It all happened when the Canada Media Fund recently hosted Analog, a conference that showcased a handful of AI experts in Vancouver, Toronto, and Montreal. The events offered an in-depth, nuanced look at the implications of AI in artistic endeavours. In Toronto, the CMF’s director of industry and market trends, Catherine Mathys, along with a computer science professor, a new media curator and critic, and an AI entrepreneur building audience demand analysis software took on the many facets of the debate around AI and the arts.
Creativity (Re)Considered
To think through the role of technology in creative practices, we first need a working definition of what creativity is. Does it refer to creating something out of nothing? Is it an ability to see invisible rules and bend — if not break — them? Is it remixing and recontextualizing ideas that are already known or in circulation? And whatever the answers to these questions may be, we need to ponder if there’s a net positive role for artificial intelligence in the creative process.
Defining creativity used to be fairly straightforward. Visual arts, music, and inventions were things made by people, using physical tools and artistic techniques. Similarly, defining ownership used to be more or less routine. With creativity melded with machines, however, it’s a different story. Once cut-and-dry issues such as copyright and authorship find themselves pondered by legal scholars and practitioners as a result.
The Computer Scientist’s Point of View
Are we using our creative brains on patterns we’ve seen before? Is our creativity anchored on things we’ve done and liked before? These are just a few of the questions University of Toronto Computer Science Professor Steve Engels asks. His work with AI finds its applications in such projects as games for the blind or to improve the cognition of seniors, and programs that generate music.
As an academic, Engels’ interest lies in the deployment of AI as a cognitive tool — something that supports human processes. One of the challenges, he says, is to figure out where the boundaries are between what’s best made by a human and what’s best made by a machine.
As a concrete example, he points to his work with the video game design community in Toronto, which has been experimenting with ‘authorless music.’ As the name suggests, that is music generated by a computer trained on algorithms. For game companies with limited music budgets, Engels says it can provide a “good enough” and affordable solution. He does, though, acknowledge what he calls “the Britney problem” — if an AI is trained on Britney Spears songs, there’s a non-zero chance that its output will sound like Britney Spears.
The Curator’s Point of View
The role of humans in creativity is critical. It is only by expressing aspects of the human condition that art ‘speaks’ to people, stressed new media curator and art critic Shauna Jean Doherty. She is therefore optimistic about a future in which AI intersects with art, but understands there are scenarios on the edge of both that represent real challenges.
She shared the example of the French art collective Obvious, creator of the first AI-based artwork that sold at auction. In the words of Hugo Caselles-Dupré, of Obvious, “we fed the system with a dataset of 15,000 portraits painted between the 14th and the 20th centuries. The ‘Generator’ makes a new image based on the set, then the ‘Discriminator’ tries to spot the difference between a human-made image and one created by the ‘Generator.’ The aim is to fool the ‘Discriminator’ into thinking that the new images are real-life portraits. Then we have a result.” The resulting piece sold at auction in 2018, and did so with a $400,000-plus price tag — over 40 times the most optimistic estimates.
She shared the example of the French art collective Obvious, creator of the first AI-based artwork that sold at auction. In the words of Hugo Caselles-Dupré, of Obvious, “we fed the system with a dataset of 15,000 portraits painted between the 14th and the 20th centuries. The ‘Generator’ makes a new image based on the set, then the ‘Discriminator’ tries to spot the difference between a human-made image and one created by the ‘Generator.’ The aim is to fool the ‘Discriminator’ into thinking that the new images are real-life portraits. Then we have a result.” The resulting piece sold at auction in 2018, and did so with a $400,000-plus price tag — over 40 times the most optimistic estimates.
The AI Entrepreneur’s Point of View
In 2016, University of Waterloo grad Jack Zhang developed an AI-based system that could weigh a plotline against some 40,000 attributes to predict ― and optimize ― the resulting movies’ chances of success. As Zhang explained at that time “before a single word was written, we used a computer program to analyze a massive amount of data to see what kind of plot elements in the film were driving audiences. So we correlated that with audience taste and behaviour data and see what type of plot would draw in what type of audience, and then feed that information to screenwriters who would closely work with our computer program to create that screenplay.”
A trailer for the movie, in this case a horror movie, was created for $30. Not knowing what to expect, the trailer was posted on Facebook. The result: 2.7 million views for a movie that didn’t exist. In other words, there was demand the supply did not meet. Zhang saw a market opportunity, something he recently explained in detail in an episode of the CMF’s Now & Next podcast.
The software he has since built out goes beyond AI assisting with the creative process. It is also capable of what he calls “audience demand analysis,” which can begin with the creative process and go all the way through to marketing and distribution.
Zhang looks at the creative process in film and TV as two-pronged problem. There’s the systematic and the random — the former being what is known, and the latter being what is unpredictable and often unexplainable. As he explained at Analog, his software seeks to combine these two domains. This does not mean it’s a replacement for creativity. But it provides the ability to run probabilities in a business known for being largely unknowable.
Incidentally, Zhang’s film, based on a trailer and some AI-fuelled instinct, began as a crowdfunded Kickstarter project, and is now financed with a $3.5 million budget, with production scheduled to begin in Q2 2020.
A trailer for the movie, in this case a horror movie, was created for $30. Not knowing what to expect, the trailer was posted on Facebook. The result: 2.7 million views for a movie that didn’t exist. In other words, there was demand the supply did not meet. Zhang saw a market opportunity, something he recently explained in detail in an episode of the CMF’s Now & Next podcast.
The software he has since built out goes beyond AI assisting with the creative process. It is also capable of what he calls “audience demand analysis,” which can begin with the creative process and go all the way through to marketing and distribution.
Zhang looks at the creative process in film and TV as two-pronged problem. There’s the systematic and the random — the former being what is known, and the latter being what is unpredictable and often unexplainable. As he explained at Analog, his software seeks to combine these two domains. This does not mean it’s a replacement for creativity. But it provides the ability to run probabilities in a business known for being largely unknowable.
Incidentally, Zhang’s film, based on a trailer and some AI-fuelled instinct, began as a crowdfunded Kickstarter project, and is now financed with a $3.5 million budget, with production scheduled to begin in Q2 2020.