What happens when creatives and AI collaborate? Here are five examples.
You’ve probably seen those headlines about Artificial Intelligence taking over creative jobs. They’re usually accompanied by an image of a Terminator-like hand perched threateningly over a typewriter.
As much as these headlines are compelling, most of them are hype. The media love pitching AI and humans against each other. And on deeper inspection, it often transpires that only a small part of the story has been reported.
In reality, artificial intelligence doesn’t have to be a creative threat.
Right now, AI is giving us creatives the chance to stretch our thinking and to start creative quests in exciting new places.
Or to put it another way, AI is our new creative muse.
So, what does human-AI collaboration look like? And how can it inspire our creative thinking?
Here are five of my favourite creative human-AI projects from the last few years.
1.AI-generated chair designs
The ChAIr Project by Philipp Schmitt and Steffen Weiss is a lovely example of human designers working alongside AI. Here, the two designers trained a generative adversarial network (GAN) on a dataset of iconic 20th-century chairs.
The GAN generated a range of possible designs, which the humans curated and used as inspiration for a new series.
As a GAN doesn’t have preconceptions of what a chair has to look like, the generated designs were offbeat and unconventional.
This meant the designers started their thinking in a new place. They played with the concept of a double-backed chair and toyed with unexplored shapes.
2. AI-generated TV ad script
Back in 2018, ad agency The & Partnership worked with their client Lexus to create the world’s first TV ad scripted by AI.
Creating the ad involved training a neural network on 15 years’ worth of Cannes Lions award-winning car and luxury goods advertisements. The program was also primed with emotional intelligence data from Unruly to teach it which moments of the chosen adverts connected best with viewers.
The final AI-generated script was then handed to the award-winning film director Kevin MacDonald, who did a great job of bringing it to life.
Whilst this is not the best ad I’ve seen, it certainly highlights AI’s writing capabilities and helped Lexus to build on their innovative brand story. The fact it occupied a lot of column space didn’t do them any harm either.
3. A grammy-nominated album
Last year, indie dance band, YACHT, used AI to help them craft their highly acclaimed album, Chain Tripping.
The process involved notating their back catalogue of 82 songs in MIDI, extracting the most exciting riffs, drum patterns and melodies, then running these through a machine learning model.
“It wasn’t something where we fed something into a model, hit print, and had songs,” says the band’s singer, Claire L. Evans.
The human members of the band carefully curated the model’s output, crafted the lyrics and decided on the overall structure of each track.
The result was a GRAMMY-nominated album — and strong evidence that AI can viably assist our creative thinking.
4. AI clothes design
The Little Black Dress or LBD is a wardrobe staple for many. So when I heard MIT were using AI to create LBD designs, I was intrigued to see what an algorithm could bring to the sartorial party.
Just like the chair project, the team trained a GAN on thousands of dress designs until the AI algorithm learned how to generate new designs.
Humans then curated the most exciting elements of the AI’s output (asymmetrical neck, pointy shoulders and one-sided bell sleeves) and turned these into the ultimate black dress.
You can see the final chosen design in all its glory here.
5. AI cocktails
I couldn’t write a blog on human-AI collaboration without mentioning one of our experiments at Tiny Giant.
Over the last year or so, we’ve had a lot of fun creating AI-generated cocktail recipes and turning them into reality.
We trained a recurrent neural network on hundreds and hundreds of cocktail recipes, until the AI algorithm suggested recipes of its own.
We selected the most interesting and delicious, based on our own exhaustive taste-testing (a tough job!). Then worked with a brilliant human mixologist, who crafted the output to perfection.
The journey from fear to fan
What I loved most about this project (apart from the tasting) was the mixologist’s journey with us.
At first he was terrified that AI would be better than him. But after working together, he changed this viewpoint.
As the AI-generated recipes were often unconventional, our mixologist was working with ingredients he wouldn’t normally use. He developed cocktails that he’d never considered before — and was excited by the results.
If you go to his bar today, you’ll see a whole bunch of bizarre ingredients in little drawers behind the bar — and a recipe on his menu that looks remarkably like one of the AI-generated ones.
As it happens, our cocktails went down well outside of the mixologist’s bar, too. Our weird and wonderful creations have been served up at events all over the place, from Berlin to Bristol and Texas to the top of an Austrian mountain.
Which AI cocktail is my favourite? It has to be the grapefruit and pomello gin creation with Yuzu sorbet, Kwai lychee liqueur, orange bitters, cranberry and burnt rosemary and thyme syrup.
Turns out human-AI collaboration doesn’t have to leave a bitter taste in your mouth after all!
