AI and education collide, enlightening a new era of education. At present, artificial intelligence has not only become a catalyst for the development of education, building a bridge between reality and virtuality, but also leading educational institutions into a new era of intelligent learning, lighting up the direction of personalized education, and cultivating multi-functional talents with independent thinking and full of innovative spirit for the country. Under the wind of science and technology, what kind of sparks will AI and education collide? How will the big model subvert the educational learning products and industry chain?
On the afternoon of June 14, the fourth forum of "Discovering New Forces in Education", co-hosted by Tencent Cloud, Tencent Education and Duowhale, was held online. At the fourth forum, Bao Jianwen, founder of Imagination Education, Zhou Qiang, founder of Rush Exam, Lu Shenbiao, founder and CEO of Mr. Sai, Tan Meihong, CEO of Qudubbing, Zhao Song, dean and co-founder of Deep Blue College, and five experts and scholars, business leaders and innovative elite guests from various fields such as quality education, language education and vocational education gathered in the cloud to discuss the application of artificial intelligence in education teaching and talent training online.
As a global English learning institution featuring intelligent teaching, Imagine Learning has been conducting research on artificial intelligence knowledge learning since 2004. By 2014, in just a decade, Imagine Learning had more than 15,000 campuses worldwide and a market capitalization of more than $10 billion. In 2014, Imagination Education began to cooperate with Imagine Learning to introduce and localize development. By 2019, Imagination Education has more than 730 large-scale campuses in China, serving more than 240,000 students, forming a highly intelligent and internationally integrated intelligent campus network.
(Bao Jianwen, founder of Imagination Education)
Bao Jianwen mentioned that as early as 2001, he founded the world's first intelligent teaching platform for Chinese learning in the UK. After returning to China, he created an industrialization platform with innovation and creativity as the core. This spirit of innovation is also infused with the cultural blood of imagination. All educational materials of Imagination Education are original to the company and have hundreds of copyrights, trademarks and patents. Bao Jianwen was personally named one of the ten entrepreneurs in "Looking for Future Unicorns". All kinds of honors are added, and what Bao Jianwen is most happy to see is the substantial growth of students. He mentioned that in the process of funding and helping children in poor areas over the past ten years of imagination education, the power of emerging technologies such as artificial intelligence has been indispensable.
The intelligent teaching of imagination education is embodied in many aspects. The self-developed artificial intelligence adaptive system provides students with a personalized English learning experience, "100 courses for 100 students". In terms of teaching resources, students' speech recognition and human-computer dialogue interaction have become the highlights of stimulating children's interest in learning. Bao Jianwen showed an animation. In the picture, when the child says "sit down", the little ladybug will sit down in the room "following orders". Cartoon-style animation and intelligent interactive learning methods not only cultivate students' ability to learn independently, logical thinking, social communication and other aspects, but also make the English learning process full of fun and imagination.
Bao Jianwen pointed out that in the current national policy, it is clearly mentioned to "actively explore the use of artificial intelligence technology" and improve school education "with experiential, inquiry-based, project-based and comprehensive learning as the main methods". Therefore, Imagination Education has developed an interesting and effective learning system and intelligent teaching aids for young children, and supplemented offline teaching with emotional companionship experience, creating a "warm" learning system of "intelligent system + teacher", which has ushered in explosive growth in business.
Bao Jianwen believes that the biggest charm of intelligent systems is that they can achieve "teaching according to aptitude" and carry out personalized guidance for each student's learning situation. The intelligent learning system built with massive data can sort out the learning map of students' knowledge points by difficulty and level, which can not only help students consolidate weak links, but also help students draw inferences from problem-solving methods and skills, and reduce the burden on students' parents. To build such an intelligent platform, Bao Jianwen showed in the video that there is no shortage of a high-precision, passionate and dynamic international education talent team. For students, the warm and contexted teacher care like "school managers" is also the nourishment of "emotional attributes" necessary for the growth process, and it is the content that imagination education attaches importance to in addition to the development of intelligent teaching systems.
Finally, Bao Jianwen introduced the application of ChatGPT in English teaching research and development. As early as 2016, the company invested tens of millions in the research and development of a human-machine dialogue system based on neural networks. After withstanding the impact of the epidemic, more advanced human-machine dialogue systems are now being developed. The emergence of a large number of open source ChatGPT has greatly reduced the capital and time costs of manual production of picture books and animations, and greatly improved the production efficiency of content. If ChatGPT, a powerful tool, is incorporated into an intelligent teaching and service platform, it will greatly promote the learning effectiveness of education. However, Bao Jianwen also observed that the production content of ChatGPT is not effective in children's emotional education and character cultivation, which requires giving full play to human initiative, stimulating children's creativity and imagination, and adding emotion and warmth to education.
