Neural networks can do some awesome things, from allowing cars to drive on their own to instantly translating dozens of languages. But when it comes to our personal favorite use cases, we’re going with one recently engineered by Australian artist Chris Rodley.
By combining a deep-learning algorithm, a book of dinosaur pictures, and a book of flower paintings, he’s created some of the trippiest images this side of an M.C. Escher or Salvador Dali.
“The tech that makes this work isn’t mine, but was created by a team of researchers from Germany who published a paper on it two years ago,” Rodley told Digital Trends. “The idea of the group, led by Leon Gatys, was to transfer an art style as represented in a sample print to a target photograph. They called it ‘style transfer,’ and it was a huge success. Many people have been using it to see how they’d look painted in the style of [Georges] Seurat or [Pablo] Picasso or whatever. But the most interesting work I came across wasn’t just applying form; it was actually taking content from one image, and using it to reconstruct the original target image.”
As well as flowers, Rodley also attempted combining dinosaur images with still-life paintings of fruits and vegetables, along with several other objects, with the results being truly mesmerizing.
“It turns out, through trial and error, that dinosaurs are perfect targets for this algorithm, because their shapes are instantly evocative,” Rodley continued. “[Whereas something like] the Mona Lisa is not so great, since its composition isn’t especially distinctive; rather it’s in the detail.”
It’s a stunning example of some of the creative possibilities available to artists when they utilize artificial intelligence as a tool that essentially transforms into brushes that can think for themselves. “AI is developing so fast now that there seem to be more creative tools than there are creative people to use them,” he said.
Since uploading the pictures to his website, Rodley said he has been overwhelmed by requests from people, many of whom want to know where they can get copies of the images themselves. “I added a contact form to my site since so many people were asking for prints,” he said. “I’ll try and make that happen.”
- Kindle or print? What readers choose depends on what they read
- Looking for a good read? Here are the best, most eye-opening books about tech
- 15 epic sci-fi novels you should read before they become blockbuster films
- Microsoft Surface Book 2 13-inch Review
- Can an algorithm be racist? Spotting systemic oppression in the age of Google