- May 22, 2017
- No Comment
What Happens When You Ask Machines to be Creative?
The concept of machines taking people’s jobs is a heavy thing. It’s played a part in scaring a whole generation of young people away from the manufacturing industry; and has tech leaders like Elon Musk calling for a Universal Global Income to make up for all the human jobs expected to be overtaken by machines.
While the speed of this AI expansion is startling, there are certain roles it’s always been thought the machines could never do. Most prominent of these select occupations is that of artists and creatives. However, scientists and boffins the world over are conducting experiments in which machines are demonstrating abilities that, on the surface at least, look a whole lot like creativity.
DeepDream: mind bending art created in seconds
Built by Google, DeepDream is an artificial neural network designed to detect patterns and faces in pictures. It was designed to aid with image searching and classification and is a useful tool. But, like a Pink Floyd album, things get a bit crazy when you run it in reverse. The AI has a massive database of images at its disposal and, when the algorithms that drive it are flipped, it over-analytically finds and enhances patterns. The results, which generate in seconds, look like psychedelic art that would take a human exponentially more time to create. You can play with the DeepDream software yourself online and make instant artworks like this out of any image you care to upload.
These creations are like a machine version of the cloud gazing you did as a kid—finding animals and faces and mythical creatures—only there’s no grass, no breeze, no ants crawling on anyone’s legs. It’s all done in labs with machines and algorithms. Things start getting even more interesting when you rerun DeepDreamed images through the software.
In this instance, you cal see the AI has taken its ‘inspiration’ from Van Gogh’s Starry Night. It’s performed an analysis, consulted its database and followed human-designed programming to string together concepts into a unique image. Does this qualify as creativity?
You can see from the comparison that the work has definitely been re-imagined. It is different to Van Gogh’s creation yet undeniably inspired by it. The slightly scary thing is, this is a characteristic of human creativity too; taking inspiration from those who’ve gone before to create something new. Yet, somehow, when the machine does it, it feels different.
AI creativity with words and colours
Research scientist, Janelle Shane, developed a neural network of her own, designed to complete a different creative task: naming paint colours. Her first attempt resulted in meaningless letter-tangles:
Not particularly indicative of the colours they’re supposed to represent. Shane adjusted her algorithms to help a machine out and found herself a step closer to some kind of meaning.
This time the AI actually managed to reference colours, though it seemed to have a green/gray obsession bordering on complete colour blindness. After another round of tinkering, Shane’s colour naming machine finally came into its own. While its choices may not be everyone’s interior design fantasy, they are undeniably entertaining, with a distinctive sheen of creativity.
Creative or not, it’s hard to imagine people sipping wine and explaining to friends that they’ve painted their living room walls in Turdly and are tossing up between Bank Butt and Stanky Bean for the kitchen.
Machines making horror movie trailers
Having created a horror movie centred around AI, 20th Century Fox decided to embark on a wildly appropriate experiment: have AI design their teaser trailer. The project was handed over to IBM and their super-intelligent machine, Watson. After an accelerated learning period and a few horror movie marathons, Watson conducted detailed analysis of the movie in question (Morgan).
The process was similar to the examples explored above. The machine’s output has the appearance of creativity but there’s no original thought involved; only reproductions and remixes of what it’s been taught. The problem is, as soon as you feel comfortable saying that’s not ‘real’ creativity, an awkward question pokes its chubby fingers into your brain fat: how different is that from what humans do? We too are taught, build up a bank of information and then make all our choices (creative or otherwise) from our analysis of stored data and the information at hand.
According to John Smith, leader of the IBM team behind Watson,
“we still have to define what creativity means. We know some of the attributes have to do with finding something novel, unexpected, and yet useful.”
While the AI can manipulate creative input based on its algorithms and database access, we are yet to see evidence of creation without guidance. In much the same way, Watson can, on an analytical level, understand what is scary to humans (if properly programmed by said humans), but it doesn’t feel the fear.
Check out the trailer Watson helped create below:
What about our creative jobs?
We don’t even know if it’s possible to teach a machine to be creative but, the people pursuing this field say that’s not what they’re trying to do anyway. Just like in other industries, automation and AI are being designed to augment human activity, freeing us up for higher pursuits and, in the creative fields, providing a new inlet for inspiration.
“It is not our goal to recreate the human mind. That’s not what we’re trying to do. What we’re more interested in are the techniques of interacting with humans that inspire creativity.” Rob High, Vice President and Chief Technology Officer for IBM Watson
Watson’s team leader, John Smith, elaborated on this point and, while he’s speaking specifically about the film industry, his sentiment can be carried into other fields as well, from manufacturing to construction.
“With film-making, 99% of the work is actually very mundane. It’s going through hundreds of hours of video in some cases to arrive at the core pieces to use. So there’s still a very good reason to use technology as an assistant here, rather than replace the human in the loop.”John Smith, Manager of Multimedia and Vision at IBM
