Creative AI is already here
Currently the discussion about the opportunities and threats posed by the development of artificial intelligence (AI) is boiling hot. Extreme scenarios of the future are being tossed around but no one can what the future really has in store. But we do have the luxury of knowing the past and present. It turns out that many of the things we imagined about the future have already happened!
This finding applies to creative or generative AI as well.
Although at first it seems that the capacity to create would only belong to the scope of artificial general intelligence, whereas we generally consider to be living times of “narrow” AI. General AI refers to human level intellectual capacity, whereas narrow AI would mean only a subset of human intellect. But it turns out that narrow AI is adequate to boost and augment human creativity so humans can achieve more with the help of machines.
Here’s the top of the line / three convincing examples that have not suddenly burst out from nowhere but go some way into the history.
AI actors in movies
One of the defining movie moments early this millennium was Peter Jackson’s Lord the Rings trilogy. Besides commercial success, the production was awarded for its digital effects. You could argue that these two were probably related because the AI technology underlying Lord of the Rings visual effects production helped to keep personnel costs in check and ambitious production feasible mean while adding to the visual attractiveness of the movie series.
Jackson pushed Stephen Regelous and Weta Digital studios to develop Massive, a software engine that creates computer controlled and animated characters in amazing numbers. Characters are AI-driven autonomous agents operated with fuzzy logic, the prevailing standard before the current period of AI. This software has since been commercialized by Regelous for use outside film industry, for instance to enliven architectural and engineering simulations.
Content creation for entertainment
Staying in the entertainment domain, we can spot another, more recent success which could be attributed to AI, or broadly speaking, advanced analytics. This is the case of House of Cards, a Netflix original series which they commissioned for two seasons in 2011 based on analyzing viewing behavior on their streaming platform.
Although the success of the Netflix move was largely attributed to “big data” that was still a hot topic in the early 2010’s, it clearly says the value is not in the data but in the insights. This attitude was also reflected in the New York Times story from 2013 that quotes John Landgraf, TV network executive.
“Data can only tell you what people have liked before, not what they don’t know they are going to like in the future,” he said. “A good high-end programmer’s job is to find the white spaces in our collective psyche that aren’t filled by an existing television show.” This is an excellent demonstration of how smart ways to utilize data augment human intelligence.
Netflix’s approach was to understand customer behavior with existing movie and series content in order to figure out which genres and actors appeal to their target audience. You may question whether this is an application AI at all, doesn’t it just sound like number crunching? But I would like to argue that this is part of the same continuum, although at the very start.
Composing music
Algorithmic music has long been a known concept, but it has not yet publicly taken the charts. In the curious world of popular music production, algorithms undeniably play apart but this for they get no credit for it. There are aspects of music creation that simple AI technologies, including symbolic AI, have been good for a long time. There are many musical styles that conform to patterns that can relatively simply be formalized and digital formats such as MIDI that provide an easy platform to define music.
But the difference between computer and human generated music has long remained clear.
Currently modern AI technologies are actively pursued to build AI music creation software. For instance, it has been speculated whether Spotify is utilizing AI to create content for some of its automated playlists just to keep royalties from ‘Meditation and relaxation’ background playlists for themselves.
Outside the chart topping this, there is public and legitimate music creation going on. At least two startups, Jukedeck and Amper are currently making progress in the domain computer generated music for the purposes of background music, opening themes and other applications were music is a central but not the primary element.
Listen to a track generated by Jukedeck:
Convinced?
Examples of AI applications in the creative domain range even further than these. For instance, marketing and advertising increasingly utilize AI. Literary prize has also been (almost) awarded to an AI authored text. Robot journalism is developing quickly and robots are already supplementing human reporters in certain domains.
This article opened with several images from American artist Robbie Barrat who has explored AI technique called generative-adversarial networks (GANs) to create series of “nude paintings”. Although his experiment didn’t quite match classical masterworks, the output is visually very interesting.
Going through all of creative AI examples will not change the take away here:
AI technologies are quickly ripening to seriously disrupt creative work
We at Fourkind are looking forward to latter part of 2018 to begin multiple project on creative AI. The train is moving, jump along or you’ll be late!
Thanks to Robbie Barrat was permission to include his artwork here
Learn more
Word Economic Forum’s report on emerging technology’s impact of creative economy
https://www.weforum.org/whitepapers/creative-disruption-the-impact-of-emerging-technologies-on-the-creative-economy
Check out the Stephen Regelous explaining the construction of artificial brains created for the Two Towers movie:
https://www.youtube.com/watch?v=WXqNO9Yi2ZU
Very good description of how Netflix created House of Cards can be found from a TEDx talk by Sebastian Wernicke
https://www.ted.com/talks/sebastian_wernicke_how_to_use_data_to_make_a_hit_tv_show#t-226877
