laitimes

The big model for education companies: a futile arms race

author:Multi-Whale Capital
The big model for education companies: a futile arms race

ChatGPT is a box that can fit everything.

Since 2023, because of the great value of ChatGPT, not only major technology and Internet companies have stepped up the layout of AI large model related businesses, but also a number of education companies have released large model products related to education.

For example, on May 5 and May 6, 2023, Xueersi and iFLYTEK officially announced the self-developed mathematical model and the Spark cognitive big model.

Education enterprises are rushing to launch education-related big model products, obviously to meet the development trend of the times, hoping to innovate the traditional education model and achieve business upgrading through AI big model technology. However, it is worth noting that AI large models require huge data and huge investment as support, and cloud computing and technology companies have the advantage of being the first to get ahead.

With the gradual maturity of the AI big models of technology and Internet companies and the vertical commercial use, it is still unknown whether education companies that only focus on the education big model track can show sufficient competitiveness.

The big model for education companies: a futile arms race

In fact, not only AI big models, in the past few years, with the gradual maturity of the Internet, big data, IoT and other technologies, relevant departments have seen the significance of science and technology for education, so they have repeatedly issued policies, hoping to further empower the education industry with science and technology.

For example, the 13th Five-Year Plan for Education Informatization released in 2016 clearly stated that by 2020, an education informatization system that "everyone can learn, can learn everywhere, and can learn at any time" and is compatible with the national education modernization development goals will be basically built.

Coincidentally, the "Implementation Plan for Accelerating the Modernization of Education (2018-2022)" released in 2019 also shows that it is necessary to promote the deep integration of information technology and education and teaching, support schools to make full use of information technology to carry out talent training models and teaching method reform, and gradually achieve full coverage of teachers and students in the application of information teaching and learning.

Thanks to the maturity of technology and the vigorous promotion of policies, the scale of China's education informatization market has steadily increased. According to data from China Business Intelligence Network, it is estimated that the size of China's education informatization market will be 557.3 billion yuan in 2023, with a compound annual growth rate of 8.3% from 2021 to 2025.

If technologies such as the Internet, big data, and IoT can improve teaching efficiency under the dual structure of teachers and students, then the AI big model may completely subvert the traditional dual structure of teachers and students, and then reshape the way knowledge is disseminated.

Taking ChatGPT-4 as an example, with more than 100 billion parameters, it scored higher than 88% of test takers in the American Bar Test, Advanced Classification Test, SAT and other tests, showing strong professional quality.

In response to the possible impact of large models on the education industry, in an interview with CCTV's digital intelligence channel "Guanlan", Zheng Qinghua, president and professor of Tongji University, said, "The big model represented by ChatGPT will change the way humans obtain knowledge. It used to be a binary structure of teacher and student. Today, machines do a better job than teachers in many areas. Therefore, the future teaching model may be expanded to a triadic structure of teacher, machine, and student."

Obviously, as AI big models mature, they will reshape traditional teaching methods and models. For example, large-scale models can monitor students' learning process, and then develop targeted learning strategies, avoiding the problem that traditional teachers cannot understand the needs of each student due to limited manpower and time.

In addition, the large-scale and online model is also conducive to breaking the hierarchical educational resources, allowing learning to truly get rid of the limitations of time and space, and realizing the same rights in education for all.

Therefore, whether it is to cater to the orientation of the capital market or to catch up with the wave of large models reshaping the education industry, education enterprises need to launch education-related large model products in a timely manner.

The big model for education companies: a futile arms race

In May 2023, NetEase Youdao, Xueersi and iFLYTEK launched education-related large model products at almost the same time.

On May 5, NetEase Youdao's official video account released a demonstration video of an AI speaking teacher developed based on the "Ziyue" large model. The video shows that NetEase Youdao AI speaking teachers can provide a variety of practice scenarios and will play a variety of roles according to students' needs, thereby solving the problem of "difficulty in speaking" that has long plagued students.

On the same day, Xueersi said that the company is developing a large mathematical model, named "MathGPT", for mathematics enthusiasts and scientific research institutions around the world, guided by problem-solving and problem-solving algorithms in the field of mathematics, and will launch product-level applications based on the large model in 2023.

On May 6, iFLYTEK released the iFLYTEK Spark Cognitive Model, which has 7 core capabilities such as text generation, language understanding, and knowledge Q&A. At the same time, iFLYTEK also launched the T20 series of AI learning machines based on the above models, which will be the first to apply large models in the main discipline of language, mathematics and English.

Comparing the landing scenarios of the large models of NetEase Youdao, Xueersi and iFLYTEK, it can be found that although they are all players in the education industry, these companies have great differences in the specific landing application of AI large models. This may be because the business models and brand precipitation of different enterprises are completely different, so they have different judgments on the landing scenarios of AI large models.

Taking NetEase Youdao as an example, it started in the translation business, and its current business focus is also strongly related to translation. According to the Q4 financial report of fiscal year 2022, Youdao smart device revenue was 407 million yuan, a year-on-year increase of 28.1%. The growth was mainly driven by the "welcome of Youdao's new products, such as Youdao Dictionary Pen X5".

