With each new development, Artificial Intelligence continues to outstrip our best human efforts. It’s now more efficient at predicting heart disease than us, it’s able to detect pregnancy before our loved ones can, and it operates a mental horsepower far exceeding several thousand of our brains combined. But while it’s certainly superior at complex calculations and predictions, there’s one “human field” AI has yet to master: creativity.
For many, creativity represents the last truly human domain – the final battleground in a confused war against super intelligence. We’re proud of our erratic and unpredictable human brains, and machines, it has been decided, are simply no match.
So it’s completely understandable that new advances in “creative AI” make us feel uncomfortable. But once we move past our basic defensive reflexes, opportunity quickly outweighs perceived threat; the future of AI and creativity is already taking shape, and there’s a lot to be excited about.
AI as a creative assistant
When we think of creativity, we often think of artists and inventors having “lightbulb moments”. In reality, everyone has creativity, but turning initial ideas into finished products is complex. Thanks to AI, this process could become dramatically simpler.
As creative agencies will attest, many of the tasks involved in realizing an idea are mundane. Planning needs to take place, data needs to be analyzed and accounts need to be kept. AI offers the best means of facilitating these processes; either through helpful insight (like the data prediction tool iSeek), or complete automation (like automatic time tracker Timely). In these cases, AI is essential for removing the low-value tasks that disrupt and sink our creative energy.
But the promise of AI goes beyond reducing creative admin – it can actually enhance our creativity. AI can take thousands of data points from our purchases and use these to identify emerging behaviours and tastes. In fashion, trending colours, patterns and styles are being recorded and analyzed by AI to help designers create innovative clothing that appeals to us.
Interestingly, it’s a promise that creatives themselves are actually embracing; the vast majority of creatives in a recent Adobe study said they were not worried about AI replacing them; rather, they saw the potential for these tools to make their lives easier and enhance creative capabilities.
AI as a producer of art
If the only difference AI is making to the creative process is improved efficiency and more information, then nothing really distinguishes it from earlier tech advancements. Especially when you place it against the example of the internet, which completely revolutionized creative possibility by giving people around the world access to information. The real question then, is whether AI can come up with new ideas and creative products.
In a number of industries, AI is already there. IBM, for example, made AI the central decision-maker when producing the movie trailer for the horror film, Morgan. The algorithm was taught the emotions and sentiments related to different visual, audio and scene compositions – and essentially “learnt” the features that would make a scene frightening. From this, it selected the movie snippets that it deemed the most suspenseful for the trailer.
In the art world, even more complex “creative algorithms” have been developed. An AI was given a large number of paintings and taught the artistic style of each artwork. It was then tasked with creating a new piece which did not readily fit into any of the categories it had learnt. Another AI then judged the artwork to ensure that the output was still art, instead of some other kind of image (judge them for yourself!).
Both of these examples fulfil many of the criteria of creativity, in the sense that they’re innovative and the product of new ideas. Perhaps more important though, is the success they had in evoking emotions; public reviewers actually liked the work they produced.
Can AI be creative?
Both the AI-generated painting and movie trailer were produced through a type of AI algorithm known as “supervised learning”. In a nutshell, an algorithm is given a dataset and develops “meaning” based on how a programmer tells it to label the information. Essentially, humans tell the computer how it should understand ideas like ‘beauty and ‘fear’ – and, based on that training, the AI can then make judgements on new sets of data.
Broken down this way, it’s difficult to label this process “creative”. Human creativity involves taking ideas from different experiences of life and piecing them together in unpredictable ways. There is no pre-defined formula for “success”. In fact, creativity often stems from challenging the assumptions we take most for granted in life. It is an expression of escape, not a replicable equation.
In contrast, the “creativity” of supervised AI is inherently controlled and pre-defined, and humans essentially still call the shots. Though the AI has autonomy in weighing the relative importance of each feature it is considering, the parameters it judges by are defined by humans. Because of this, it is better to view supervised AI as an extension of human creativity, and not independently creative.
The future of creative AI
So, can we ever move beyond this “humanly prescribed” version of creative AI to produce algorithms which determine relevance by themselves? Essentially, which think creatively for themselves?
It’s completely possible. In fact, this type of algorithm is already frequently being used in marketing. Take Amazon for example: it uses unsupervised learning to create clusters of users which the AI defines as having similar interests. Through these AI generated groups, Amazon is able to recommend the products it considers most appropriate for the members of this artificial group.
That being said, attempts to utilize unsupervised learning for creative purposes have so far proven major failures. Perhaps the most entertaining example of this was an AI-written screenplay which analyzed the scripts of popular sci-fi films. Unlike with IBM’s trailer, this AI was given no guide on sentiment or the logic of human interaction. Unsurprisingly, the output of these self-taught rules was a series of film clichés which made no sense.
So, AI is still very much a facilitator and extension of our own creativity; not its equal, and certainly not its superior. While this status is unlikely to change in the short-term, we can’t rule out the possibility in the more distant future. For now, at least, we have full licence to enjoy our unprogrammably beautiful, creative human brains.