Caixin
Nov 07, 2020 09:00 AM
OPINION

Opinion: How Artificial Intelligence Can Democratize Education After Covid

Students experience AI-based music education in Hai'an, Jiangsu province, on Aug. 26.
Students experience AI-based music education in Hai'an, Jiangsu province, on Aug. 26.

Craig Smith is a former New York Times correspondent and host of the podcast, Eye on AI.

The ongoing Covid-19 pandemic, which has disrupted classroom instruction around the world, presents an historic opportunity to democratize education with artificial intelligence.

Since Socrates taught Plato and Plato taught Aristotle — or Confucius taught Yan Hui — man has known that the best education is delivered one-to-one by an experienced educator. But that is expensive, labor intensive and cannot scale. The result is the imperfect classroom-based instruction that we live with today.

Yet, new forms of AI, based on deep neural networks, can now uncover patterns about how students perform and help teachers optimize their strategies accordingly. “AI tutors,” software systems that students interact with online, can give every student greater access to the individualized attention they need. These advances, together with widespread internet use, can remake education as we know it.

“Future learning environments will certainly leverage AI,” said Paul Kim, associate dean and chief technology officer at Stanford University’s Graduate School of Education, adding that AI’s impact will range “from student counseling to student project assessment and from achievement predictions to program planning.”

Kim is part of an effort to make AI solutions available on portable devices through the Stanford Mobile Inquiry-based Learning Environment, or SMILE, software that helps engage students in inquiry-based learning sessions and generates real-time learning analytics.

He predicts that most educators will welcome such changes, but warns that the current digital divide will mean that only those with access to advanced technology infrastructure and personal devices will benefit.

“Addressing these immediate challenges should be central to the discussions among educators of the 21st century,” Kim said.

The earliest computer tutoring systems appeared in the 1960s, but they were expensive and did not move beyond well-funded research institutes. By the 1970s and 1980s, however, researchers developed new tutoring systems using rule-based artificial intelligence that led students through lessons, giving them hints. But those systems failed because they were expensive to program and maintain.

Since then, most computer teaching systems have been based on “decision trees,” elaborate forking learning paths that send students in one direction or another depending on their performance.

But the recent AI revolution, based on networks of algorithms that learn over time, has changed that. Today, “deep learning” algorithms can discover patterns in an individual’s performance and optimize teaching strategies accordingly.

The best of these systems can raise student performance well beyond the level of conventional classes and even beyond the level achieved by students who receive instruction from human tutors. AI tutors outperform, in part, because a computer is more patient and often more insightful.

One of the first commercial applications of machine learning to teaching was by a company called Knewton, founded by a former executive at the private education giant, Kaplan Inc. After answering a few questions, Knewton can determine a student’s level and deliver appropriate content. But Knewton ran into financial difficulties and was sold in May 2019 to the education publisher John Wiley & Sons for a tenth of what investors had poured into it.

Since then, South Korean startup Riiid has taken the idea further, developing a suite of algorithms that track student progress, sense when students are bored or frustrated and optimize content to keep them engaged and motivated. Their English-learning app has been used by more than a million students in Korea and Japan, allowing them to amass what they say is the world’s largest dataset of student-AI interactions.

Armed with the validation provided by their English-language app, the company is now providing backend solutions to private education companies and education administrations in several countries, including the U.S.

“We originally thought that our technologies were specifically for the test-prep market, but right after Covid-19 struck the world, the education ministry of the UAE and one of top technology companies and other non-test prep, educational organizations contacted us to ask how they can use the AI technologies that we have developed,” said YJ Jang, Riiid’s founder.

To further research in the field, Riiid has launched a global competition – the Riiid AIEd Challenge – to engage global AI talent in designing better teaching algorithms. The inaugural challenge, announced in October, invites teams to use Riiid’s EdNet dataset to build ‘knowledge tracing’ algorithms for Google’s Kaggle competition platform. More than 1,000 teams are participating so far.

The top five teams, in what is expected to be an annual competition, will share $100,000 in prize money and have their solutions presented at a workshop at the Association for the Advancement of Artificial Intelligence conference in February next year.

Such challenges have been used in the past to focus the world’s best academic minds on difficult problems, and are responsible for some of the most significant advances in artificial intelligence. The ImageNet Large Scale Visual Recognition Challenge led to breakthroughs in computer vision, for example, and the US DARPA Grand Challenge kicked off the development of self-driving cars.

“As an adviser to this important event, I look forward to reviewing AI-backed solutions that can give educators insightful directions to best coach young minds,” said Stanford University’s Kim.

Riiid is not alone. A growing number of startups are now drawing on academic work to create commercial solutions that they hope will remake education as we know it. Market research firm HolonIQ predicts that more than $6 billion will be spent on AI education by 2025.

China is leading in AI education with more than $1 billion invested in the sector so far. Tens of millions of students now use some form of AI to learn. The most prominent player is Yixue Education’s Squirrel AI, which has more than 2,000 learning centers in 200 cities across the country. The company has partnered with Carnegie Mellon University to create the CMU-Squirrel AI Research Lab on Personalized Education at Scale.

Yoshua Bengio, one of the world’s most prominent deep learning researchers and a recipient of the 2019 Turing Award — the Nobel Prize of computer science — has spun off a company called Korbit together with a researcher at Cambridge University and one of his students at the University of Montreal.

For now, Korbit is focused on teaching machine learning, using machine learning. But Bengio envisions a future when such AI-powered systems can teach any subject to any student with an internet connection.

“The potential impact of these kinds of technologies could be huge,” Bengio said in a recent telephone interview. “There are just not enough skilled educators to address the needs of a large number of young people around the world.”

Everyone involved in AI for education stress that AI tutors are intended to be a teacher’s aide, not a replacement for teachers.

“What you will have is a human teacher who helps a group of a certain size but then each of these individuals can get personalized guidance,” said Bengio. “That personalization is really where machine learning comes in.”

Researchers are now working on incorporating chatbots and even realistic human avatars into these tutoring systems to make the teaching, now done primarily through text, even more engaging. Bengio likens the current AI tutors today to the clumsy cellular telephones of the early 1990s. “What you're seeing is the first generation,” he said. “It’s going to take decades to refine them.”

Nonetheless, AI and the internet hold the promise of democratizing quality education. By focusing the best AI minds in the world on the improvement of AI-enabled education, researchers and teachers together can design a new paradigm for education in the post-Covid era.

The views and opinions expressed in this opinion section are those of the authors and do not necessarily reflect the editorial positions of Caixin Media.

If you would like to write an opinion for Caixin Global, please send your ideas or finished opinions to our email: opinionen@caixin.com

Support quality journalism in China. Subscribe to Caixin Global starting at $0.99.

loadingImg
Register to read this article for free.
Register
Share this article
Open WeChat and scan the QR code