Inside the offices of Toutiao, it’s onerous to inform you’re simply outside of Beijing and not within the heart of geographic region.
The walls area unit embellished with fashionable graffiti, and there area unit all the standard perks: a rumpus room, a gym, and a luxurious eating house. however Toutiao doesn’t solely look the half. the corporate makes Associate in Nursing insanely standard Chinese news aggregation app referred to as Toutiao (meaning “headlines”). Toutiao boasts some 600 million users in China, with quite sixty million active daily users. the corporate gets most of its revenue from advertising, that is additionally tailored to users’ likes and interests.
Toutiao’s parent company, Bytedance, has adult at a dramatic rate. In 2014 Bytedance wanted investment that valued it at concerning $500 million. With its latest fund-raising spherical, the corporate is reportedly seeking a valuation in more than $10 billion.
The success of Toutiao has been achieved, in massive half, by mistreatment machine learning to work out users’ interests and tastes, and craft its offerings consequently to urge additional clicks. Facebook and Twitter conjointly use machine learning to refine recommendations, however they believe additional heavily on social connections.
When you open the app, you’ll see a stream of posts, starting from serious news reports to cute videos. The additional you click, the additional Toutiao learns your tastes and interests, and refines its recommendations. Toutiao was created in 2012 by Zhang Yiming, a 33-year-old bourgeois.
When I 1st opened the app, i used to be conferred with a broad vary of content. the highest story was a post concerning coming back economic reforms. This was followed by a bit on the amount of pollution in Beijing that day, and so a video of a preferred slapstick comedian talking concerning the air pollution. more down was a salacious story a few noted histrion having Associate in Nursing affair, and, curiously, a piece of writing on the way to recognise the sex of newborn chickens (I’m told this was aimed toward rural readers).
In the coming back year, Toutiao aims to develop additional AI-focused services, and to expand internationally. the last word goal is to become the world’s number-one on-line content destination, eclipsing the likes of Facebook and Twitter, likewise because the likes of the big apple Times and Buzzfeed in their own markets.
Indeed, signs of high-tech hustle were everyplace once I visited last month. lots of coders, designers, and marketers were busy operating away during a trendy, unrestricted house, with desks organized around a futuristic-looking circular glass meeting area. In one corner, a pile of many hundred unpacked ThinkPad laptops Saturday prepared for brand new staff to say.
I met with Lei Li, WHO heads up Toutiao’s AI laboratory, having joined the corporate from another Chinese net company, Baidu, in March of last year. Besides developing higher ways that of crucial what posts to suggest to users, Li’s team has created a man-made communicator, referred to as Xiaoming, that’s capable of generating short news reports on European football games. Li’s team is additionally performing on ways that to spot users that may be ready to answer specific queries, with a read to giving a service the same as Yahoo Answer or Quora. Toutiao opened this drawback up to outside researchers with a contest control in Dec.
“We really need to use machine learning to grasp question-and-answering and comments and news,” Li told Pine Tree State. “Toutiao Lab’s mission is to push the boundaries of machine learning and AI and to transfer that into product.”
Toutiao already has many ways that of distinctive pretend news. In China this principally consists of dubious health recommendations. It will this employing a mixture of human reviewers in conjunction with machine-driven analysis of posts and comments associated with them.
Li conjointly argues that Toutiao’s method of egression content results is a smaller amount of a filter bubble than is commonly seen on Facebook, wherever folks area unit usually exposed solely to posts from those with whom they already agree. This involves coming up with the advice algorithms to supply some content albeit a user hasn’t signaled a preference for it to this point. “Our [machine learning] models can try and expand your horizons, your interests,” says Li.
Jie Tang, a academician at China’s Tsinghua University WHO studies data diffusion and social network dynamics, says as Toutiao becomes additional standard it'll become more and more vital for it to form recommendations that area unit correct and compelling. He says the corporate may additionally ought to generate additional of its own content. “You would like one thing new,” he says.
Toutiao is already performing on this. Besides linking to outside content, a growing range of posts and videos come back from those posting directly on Toutiao’s platform. Users embody news retailers, government agencies, and celebrities.
To some extent, the emergence of Toutiao symbolizes China’s transition from a low-wage, low-tech producing powerhouse to a additional technologically subtle market. fitly enough, the company’s headquarters area unit placed in what accustomed be Associate in Nursing region producing industrial plant.
Bytedance additionally offers associate degree English-language app referred to as TopBuzz. Its future strategy can embrace finance in and partnering with native media firms. Last year it endowed $25 million in associate degree Indian human referred to as Dailyhunt. the corporate is additionally longing for revenue opportunities outside of advertising. Last Sep, Toutiao signed a affect JD.com, a Chinese e-commerce website, permitting users to shop for merchandise through the app. Current investors in Toutiao embrace cypress Capital China and therefore the Russian have Yuri Milner.
Ultimately, Li says, the company’s efforts to expand internationally are going to be hep by the work his team is doing. “Many of the teachings we've learned will be shared for our international merchandise,” he says. “We have designed the biggest machine-learning platform for content. That’s our weapon.”












