Rarely does a routine business meeting turn out a revelation. Neither did waiting for representatives of iQiyi, China’s leading video streaming platform, in an opulent hotel bar in east Beijing portend anything but a formal half-an-hour trivia exchange. It starts as usual, with handshakes, business card swapping and ritualistic queries on whether or not we are familiar with the company’s content. Of course, we are. “The Rap of China is pretty good,” I respond with an air of a seasoned reality show pundit. Unimpressed, the iQiyi envoy continues, “Oh, but did you know that GEM (a popular Hongkongese singer) was chosen to be part of the jury by an AI algorithm?”
My whole understanding of content production came crushing down. Seems like the days of irrational decision making and stodgy brainstorming sessions are over. We are in the future indeed and we can entrust machines with something that seemed to be the last outpost of humanity in out briskly digitizing word – entertaining us.
Bear with me here. Computer games have been around for a while, but every plot turn and UX decision in them has always been determined by humans. We tend to personify the success of films, songs, books or any other possible chunks of content by associating them with their human creators. And it’s obvious that AI algorithms are also created by our own kind, but from there on they seem to branch out on their own.
“Artificial intelligence algorithms can calculate celebrity popularity, popularity, professional level, music type, fan portraits, music characteristics, word of mouth potential and other dimensions of data. After all, the matching degree between GEM and Will Pan (singer, member of Rap of China jury) is about 90 percent,” notes iQiyi program producer Chen Wei.
Not only did AI help iQiyi select a perfect jury for their show, but it also helps with editing and generating subtitles. Now the editor only needs to type in “Chris Wu smiling”, and the algorithm will immediately find and slice up a fitting fragment from the hours and hours of recordings. The company also claims that their AI deciphers subtitles with little human editing, which is pretty impressive in case of The Rap of China, a show that almost entirely consists of double-time rapping. Yet, while the latter two applications of AI make us marvel at how handy it can be, the first one might be harder to grasp.
Human creativity is what we take pride in, but with AI servicing art, arguably the pinnacle of our creativity, what has once been celebrated as a grand testimonial to humanity appears to assume a very earthly commercial tone. But what does AI actually do? Put very crudely, it analyzes numbers. The algorithms developed by iQiyi analyze personal data of actors, estimate the average success rate of their films, assemble information on hottest discussion topics online and thus make casting and screenwriting suggestions. The same mundane work has always been done by hordes of humans equipped with calculators and spreadsheets. Hardly has any film (or a TV Show) in the past 80 years been made without prior research. In this respect, AI is nothing but a weight off the studios’ shoulders, allowing them to focus on things that are actually creative – say, good quality drama.
Hollywood has also been trifling with algorithms, but unlike China, where efficiency is paramount, Americans have been more cautious about making quick conclusions on the role of AI in content production. Nowadays, however, most prominent studios are starting to succumb to the irresistible convenience of using AI, yet few would comment on it, choosing not to propagate information that could be perceived negatively by an average consumer.
Netflix, iQiyi’s main foreign counterpart, is one of the companies that did not hesitate to jump on the AI train early on and played a crucial role in demystifying and destigmatizing the use of AI in entertainment, be it in making personalized recommendations, choosing appealing thumbnails or assigning filming locations based on factors like actor/crew availability, budget, and scene requirements.
“On a film set now, it’s robots, it’s drones, it’s super high-tech, but the business side hasn’t evolved in 20 years, people use Excel and Word, fairly simplistic business methods. The data is very siloed, and there’s hardly any analytics,” says Tobias Queisser, CEO of a Los Angeles-based AI startup Cinelytic to the Verge.
Apart from Cinelync, that collects historical data about film performances to make viable marketing suggestions, a whole bunch of other American and European start-ups are also striving to carve out room for themselves in the evolving entertainment industry. ScriptBook, founded in Belgium in 2015, sells a technology that can presumably predict film’s success through its screenplay. New York-based Qloo uses AI-generated profiles deducted from scores of personal data to identify viewer preferences. Amazon, as one of the biggest AI propagators, also has stakes in the industry through services like Amazon Rekognition that help content creators conduct comprehensive video analysis, detecting objects, scenes, activities and facial expressions, indicating people, recognizing celebrities and even assessing the aptness of inputted fragments.
AI is also infiltrating music through popular digital audio workstations like Logic or Ableton where it can help musicians create drum patterns or deduct melodies from samples. It can even write music independently, and an AI music startup Amper is capitalizing on these capabilities by selling access to its algorithm to people and companies looking for simple and royalty free background tunes.
“We tend to think of technological advances as destroying what’s gone before, but that doesn’t usually happen. This could lead to a different way of making music,” once said former Pulp frontman Jarvis Cocker.
For one, AI could help us take our heads off the tedious technical aspects of content production and focus entirely on the artistry behind it. Any musician would know how exhaustingly dreary and expensive mixing and mastering are (last steps before a soup of clashing sounds can turn into a well-balanced song). Services like Landr analyze a selection of industry staples and provide users with an instant mastering solution, which is still raw, but promising.
Similar algorithms are being adopted across all content dependent industries, even book publishing and mass media. AI’s like Booxby identify and predict readers’ possible reaction to any given book based on its text. Furthermore, some major media publishers have even started using AI to help them with writing stories. Washington Post’s automated reporter managed to publish as much as 850 articles in its first year, noticeably freeing reporters of their newswriting routines. The New York Times, Reuters and other global media mainstays are no strangers to AI either, and hardly could anyone blame them.
The Terminator scenario, even if possible, is still fairly distant, with most existing AI’s merely servicing our demands. Even with the most accurate strategic advice no film production or a book could be complete without human input. And even with AI’s learning to edit videos and produce songs, it will always take a human to approve the results and add some final touches.
Like pretty much any other human invention, AI is all about efficiency and convenience. One good example is iQiyi’s ZoomAI technology that has been nothing but a boon to the Chinese film industry. Now, a 2-hour movie, processed through ZoomAI can complete repair enhancement and go online in only 12 hours. Repairing it manually would require a team of 10 people sweating for 20 days. One example of an iQiyi restored film is the 1922 classic Labourer’s Love.
“We find that a lot of the more manual, mundane jobs become easy targets [for AI automation] where we can have a system that can do that much, much quicker, freeing up those people to do more creative tasks,” said Darren Hendler of a special effects company Digital Domain in a conversation with Futurism. His firm was behind the visuals of such monumental productions as Avengers: Endgame, where among a slew of other tasks, it dealt with turning Josh Brolin into Thanos. The challenge was accomplished with a help of an AI that made the transformation seamless and allowed the actor to immediately see his altered self when he first came to set.
According to representatives of iQiyi, the company’s staff is currently roughly divided into two camps: content creators and engineers – while the former stay on guard of the human side of content production, the latter make sure that the algorithms function properly and handle repetitive insipid tasks. Ultimately this is the direction of all human progress, Henry Ford took automobile production from chaotic scurry to precise assembly lines and now robots can handle their operation. This might be the time we left the dystopian perception of technological progress behind and imagined what the world would look like if all robots from The Terminator were benevolent and highly subservient.