In China, where many parents still believe that going to a good university is a sure bet for a successful life, little expense or effort is spared to prepare their children for the annual university entrance examinations.
Known as Gaokao in Chinese, the exams take place over a few days and tests each June, quizzing students on subjects that include Chinese, mathematics, the sciences, and a foreign language. The combined score, similar to the Grade Point Average (GPA) in the US, is then used to rank and corral students into different universities. Top institutes like Tsinghua and Peking University, like Harvard and Yale, get the ones with the highest GPA-equivalent scores.
But what if students could get the help of robot teachers to help them prepare for the examinations? Machine tutors that know the students “much better than they understand themselves” – in the words of Yixue Education co-founder Li Haoyang.
In a competition between Yixue Education’s artificial intelligence-backed software and teachers with an average teaching experience of 17 years, the machines were able to improve test scores more than their human counterparts. The contest, which involved 78 students, took place in October in Zhengzhou, Henan province, under the supervision of the local education bureau and was monitored by internet analysis firm iResearch.
“Traditional teaching in the classroom is very low in efficiency as the weaknesses of each student varies,” Li, who declined to give his age but should be in his late 30s, said in an interview in Guangzhou. “Many students waste time learning knowledge that they have mastered repeatedly, as the teachers want to take care of those who have yet to know it well. But even then, slower learners could still be left behind as teachers might not be able to spend enough time to help them.”
Yixue’s learning methods use AI to identify a student’s weaknesses, then tailors the tutoring to reinforce those areas, according to Li, one of four co-founders of Shanghai-based Yixue. For now, the software is geared toward students from elementary level to high school, and the company plans to introduce courses for students taking the Gaokao next year.
Yixue is competing with other start-ups like Hujiang EdTech for a slice of an expanding pie in China’s online education industry, which iResearch projects will increase more than 70 per cent from last year to 269.26 billion yuan (US$41 billion) in revenue in 2019. Other educational start-ups in China include Liulishuo, which uses AI-powered software to teach users English.
After four days of intensive training, students using Yixue’s software saw their average mathematics test scores increase 36.13 points, compared with 26.18 points for students tutored by the human teachers, according to the company.
Henan, where the Yixue trial was held, is the most competitive province in China for a college position in the country. This year, over 860,000 students took the Gaokao in the province, but only 10 per cent of them – the lowest percentage in the country – were able to go to “key universities” that are regarded as well-known and receive a high level of government support.
For Yixue, which raised 270 million yuan in an early funding round in mid-2017, the next step is to expand to Hong Kong and Taiwan, followed by Southeast Asia and eventually to Europe and the US, according to Li. The software is adaptable to students worldwide regardless of language, he said.
The company currently has more than 100,000 paid users, with about 10,000 spending more than 10,000 yuan each on the platform, according to Li. Some students can spend as much as 50,000 yuan as they take multiple courses. Gross profit margin now stands at about 80 to 90 per cent, though that may change as Yixue explores plans to open 2,000 physical training centres to serve parents who are still accustomed to sending their children to a physical classroom.
Yixue was also recently in discussions with Hong Kong-based firm Hanson Robotics, which is behind the Sophia humanoid robot, to create human-looking droids to interact with students in a classroom setting, Li said.
“The children really have it very tough,” Li said. “By building this system, I hope to be able to free them.”