Looking at the History of Technological Development: Behind the Rise of 'AI Interactive Courses', Where is the Path for AI + Education?

[baoshi.rao]

Since the concept of 'AI' became popular, many industries have embraced 'AI+'. The 'AI + Education' sector has also become one of the hottest topics in recent years.

Since 2015, as AI became a hot topic, 'AI + Education' has emerged as one of the most discussed trends in the education field. Whether it's traditional education giants like New Oriental and TAL Education, or emerging online education unicorns like Liulishuo, or the widely discussed 'Fengbian Programming' in 2019, all have set their sights on the goldmine of 'AI + Education' or AIED (Artificial Intelligence in Education).

For the education sector, where teacher costs are particularly high—such as in online one-on-one teaching models—the four leading companies VIPKID, DaDa, 51talk, and VIPJR have all ventured into the 'AI + Education' space since 2017.

Recently, DaDa English was rumored to be acquired by TAL Education, while Lin Chenbin, VIPKID's head of user products, claimed that key metrics for AI interactive recorded courses have surpassed those of regular live courses. Why has 'AI + Education,' especially AI interactive courses, seemingly skyrocketed overnight?

(A Fengbian Programming ad on WeChat Moments, highlighting AI interactive courses, sparked numerous comments and exposure, far exceeding that of traditional big brands.)

Looking at the history of technological development, information technology has revolutionized industries in three major stages:

1. Replacing Manual Labor: The first stage involved using technology to replace human labor and reduce production costs. For example, computers were initially created to replace manual calculations.

   Microsoft's Windows and Office software, which accompanied the rapid development of computers, replaced paper-based documentation, storage, and retrieval, lowering labor costs while improving efficiency.

   In the education and training industry, this first manifested as optimizations in administrative management. Tools like Excel for lead grading, student scheduling, course management, and later specialized CRM and administrative software became the 'information-based' management tools still used by many traditional offline education institutions.

2. Online Transformation: The internet emerged, with platforms like Yahoo, Google, Facebook, and Amazon meeting consumer demand for information through various aggregation and organization methods.

   Businesses began leveraging websites, search engines, blogs, Weibo, official accounts, and Xiaohongshu to publish content, aiming to reach more potential customers and capitalize on the rapid dissemination of online information for marketing.

   In this stage, education and training companies reorganized traditional offline courses into digital formats (text, video, etc.) and disseminated them online, giving rise to 'online education' and 'e-schools.' Simply moving courses online, making them searchable and shareable, fulfilled the core mission of this phase.

3. Decision-Making and Personalization: The final and most impactful stage involves technology improving quality of life by aiding human decision-making or resource allocation.

   By the mobile internet era, information shifted from scarcity to surplus. From a supply-demand perspective, the most commercially valuable direction became meeting consumer needs more efficiently, thoroughly, and personally.

   How can consumers access needed information more efficiently? How can users be helped—or even automated—to make decisions?

   Examples like Amazon's smart recommendations, news feeds, and lifestyle services show that the third stage isn't just about accessing information but completing service loops—like dining, socializing, and education—online that were previously offline-only.

   For education and training, as users increasingly make decisions online, information overload intensifies competition. Institutions must constantly innovate ways to influence what users see and guide decisions in their favor.

   Meanwhile, with the maturity of mobile internet infrastructure—especially China's WeChat ecosystem, including WeChat Pay—institutions must not only market on WeChat but also complete the entire teaching loop within its ecosystem. This has spurred new models like 'micro-courses' (compared to traditional MOOCs) and rapid growth for SaaS services like Duanshu and Xiaoe Tong, which outpace traditional e-schools.

So, where does 'AI + Education' fit? It belongs to the final part of the last stage.

AI's core value in education lies in personalization, followed by efficiency improvements and cost reductions.

However, meeting user needs requires universality and service loops to precede personalization. Recent emphasis on 'study supervision' in online education addresses the lack of interaction and oversight in traditional models, which is crucial for educational effectiveness.

Only with robust supervision can online education achieve a true 'closed loop.' For example, Duanshu, a lightweight online education SaaS provider, focuses on 'AI interactive courses' to innovatively expand AI + education solutions after establishing this loop.

Thus, 'AI + Education' cannot be a castle in the air. It must build on sufficient informatization and information flow to realize its ultimate value: helping users make decisions. For education institutions, without ample online course content (data sources), how can intelligent distribution occur? Even selecting the most suitable courses for students becomes challenging, let alone delivering personalized, efficient teaching.

This explains why, despite the buzz around 'AI + Education,' only internet giants like Tencent, Baidu, and ByteDance, education leaders like New Oriental and TAL, and top players in online education niches are seriously investing in it.

Beyond high R&D costs and implementation challenges, institutions face stage-specific hurdles. From an AI perspective, algorithms are just one part; having enough source data for large-scale training is the core competitive edge every AI company seeks.