It is precisely because the business fundamentals are strongly related to translation and leave the impression of a "translation company" in the consumer market, NetEase Youdao chose an AI speaking teacher for the first stop of the big model.

The reason why iFLYTEK launched AI learning machine hardware is also because education hardware is an important revenue pillar. According to the financial report, in 2022, the revenue of iFLYTEK's education products and services will be 6.161 billion yuan, a year-on-year increase of 2.56%, accounting for 32.74% of the total revenue, which is the business with the largest proportion of revenue.

Horizontal comparison can find that although all education companies see the value of large models, they are gradually differentiated in the educational application of large models, and most of them have made product layouts along the previous advantageous businesses. This strategy does help the rapid landing of large models, but it may also face the dimensionality reduction blow of "all-rounders".

The big model for education companies: a futile arms race

In this domestic model competition, not only education companies, but also technology and Internet companies are angry and strong, lest they lag behind the times.

The big model for education companies: a futile arms race

Summary of domestic large models Source: Minsheng Securities

Minsheng Securities Research Report shows that at present, more than 30 enterprises in China have laid out large models. The enterprises entering the market mainly include Internet giants, server leaders, AI companies and scientific research institutes, among which Internet and AI companies are the majority.

For example, on March 16, 2023, Baidu released a large-language model and generative AI product "Wenxin Yiyan". On April 11, Daniel Zhang, Chairman and CEO of Alibaba Group and CEO of Alibaba Cloud Intelligent Group, released the "Tongyi Qianwen" and said that all Alibaba products will be connected to the "Tongyi Qianwen" model in the future.

Many technology and Internet companies have invariably increased the size of AI models for similar reasons to education companies, and as the technology matures, it may become the "iPhone" of the new era. In this regard, in an interview with the German business newspaper "Handelsblatt", Microsoft co-founder Bill Gates said that "ChatGPT is as important as the invention of the Internet and will change the world".

Although all companies can enter the AI big model track, because related technologies require a huge amount of data and a lot of capital investment, technology and Internet giants have the advantage of being the first to get ahead compared with vertical enterprises. Minsheng Securities pointed out that "after the surface of large models is no longer scarce, the substantive pattern is more like 'Tai Chi': easy to learn and difficult to refine".

Alibaba CEO Daniel Zhang also said that large models are an all-round competition of "AI + cloud computing", and the research and development of large models with more than one trillion parameters is not only an algorithm problem, but also includes complex systematic engineering in many fields such as the underlying huge computing power, network, big data, machine learning, etc., which requires the support of ultra-large-scale AI infrastructure.

According to IDC's China Public Cloud Services Market (Second Half of 2022) Tracker, Alibaba ranks first in the IaaS market with a share of 32.6%. Considering Microsoft's huge investment in OpenAI, it is largely turned into computing for Azure cloud services. Alibaba's AI big model, which is dominant in the field of cloud computing, is obviously higher than that of other education companies.

What's more, large AI models need to be applied at scale to offset the high upfront costs and achieve profitability. Most enterprises are also planning to open up AI big model technology to B-end customers.

For example, in March 2023, before the official release, Wen Xin had reached strategic cooperation with a number of education companies such as Dane and Onli Education.

Coincidentally, Zhou Jingren, CTO of Alibaba Cloud Intelligence, also said that "Alibaba will open up the ability of Tongyi Qianwen to help each enterprise create its own exclusive model." At present, OPPO, Geely Automobile, Zhiji Automobile and other companies have announced that they will cooperate with Alibaba in large-model related scenarios.

This also means that the AI big model of education enterprises is facing a huge test, on the one hand, its computing power and investment have obvious disadvantages compared with technology and the Internet; On the other hand, technology and Internet giants may fully open AI large model products in the future, so it is difficult for AI large model products of education enterprises to have sufficient comparative advantages.

In fact, iFLYTEK has seen the above contradictions, and in addition to launching the T20 series of AI learning machines, it is also digging deep into the application of "big and complete" AI large model technology.

When releasing the Xinghuo cognitive big model, iFLYTEK said that the overall layout of the big model is a "1+N" system, where "1" refers to the general cognitive intelligence large model; "N" refers to the implementation of the large model in various fields such as education, office, automobile, and human-computer interaction. From this point of view, the AI learning machine T20 series is just a vertical application of the Spark cognitive large model.

All in all, with the gradual maturity of AI large models, the education field will become one of the main application places of this technology. The application of AI technology can make education more intelligent, help students personalize learning, and improve teaching quality. Through intelligent learning aids and personalized learning paths, students can learn according to their interests and abilities, improving learning effect and motivation.

In this context, many education companies have rushed to enter the AI big model track, and closely followed the previous advantageous business to launch related products.

Although at present, the AI big model products of education enterprises are very gimmicky, in addition, technology and Internet companies with better data and funds are also studying AI large models and hope to open up relevant technologies to B-end customers.

As the AI models of technology and Internet companies gradually mature, education enterprises that do not have economies of scale and technical advantages may also directly adopt the former's technology. From this point of view, the greater significance of the AI large model products of education enterprises at this stage may only be to appease the capital market.

Read on