Caixin
Jul 29, 2019 07:39 PM
BUSINESS & TECH

Money Pours Into Industry Claiming to Turn Kids Into Learning Machines

Many investors appear to be drawn by the potential of machine learning to improve literacy cheaply for a broad group of students while also being customized to their individual learning needs. Photo: VCG
Many investors appear to be drawn by the potential of machine learning to improve literacy cheaply for a broad group of students while also being customized to their individual learning needs. Photo: VCG

Investment continues to pour into China’s burgeoning “AI education market” — a buzz-term packed industry that claims to use machine learning and other technologies to supercharge pedagogy.

App-based homework-helper Knowbox announced this month that it secured $150 million of additional funding from investors led by internet giant Alibaba Group Holding Ltd. Yet despite the money sloshing around, the slew of companies that have entered the industry are still struggling to find sustainable business models due to high marketing costs.

It comes amid a general wave of enthusiasm for online K-12 education, which refers to the core U.S. education from ages 4 to 18. According to think tank iEDU, roughly 44 companies poured money into the “AI education” market last year, with 80% of investments amounting to more than 100 million yuan ($14.5 million) and most in the 100 million yuan to 500 million yuan range.

“We don't want to invest in small and beautiful models. We only want to invest in beautiful companies that are hopeful to grow into big trees in the future,” boasted Annabelle Yu Long, the CEO of Bertelsmann China, one of the backers of Knowbox, when speaking at the announcement of its latest funding round. Her sentiment is echoed by many of the major funders of the industry.

Many investors appear to be drawn by the potential of machine learning to improve literacy cheaply for a broad group of students while also being customized to their individual learning needs. Unlike traditional online video education, in which a student speaks with a teacher over the internet, “AI classrooms” in theory are able to monitor a student’s responses to questions, to provide interactive teaching methods. Some models are experimenting with cameras that identify students’ expressions and levels of concentration.

But offline versions of the technology, such as facial recognition systems installed in classrooms, have faced a backlash from some parents and students who say they are Orwellian and question their effectiveness.

Yet the field is still in its early stages. Many of the early leaders, such as the New York-listed TAL Education Group and Gaosi Education, are introducing “AI” technology in combination with their traditional “bread and butter” business of online tuition, focusing on language fluency and homework assistance.

One difficulty for many companies is how to bring down the cost of lessons while still making a profit, particularly as parents and students have ample choice across the traditional learning market. Knowbox is focusing on ensuring that it keeps hold of students, said Liu Ye, the company’s founder. The company’s classes now have 100,000 users, with 60% to 70% of users renewing after their first round. The company expects 60% to 70% of users who reach the second round to renew.

Yet even if the price of the courses can be reduced, high marketing costs remain a challenge. Shanghai Liulishuo Information Technology Co. Ltd., one of the first companies in the field to go public, offers low cost interactive courses at prices such as 99 yuan and 199 yuan. This helped it to gain millions of users in 2018. Despite the relatively large number of students, it still saw net losses of 487 million yuan the same year, largely due to its vast sales staff, which accounted for 70% of its total employees.

Most companies’ methods are relatively basic and crude, Guan Junhui, the general manager of internet giant Tencent Holdings Ltd.’s AI education arm said at an industry conference earlier this month.

And many companies still have insufficient data to make their classes truly customizable, with data on children’s pronunciation still relatively scarce compared with that of adults, hindering a sound assessment of students’ abilities, for example.

Contact reporter David Kirton (davidkirton@caixin.com)

You've accessed an article available only to subscribers
VIEW OPTIONS
Share this article
Open WeChat and scan the QR code
GALLERY