Ping An Technology is leveraging AI music to boost stickiness of its financial products
Is nothing safe from the reach of the machines? After replacing humans at repetitive tasks, beating them at games of skill, artificial intelligence is now making inroads into the arguably the last frontier: the creative arts.
Ping An Technology, a subsidiary of China’s second-largest life insurance company, has developed an award-winning computer algorithm that can generate original melodies after “studying” hundreds of pieces of piano music.
But the company has no intention of stopping there. It is working towards using AI to create virtual singers and write original pop music based on individual tastes with a longer term goal of composing a symphony to potentially rival Beethoven.
“Be it a Chinese version of break-up songs similar to Taylor Swift’s style or a Shanghai accented voice that raps like Justin Bieber, with the upgrade of the AI music system we will be able to create whatever music you like as long as we know the genre and the type of singers you adore,” said Xiao Jing, chief scientist of Ping An Technology.
Founded in 2008 as an in-house provider of information technology services for companies within the Ping An group, Ping An Technology has been betting big on AI, including development of applications such facial recognition and voice recognition.
The new mission of the tech arm is to not only boost the group’s own business but also to export its know-how to other financial institutions, public utilities and manufacturing firms that are adapting to the digital era amid Beijing’s Internet Plus strategy and China’s national push to be a worldwide leader in AI by 2030.
The centrepiece of Ping An Tech’s development plan is the application of AI in music composition because music and songs are seen by the insurance conglomerate’s executives as a powerful weapon to boost “user stickiness” among its 400 million plus clients who mostly turn to the group for occasional financial and insurance-related services.
“Music is something people of all ages enjoy on a daily basis, so our idea is to use AI to easily create thousands of virtual singers in pop music and eventually upgrade the system to make symphonies to rival professional musicians within three years,” said Xiao, who laid out the blueprint for AI-generated music in an interview at Ping An group’s headquarters in Shenzhen.
Around 600 to 700 of Ping An Technology’s 8,000 employees are working on AI and big data. Without revealing the level of investment in AI, Xiao said more than 50 billion yuan (US$7.9 billion) has been spent on R&D over the past 10 years, producing a range of innovations to increase its business efficiency and cut costs.
In its office building, a landmark 600-metre tower in Shenzhen’s central business district, all employees use face ID to gain entry, a three second process powered by the company’s self-developed facial recognition technology.
AI-enabled solutions have also been applied in its customer service centre, which receives more than one million calls daily. Bots process some of the customer requests and AI-enabled voice recognition can even gauge the emotional state of the caller. A protocol will be triggered if the machine determines that a calling customer is “upset enough”.
Despite being a late mover in the music industry, Xiao said Ping An Technology is not worried about the established music live streaming sites operated by internet giants such as Tencent Holdings and Alibaba Group Holding, owner of the South China Morning Post. While these sites have already made AI-enabled music recommendations the norm in China, they are not yet using AI to compose music.
“Their advantage is based on the large library of copyrighted music they spend big money on. That advantage will diminish when our AI generates good original songs on a large scale,” said the 46-year-old Xiao, who is part of China’s 1,000 Talent Plan, a Chinese government initiative to lure top notch experts to return to China to help ignite the country’s innovation efforts.
Before joining Ping An Group in 2015, Xiao worked as a senior scientist for Microsoft and Epson.
Xiao’s confidence in AI music was boosted by the first place award Ping An Technology won at the International AI Music Composition Competition hosted by Switzerland’s École polytechnique fédérale de Lausanne (EPFL) in January. The event, which drew more than 100 teams from across the world, saw professional human composers judge a variety of AI-created original music ranging from jazz and blues to classical and pop.
Not surprisingly, songwriters are cool to the idea of machines creating music. Peter Li Si-Song, a Singapore songwriter who has produced albums for Asian stars such as Jacky Cheung, Sandy Lam and Stefanie Sun, said there is line between leveraging technology and being lazy.
“In music composition, you still need to use your own emotion and feeling to create something to wow yourself. Relying on new technology is not going to do that,” he said in an earlier interview.
Although an AI system developed by Google’s Deep Mind unit beat the world’s top Go player Ke Jie in a 3-nil victory in 2017, Xiao said that task would be much easier than Ping An Technology’s ultimate goal of having its AI compose a Beethoven-type symphony because it is difficult to “teach” machines good music from bad.
However, Xiao said music composed by AI technology could find wide application in films and television as it could create unique and recognisable soundtracks in a much shorter time frame. However, Xiao believes AI won’t put composers out of a job. “It could provide human composers a broader pool of inspiration from which to draw, but will never replace professionals,” he said.
“Creating music [with AI] still needs to imitate exiting genres. Only humans can create genres we have never heard before.”