SCMP’s Gary Liu is betting on tech to transform the business. This is how they’re deploying it.
New offices, new thinking. SCMP moved into its new offices in Hong Kong's Times Square in February this year. (Photo supplied)
Since taking over the reins of the South China Morning Post in January 2017, Gary Liu has created a bold data and artificial intelligence roadmap that charts the course for the SCMP to become a bionic media organization.
Over the past 15 months, Liu and his team have upended the company’s old data infrastructure, built a proprietary data warehouse, introduced automated reporting algorithms, and incorporated new third-party business intelligence platforms into its system.
These transformations are already having an impact, including by galvanising the operational speed of the entire company from one that used to reflect on business metrics on a monthly basis to a digitally driven news operation that analyses and uses data in real-time.
“We’re actually very late to it. But thankfully, we’ve been able to move fast and these changes have brought about very real acceleration,” says Liu.
And that’s just the halfway mark in Liu’s plans for a SCMP metamorphosis.
South China Morning Post CEO Gary Liu is pushing the company to embrace AI. (Photo supplied)
The future, served by AI
The 114-year-old newspaper is now turning its attentions to meaningfully using data-derived insights to enhance readers’ experience on all its digital platforms, as well as deepening its AI capabilities.
On AI-augmented user experience, SCMP is already experimenting with content recommendation algorithms that are based on its internal artificial intelligence.
“Over the next 12 months, we’ll be making our user experience more personalized. We’re going to be building and maturing our content recommendation algorithms across the entire platform of all our products,” says Liu.
Incidentally, both of SCMP’s new digital news products, Inkstone and Abacus, are already powered by content recommendation algorithms. For Abacus, which covers Chinese tech for the rest of the world, the newspaper is experimenting with code that can recommend content based on users’ scroll behavior on their mobile devices.
Scrolling north-south means that a reader has reached the end of an article and that person wants to go deeper into the story, while scrolling east-west feels more like channel flipping, explains Liu.
“We’re pioneering and testing this through Abacus. If it works, you’ll start seeing that same paradigm find its way into our other products,” he adds.
Besides making recommendations, data and AI will also be used to guide the way content is written and distributed.
Writing for the world
Based on SCMP’s data, the manner in which certain stories are written determines success in different countries. For instance, shorter stories without much background and context do better in Hong Kong than in the English-speaking international markets, where more comprehensive, in-depth articles perform well.
This new insight has, in turn, spurred the company’s programmers to start experimenting with automated reporting algorithms that customise stories for readers in different markets.
These technical developments aside, SCMP is also pursuing deep machine learning — namely, natural language processing and natural language generation.
While Liu remains tight lipped on advancements in natural language, he revealed that there are plans to release several products in the coming months.
The Post moved into its new offices in Hong Kong's Times Square in February this year. (Photo supplied)
Irreplaceable humans
Despite having made quantum leaps in data and artificial intelligence, Liu says human intelligence remains sacrosanct at SCMP.
Digital tools already play an integral role in many aspects of SCMP’s operations, including its editorial functions, news distribution, scheduling, advertising and marketing. But Liu says he firmly believes that data and AI cannot entirely replace humans — even if it were a matter of business survival.
“Human judgement and accountability will always stand supreme for us,” says Liu.
“It’s a very slippery slope to say that we will one day disqualify human editorial judgement for the sake of metrics and data. We can’t be that business because that would mean that we would eventually become an irrelevant news brand that’s only chasing numbers.”
A moral code for newsrooms
On the issue of ethics and robotic journalism, Liu says that he is aware of the possibility that indiscriminate use of AI can lead to news consumers developing a myopic worldview.
He applauds the BBC and Al Jazeera for having the foresight to develop their own moral codes on data and AI, and admits that SCMP must start working on its own framework.
“We don’t have one, yet. But I agree we need it and I think more news organizations should have it,” says Liu.
“If we were to develop a data and AI moral code, the protection of diversity of thought and discourse, and the trust that readers have placed in SCMP, should be of utmost importance. Anything we do has to be in support of those values, not antagonistic to them.”