Artificial Intelligence (AI) is more commonly associated with automation rather than creativity. But as the technology and its applications continue to advance, some creative workers, such as artists, designers, writers and even movie-makers, might be beginning to worry that the robots are gunning for their jobs.
There is no doubt that AI has plenty of use cases within the creative industries. In fact, it is already widely applied in tools, software, and services that help creatives to do their jobs. That means that in recent years, the latest AI applications have almost inevitably seen the line between automation and creativity become ever more blurred. But for all its speed, efficiency and computer-powered intelligence, can it ever truly replace the human brain’s capacity for creative thinking?
At the University of Warwick in the UK, a team of engineers has trained a neural network (a computer network that aims to replicate the neurons of the human brain) to recognize beautiful landscapes. The question this poses is, if AI can do this, could it also potentially select the most effective imagery for ads or other promotional materials, sidelining the art director along the way? Google seems to think so.
The team there made the code for their DeepDream AI neural network project available to whoever wanted to use it, auctioning off some of the resulting art. The finished work looked so impressive that people were willing to pay big money for it, with the auction raising $84,000 for charity. Those introducing the exhibition made the point that humans have been using technology to create art for centuries, with the use of artificial neural networks being a further extension of that principle.
This then begins to open up all sorts of philosophical debates about what is art. Is it all about aesthetics, or are a work’s origins and artist important to its meaning and quality? And in the case of these paintings, who is the artist? Is it the software and hardware making up the artificially intelligent network that generated the output, or is it the programmer that created the algorithms in the first place, or is it the middleman in between?
One of the creators of the artwork sold on that night, Memo Akten, described the relationship as Google having created a “better paintbrush”, but insisted that the human artist was still an essential part of the process.
Superstar artists like Damien Hirst have often used assistants to actually put the finished art together, but conceptually he is still considered the ‘artist’. He has made the point himself that nobody criticizes architects who don’t build their own houses. In the case of neural networks, you could argue that the assistants are simply machine powered, rather than human.
The debate over what is art has raged for centuries between some of the sharpest minds on the planet, and it’s not an argument we’re proposing to settle in this article. But it certainly gives you something to think about. The key question then is whether or not this neural network art have been created without input from a creative human brain – an artist – in the first place.
AI killed the radio star?
Musicians might feel they have reason to worry too thanks to another team of Google researchers who have been using AI to make music through the Project Magenta program. The background is the essentially the same for both creative art forms. Complex mathematical systems allow the machines to learn behavior by analyzing vast amounts of data. By looking for common patterns, AI can then recognize structures, whether that’s an image of a tree or the musical characteristics of a composition.
Once the machine has learned the basic structures and how they relate to each other, it can then replicate them and create new ones together based on what’s been shown to work best in any particular scenario. It’s easy to see how this could save a lot of time when it comes to basic design projects or creating a soundtrack for a movie. Google’s technology, known as NSynth, is even able to merge the characteristics of different instruments to create new sounds.
Machines are also already taking on the role of DJ, using technology rather than musical taste to choose tunes that go well together. And whether they know it or not, this is something many music-lovers are already exposed to on a regular basis. Ubiquitous music streaming service Spotify has placed machine learning at the heart of its development for a number of years, with algorithms powering playlists and recommendations by detecting patterns in the music and piecing together what it thinks people will respond to.
It’s easy to see how this usage can be translated into public performance – a club just fires up an AI-powered playlist instead of needing a DJ to spin the tunes. It may be data-based rather than artistic ‘instinct’ that is helping to fill the dancefloor, but if the results are the same for those consuming the music, does it matter?
Who made who?
It is often said that art is in the eye of the beholder, so how a person reacts to a work of art can be crucial in shaping the future direction of creativity. It may seem safe to assume that the role of art critic will always be reserved for humans. But what happens when the machines start judging each other?
A recent group effort between the Facebook AI team and a group at Rutgers University in New York went even further by modifying an algorithm known as a generative adversarial network, in which two neural networks ‘play’ against each other. Set a task, one network will come up with a solution, and the other one will judge that effectiveness of that solution. With the networks alternating between ‘creator’ and ‘judge’, the results get progressively better. Significantly, when the images created were then assessed by the public in an online survey, the visuals created by AI rated higher than those created by humans.
Many advertisers are already using AI in their campaigns. M&C Saatchi was one of the first. In 2015, the advertising agency developed a campaign that consisted of digital display with an integrated camera to record audience response. The machine relied on algorithms to assess the most effective elements of ads, such as copy, layout and visuals.
Over at IBM Research, AI was used to create a movie trailer for the horror movie, Morgan. IBM’s technology, Watson, analyzed the visuals, sound, and composition of hundreds of horror trailers and then selected which scenes from the full movie should be included based on what it learned.
Can creativity be automated?
In both of these cases, there is a great deal of artistic input that goes in before the AI can start to put together the pieces. In some ways, this could be described as artistic optimization rather than creation, using machine learning to essentially carry out focus groups at a greater scale and speed than would be possible using traditional techniques. Creatives have put together the pieces of the campaign and the different ideas, while the machine learning helps to choose the combinations that work best.
A lot of these examples explored above rely heavily on human guidance. AI might be able to create elevator music, pretty pictures, and even patch together a basic trailer for a horror flick, but much of this is automation, as much as creativity. After all, isn’t creativity about pushing boundaries and coming up with something that’s new? And is it even possible for AI to replace the human element? Certain industries, like advertising, are based on emotional response, in other words, what people feel. Algorithms can’t simulate that – not yet at least.
It could also be argued that some artisanal jobs, such as hand carving or furniture making will become more valued in future for the very reason that the work has been done by hand and not by AI. The flair and individuality employed by the craftsperson mean each piece is unique whereas automation promised by AI means standardization of the end product. It’s equally hard to conceive that AI can replace imagination or ‘thinking out of the box’.
While many advertising agencies are launching AI divisions, these will be mostly focused on split testing to discover which creative strategies work best. So it’s more likely that AI algorithms will be used to augment human functions rather than replace them entirely, freeing up creatives to focus on higher levels tasks.
And many of those developing the technology would agree with that. As Rob High, Vice President and CTO for IBM Watson says: ‘It’s not our goal to recreate the human mind – that’s not what we’re trying to do. What we’re more interested in is the techniques of interacting with humans that inspire creativity in humans.”