('What to Test, What to Pass,' a专升本 question bank, includes over 8,000 core knowledge points. To improve allocation and decision-making efficiency, it adopted Duanshu's AI interactive question bank to build a smart repository.)

In retrospect, 'AI Education' is a broad concept. Typical education and training processes—like research, lesson preparation, teaching, enrollment, and administration—aren't all equally improvable by AI, at least not with matching difficulty and value.

For example, marketing and research are currently almost impossible to solve with AI, while administrative efficiency gains through AI are limited.

Thus, AI aids education and training in two main ways:

1. Personalized Teaching: As mentioned, the primary benefit is 'teaching students according to their aptitude.'
   • For instance, Nasdaq-listed Liulishuo's 'Understand You English' not only assesses pronunciation and fluency but also completes the entire teaching process without human teachers. With growing users and ongoing course development, it achieves a closed loop of 'ability assessment + automatic course combination + AI teaching.'
   • Similar 'smart ability assessment + AI teaching' models are gaining traction.

According to iResearch data, the prospects for EdTech and course monetization are promising, with at least 40 companies announcing their entry into the field. Most English-language products have adopted what is known as the "adaptive education" model. In this scenario, teachers increasingly take on roles as "teaching assistants" or provide auxiliary support. With rapid data growth and continuous algorithm iterations, the feedback capability of data will strengthen over time, enabling more accurate assessments and content recommendations tailored to students' needs.

Beyond personalization, another core value of AI+education lies in helping institutions and teachers handle repetitive or similar tasks, thereby reducing costs. The key advantage of the aforementioned "adaptive education model" is that as user numbers and course volumes increase, marginal costs decrease, potentially approaching zero. The learning outcomes far surpass those of traditional online video courses, which lack interaction despite similarly low marginal costs.

In reality, high-quality teaching resources are always scarce compared to the vast number of people seeking quality education. This supply-demand imbalance is a conflict that cannot be resolved in the short term. We cannot double or significantly increase the number of outstanding teaching graduates in 5, 10, or even 20 years.

Thus, improving the utilization of high-quality teaching resources is a critical aspect of online education models. Whether through "dual-teacher classrooms," "virtual teacher livestreams" paired with "human teaching assistants," or "AI adaptive teaching" with "assistant support," the essence is to conserve teaching resources and reduce operational costs for institutions.

"Reducing teaching costs" is particularly crucial in one-on-one tutoring. Since the offline training era, the biggest challenge of one-on-one tutoring has been its high teaching costs, difficulty in scaling, and limited profitability. Consequently, online one-on-one tutoring has recently shifted toward small-group classes, primarily to lower teaching costs.

As mentioned earlier, online English one-on-one platforms like VIPKID and DaDa have long explored AI-based teaching, driven by the need to reduce teaching costs. On one hand, costs must be minimized; on the other, teaching quality and effectiveness must be maintained. Naturally, AI+teaching has become the optimal strategic direction.

As mobile internet enters its later stages, leading education companies have moved beyond "information and distribution" to pursue complete teaching cycles and personalized outcomes. AI+education is undoubtedly one of the key development trends.

However, in a highly fragmented market like education, most small and mid-sized players are just entering the third phase. While they aim to learn from the models and trends validated by industry leaders, limited resources and technology prevent them from undertaking such high-cost experiments. In this context, new players with different approaches are entering the AI+education space.

From an industry chain perspective, platforms targeting parents and students were the first to rise. As the industry grows, upstream and downstream models are emerging.

At the IaaS and PaaS layers, Tencent, after restructuring in 2018, officially launched its first education industry lab—the Intelligent Education Joint Lab—in May 2019. It offers smart education solutions like speech recognition, speech analysis, and image recognition based on its Teaching & Research Cloud.

At the SaaS layer, lightweight online education SaaS provider Duanshu is proactively entering the AI space, possibly becoming the first to offer application-level AI+SaaS services for small and mid-sized businesses. The logic is that small and mid-sized education institutions must follow trends validated by industry giants, but their tolerance for trial-and-error costs with new technologies and models is much lower. This makes AI+education largely inaccessible to 99% of such institutions.

Even though giants like Tencent, Baidu, and TAL have built AI+open platforms, most small and mid-sized education institutions lack the capacity to participate. After all, IaaS and PaaS layers require further R&D and integration before they can be directly applied to business operations.

Thus, Duanshu leverages AI algorithms from Tencent and iFlytek to create lightweight, usability-focused AI interactive course solutions. Its strategy targets mid-tier and long-tail users in verticals like spoken language, picture books, and early childhood education. However, challenges remain, such as meeting personalized needs and reducing the learning curve for clients.

While AI+education has a long way to go, and AI interactive courses are just one part of "adaptive education," it is clear that AI is currently one of the best solutions for making education more effective. Whether it's giants like Tencent, Baidu, TAL, and New Oriental, rising unicorns like VIPKID and Liulishuo, or startups like Squirrel AI and Duanshu, only time will tell their success.