Can Artificial Intelligence Learn To Create Emotionally?

4minute
read

In 2016, esteemed anime director Hayao Miyazaki was invited to attend a tech demonstration at Dwango Artificial Intelligence Laboratory. Miyazaki is known for his brutal behind the scenes demeanour – he lambasted fans who obsess over the genre, and told his son his film was so bad he should never make another one – but the lab’s team obviously expected they would receive positive feedback from the animator.

They were wrong.

“I strongly feel this is an insult to life itself,” he says, stunning the team into silence. They seem embarrassed, but also confused. This is cutting-edge technology, and they work for one of the largest media companies in the world. Have they really failed to impress?

Miyazaki’s issue isn’t with the technology. He’s not a stickler for the old ways. Two years after announcing he was set to retire (for the sixth time), Miyazaki returned to Studio Ghibli in 2015 to produce his first CGI film. His problem is with the team’s suggestion that the AI system running the tech demo understands such complex notions as pain in ways that humans can’t comprehend.

He shares the story of a disabled friend whose weak muscles barely allow him to raise his hand for a high five. “Now, thinking of him, I can’t watch this stuff and find it interesting. Whoever creates this stuff has no idea what pain is whatsoever.”

It’s an important point. Creativity is not just inspired by emotion – in this case, the notion of pain. It is emotion. It is in knowing pain.

No, this is not about romanticising the tortured artist, but about the psychology that drives creativity. A 2013 study entitled Does Negative Affect Always Narrow and Positive Affect Always Broaden the Mind? Considering the Influence of Motivational Intensity on Cognitive Scope found that emotions considered either positive or negative have a beneficial impact on the creative process.

Further research has confirmed these findings, and concluded that we cannot cheat our brain into becoming creative. As Imagination Institute’s Scientific Director Scott Barry Kaufman puts it“There’s something about living life with passion and intensity, including the full depth of human experience, that is conducive to creativity”.

This experience is something Artificial Intelligence lacks and, as many experts agree, will always lack. AI is being developed to exceed the limits of human capability, by moving beyond the limits that the human experience has defined. Not to override it.

Take for example Sunspring, the first AI-scripted film, which I wrote about in 2016.

Looking deeply into the film, beyond the system’s idiosyncrasies, meaning may be found, but only if we decode it. What we experience only feels human because of the human actors. Without them (i.e. when you read the script), there is no emotion. The system decided it wasn’t happy with its name (the creators called it Jetson, before it told them it wanted to be known as Benjamin), yet it doesn’t even give the characters names, one of the simplest ways in which a human artist would forge an emotional connection between their characters and an audience.

So can Artificial Intelligence learn to create emotionally, as all the great human artists do? The answer is no.

Does that mean AI won’t have a future in creative fields? Again, no.

When Dwango’s Chairman Nobuo Kawakami tries to justify Miyazaki’s dislike for the test demo by explaining it off as “just our experiment,” Studio Ghibli producer Toshio Suzuki asks what they actually plan to use the system for.

A moment of silence follows, before

“Well, we would like to build a machine that can draw pictures like humans do.”

Miyazaki and Suzuki are dumbfounded. No explanation is forthcoming.

“We humans are losing faith in ourselves,” the former declares.

In March of 2016, another Japanese company, advertising agency McCann Erickson, announced they were hiring an ‘AI Creative Director’ who would use ad metadata to provide “logic-based creative direction” in an upcoming campaign.

By June, McCann announced the first of these campaigns was for Mondelez brand Clorets Mint Tabs. Two commercials were released online: one by the AI, one by a human. Viewers were asked to vote on which they thought was best.

The human-directed ad won…but not by much. The final margin was a narrow 54% to 46%.

Why were the results so close? It might be argued that it’s indicative of the less artistic, more logical approach that creative industries have been taking long before an Artificial Intelligence system was asked to direct an ad, so that a computer’s gaps in understanding are unnoticeable. It might simply come down to preference.

Whatever the case, I feel confident saying that the best art will always be the most human art.

too many entries