece, to prove that he has creative potential of the Man. Robot replies he can’t but then, acting like a plotter, draws precise moment from his vision with a marker.

Is that an art? Quite possible, it is.

Now, let’s consider same things are being done for first time in real life by computers. Washington Post had recently shaken the world of arts (well, I was shaken, at the least) with their article on how computers are able to analyze artistic style and mimic it [1]. Great painters of days of yore spent their entire lives to find these styles, unique touch of brilliance that made them distinctive and outstanding, even though the act of painting took some small time, preceding efforts were numerous and plenty. And now, with neural networks technology, virtual robot artists are able to decrease even those time periods to create if not a new masterpiece, but an enchanting work of art, or, at the least, a worthy souvenir.

The algorithm took the top left image and applied the style of the small, inset images to create works with very similar styles. (University of Tuebingen)
The algorithm took the top left image and applied the style of the small, inset images to create works with very similar styles. (University of Tuebingen)

By giving an initial image (say, a photo of some housing or a portrait of a movie hero) and picture which style to refer to (like, say, Van Gogh’s Starry night), German scientists[2] are able to enact the algorithm that turns artist’s style into a block of data, and suggests how would lines and colours of initial photo be transformed according to what program considered to be artist’s style. Results are amazing, although they look to me like postmodern works of some followers – but they already beat many folks’ achievements in field of visual art.

Here’s a demonstration of how the network creates the new image. (University of Tuebingen)
Here’s a demonstration of how the network creates the new image. (University of Tuebingen)

Or course, the power to invent and set a new style still belongs to humans – but, really, how many are there those genial trendsetters now? And world could go with beautiful derivative works.

  1. Featuring Neural nets art
  2. Original scientific statement

 

 

 

 

Leave a Reply