Creative machine thinking (Column)

Don't dismiss AI's overhyped potential

Illustration by Celia Krampien

We have too grand of a view of our own creativity. The underlying reality in all of this… is there is nothing magical about the human brain

Pedro Domingos, University of Washington

When Google DeepMind’s AlphaGo defeated world champion Go player Lee Sedol back in March, it was hailed as a breakthrough in machine learning and artificial intelligence — 10 years sooner than most experts predicted.

AlphaGo was designed to win a board game, but the nature of that board game and how AlphaGo conquered it portends profound new (some might say unsettling) possibilities for the creative industry where AI systems could replace creative professionals.

While the premise of Go is simple enough (two players try to command a game board by taking turns placing black and white stones), the game play is enormously complex. For every move, the opposing player has many more countermoves than in a game of chess. Google claims there are more possible configurations on the board than atoms in the universe. Because of that complexity, an AI program couldn’t evaluate every possible move as it did when IBM’s “Big Blue” defeated chess Grandmaster Garry Kasparov 20 years ago. It’s why Go is often considered a game of intuition rather than logic. DeepMind engineered AlphaGo to function in a fashion that mimics human intuition.

“The search process [to evaluate moves] is not based on brute force, it is based on something more akin to imagination,” said DeepMind’s David Silver in one interview. The climactic moment in the contest with Lee Sedol came when the AI made a move so totally unexpected, it knocked the champion off balance. He literally had to leave the room to compose himself. The move was described as “creative.”

Intuition, imagination, creative — each used to describe the neural networks underlying AlphaGo. If you let your mind play a bit with that reality and extend the trend line a little, it’s not hard to envision a future where AI systems create ads and come up with strategic concepts for brands.

The natural, human, reaction for many will be to reject the premise, to believe that creativity is a defining attribute of our humanity not replaceable by algorithms or lines of code.

mrkt03_filter_gc01-aBut if you ask scientists, there’s a good chance they will respond with a version of the pop-psychology maxim that all things psychological are biological. Human emotions, intuition and instinct — including moments of creative inspiration — are all, fundamentally, about some of the 100 billion neurons in the body transmitting information to other neurons. An unusual volume of information, processed and combined in unusual ways is what we call creative thinking. Computers are good at processing information and we can make them combine that information in far more random and numerous ways than humans can handle — in other words, to think creatively and come up with original ideas.

DeepMind founder Demis Hassabis, at a London School of Economics panel talk last year, said “I don’t think that creativity is necessarily this mysterious thing… Creativity itself is something that we would hope our machines would have.”

Pedro Domingos, a professor at the University of Washington and author of The Master Algorithm, is a bit more blunt: “We have too grand of a view of our own creativity. The underlying reality in all of this — if you are a reductionist like most scientists — is there is nothing magical about the human brain. So whatever it is doing, we can do with a machine.”

Simon Colton, a professor of computational creativity at the University of London, believes that computers can be useful in producing the building blocks of the creative process: ideas. He’s already produced computer-generated paintings and a musical. His latest undertaking is the What-If Machine project, a joint venture to produce software that will “invent, evaluate and present fictional ideas” with real value for stories, films, paintings and advertisements. The interactive site allows visitors to generate their own what-if scenario, choosing different variables from pull-down menus. Some of the examples seem absurdist and abstract, while others, like “What if there was a merchant who woke up in a sea as a fish but could still use deodorant?” wouldn’t seem out of place in an Old Spice pitch meeting. (And someone working for Cadbury must have said at one point, “what if we had a gorilla play the drum solo of a Phil Collins song?”)

“We are not yet producing great advertising ideas or anything close to them,” said Colton. But if the ideation process typically sees creatives churning out 10, 20 or even 30 ideas to arrive at one good one, that process of combining bits of information or elements of a story in different ways could be replicated by software. “That is classic computations creativity,” he said. Lately, he’s been talking with more advertising creatives who are intrigued by these possibilities, including Edelman UK’s executive creative director Bo Hellberg.

Hellberg believes artificial intelligence could provide creative departments with a new supply of creative thinking. “Collaboration with a machine that will provide more ideas, an abundance of ideas and perhaps explore territories that you wouldn’t think of, or wouldn’t have time to explore,” Hellberg said.

It may be unsettling to some, but the more time Hellberg has spent looking at the power of AI, the more he’s convinced that machines could one day do much of the work that creatives are doing today. “I can’t see a logical reason, a scientific reason with the way we are right now with AI, that they won’t get tot hat point,” he said. “I can’t see why not.”

Sidebar: The AI CD

There are already some AI creative directors, but their output so far is mere novelty — projects to get public attention. But they do demonstrate a level of skill that you may find surprising.

IBM‘s posterbot and Jeopardy Champ Watson generated a trailer for the film Morgan this summer. It scanned hundreds of horror films, learned the identify scenes of action, fear or emotion, applied what it learned to Morgan‘s emotional map, then selected cuts for an editor to assemble. The cuts do feel off, but the result is creepy… which, because Morgan is a horror movie, is surprisingly effective.

This story originally appeared as David Brown’s “Getting Creative” column in the Fall 2016 issue of Marketing

Add a comment

You must be to comment.

Tech Articles

Ourdata offers a more charitable ad blocker

B-corp's 'ad enabler' wants to help both publishers and consumers with 'data union'

Amazon unveils a store with no checkout

Sensors register shoppers' items and automatically charge them to Amazon app

GroupM integrates data offering with new platform

The media investment group has announced the global launch of [m]Platform

Increased demand drives Grocery Gateway’s growth

Longo's CEO says online grocery shopping has 'come of age'

Luxury retail must go digital or be forgotten (column)

AJ Dalal says luxury retail ignores the connected shopper at its peril

Canadians warm up to social commerce

PayPal and Ipsos research shows "Shop Now" buttons are gaining traction

Online ad exchange AppNexus cuts off Breitbart

Popular online ad exchange bans site for violating hate speech policy

Videology brings Bryan Segal on board

Former Engagement Labs CEO to lead Canadian operations

A CEO’s tips for using DIY video in consumer marketing (Column)

Vidyard's Michael Litt argues against outdated 'text tunnel vision'