In recent years, artificial intelligence has been getting more mainstream media attention than ever. Once relegated to pulp fiction and sci-fi movies, AI has become the mainstay of the tech industry, creating jobs – while conversely, feeding the fear that they are taking them.
As AI becomes more complex, the breadth of occupations at risk is growing; moving away from retail and service industries, into traditionally more nuanced jobs such as finance and as of recently, law.
But what about our little corner of the market, what about creativity? Would an AI ever be able to possess the subtlety and emotion for creative thought? To dig a little deeper into this we have to veer slightly off topic to old chess computers and the ancient Chinese game of GO.
Would an AI ever be able to possess the subtlety and emotion for creative thought?
Over the past few years Deep Mind, Google’s AI division, has been refining Alpha GO, an AI that not only has the functionality to follow direct orders, but also has the capacity for learning.
A chess computer can easily play every possible combination of moves to the end of the game and pick the best one. Simple…ish but more importantly, achievable with current technology. With GO however, it is said that there are more possible board states than there are atoms in the known universe. Let that sink in. Running every move would take more time than any of us have here, and would be far too inefficient to be deemed practical. A different tact was needed. Alpha GO had to learn how his opponent played and to do this Deep Mind utilized machine learning, which in essence is programming an AI to learn on its own. It was given a series of directives; winning = good, losing = bad, taking territory = good, losing territory = bad etc., and then was set to play itself repeatedly to learn the nuances of the game. It went up against the 18-time world champion Lee Sedol in a best of five and lost a game, but having seen how its opponent played, won the rest of the series convincingly.
This is clearly an incredible feat and one that should be applauded. However, while the path to the game’s conclusion is complex the outcome is binary – either you win, or you don’t. With creativity, the perception of the finished product can be just as intricate as the process used to create it. There are so many extraneous factors that can sway a person’s appreciation of something, be it nostalgia, taste or even tiredness. All this could change whether a given person likes, dislikes or is left cold to a certain piece. Meaning that even if an AI was able to create a perfect product, the end user may just not like it, for no other reason other than they don’t. This raises the question, could an AI ever really take the place of a human in a creative role or would we coexist, the AI standing alongside us, another creative mind to be either spurned or applauded based on the viewer’s enjoyment of their work?
There are also so many other factors needed to operate as a creative, dealing with disappointment, the ability to adapt to an ever-changing brief, to attempt to bring a client around to your way of thinking, or simply compromise.
This is all taking for granted that we could create an AI powerful enough to comprehend the task, which as it is understood at the moment, humans will never be able to do themselves. We’d have to teach an AI to learn how, then leave it to that AI to program the other; so, who knows how it might function. Would an AI even create work we could comprehend?
Facebook’s AI team (FAIR) came into an interesting discovery when trying to create a negotiation algorithm. It set two AIs against each other, in a similar vein to Alpha GO, to see how they would adapt and learn. They were programmed to negotiate using the English language, however no rule was established that stated they had to use traditional sentence structure. Very quickly, the bots veered away from what we would understand as English, using the right words but in a nonsensical order – to us at least.
Similarly, an AI named Benjamin was tasked with writing the script for a short sci-fi film; it was fed a range of sci-fi books and was tasked with writing the film. The outcome was hilarious … for ALL the wrong reasons. The final film is called Sunspring and I suggest you give it a watch.
With these two examples in mind, it might be interesting to think that when AI is powerful enough to create, it may not create for humans at all, but for other AIs. The way in which we process the world around us would be so wildly different, that what was deemed creatively correct for one may not work for another. Galvanizing AIs as a new species with its own creative needs; creative agencies, art and music tailored for them, by them.
It’s here we start to leave reality as we understand it and delve back into the realms of sci-fi; but for how long?
My guess is our jobs are safe. For now, at least.