As a leading intelligent learning platform for all subjects in China, the champion has a research and development and application foundation for intelligent learning systems for many years. Zhou Qiang pointed out that the future market of the education and training industry will be a market that combines intelligent AI equipment and comprehensive teaching mode. Today, AI technology has been very developed, and technical barriers that were previously considered unattainable have been broken, including the now well-known OCR recognition technology, TTS real human voice technology, and the latest AI knowledge point labeling technology, sitting posture detection technology, NLP speech recognition technology and AI topical technology. These technologies combine large-model algorithms with low-level development to produce many amazing application scenarios.
(Zhou Qiang, founder of the Champion Examination)
Zhou Qiang introduced several important scenes. First of all, the champion developed a knowledge graph learning method, which deeply labels children's knowledge points through large models and automatically recommends corresponding weak links. Red and yellow indicate mastery, green indicates mastery, and personalized learning can be carried out to achieve accurate learning schedule. Secondly, an intelligent assessment system has been developed, which can be tested according to the child's version, understand their mastery of each knowledge point, and automatically distinguish the difficulty factor. The system can also analyze the distribution and progress of knowledge points, and can also compare it with the whole country. Large models can also predict children's progress by training algorithms, providing targeted suggestions and random training. In this way, children can "spend time on the cutting edge", consolidate weak knowledge points, selectively skip the unwanted parts, and achieve efficient learning.
Zhou Qiang mentioned that this is the concept of adaptive learning, which has existed for a while but has been well applied in the education and training industry.
Then, Zhou Qiang introduced the application of AI intelligent test paper analysis system and essay marking system. He said that Rush Champion was one of the first companies to propose the concept of intelligent paper analysis system. Different from the traditional test paper correction, the system recognizes the child's test paper by taking photos, and realizes automatic cutting and labeling. The system will simulate the role of special teachers, analyze the weak points in the test questions, and automatically add the wrong questions to the wrong question book. Raise label accuracy to a high level with an intelligent speech recognition system.
For example, it can achieve 98% accuracy in science, which is much higher than the accuracy of traditional search models. In addition, the intelligent paper analysis system has multidisciplinary and interdisciplinary capabilities, covering mathematics, politics, history, geography, science, biology and other disciplines. In the liberal arts, the Chinese composition marking system can quickly detect typos, identify excellent words and sentences, and give points. The system can also generate a comprehensive evaluation report, including tests in multiple dimensions such as relevance degree, central prominence, content enrichment, and mental health. Similar systems are also applied to English composition grading and oral assessment, realizing a comprehensive assessment system.
Refinement and personalized inspection and filling of gaps have brought subversive innovation to education and learning. These intelligent education solutions have been applied in practical scenarios, bringing significant effects to students' performance.
In addition, the champion is developing a large model system for teaching and research. The system can automatically upgrade lesson plans through AI large models and automatically write course scripts. In addition, you can also combine virtual AI characters to transform the appearance of teachers through technical means to create a cool and handsome virtual teacher image. The system can also quickly carry out the question table system and convert text to audio through TTS technology. The role of the virtual AI teacher can facilitate the learning process, greatly improve the efficiency of the teacher's classroom, and optimize the lesson production experience.
Zhou Qiang revealed that the self-developed AI topic technology has entered a later stage and will be launched soon. The technique includes explaining children's frequently frequently used questions and using conversational interactions. The system will not only inform the child of the correct answer and ideas, but also allow for simple emotional communication.
Finally, Zhou Qiang introduced that AI has been commercialized in many aspects in the field of education, and more than 30 AI tools have been applied to intelligent learning systems, including online learning and traditional teaching. At the same time, the champion also plans to combine the underlying capabilities of the large model with its own content and secondary training, and has achieved a number of technological breakthroughs.
Recently, the Ministry of Education and 18 other departments jointly issued the "Opinions on Strengthening Science Education in Primary and Secondary Schools in the New Era", and in-depth science education was put on the agenda. As a key part of STEM education, policies help give wings to the development of STEM education. Under this trend, how to make STEM education software and hardware parallel and develop the scale of the industry in a healthy and standardized manner?
As an education company with quality education as its business, Mr. Sai has been working in the STEM market for many years.
Mr. Sai's founder and CEO Lu Shenbiao suggested that the first thing entrepreneurs need to think about is the potential of the STEM education market for entrepreneurs preparing to enter the STEM industry. He began by asking a key question: How big is the market for STEM education? He pointed out that the root cause needs to be found in the communication with parents, salespeople, teachers and class teachers, because they are part of the actual operation of the industry and can provide more valuable feedback. He believes that there are three questions to assess the awareness and needs of clients and parents for STEM education. The first question is what is STEM education; The second question is why children need to learn STEM today; The third question is why start learning now. Parents who can answer these three questions have more potential and motivation to invest in their children's STEM learning.
According to Lu Shenbiao's years of practical experience and exchanges with peers, he pointed out that the current STEM education market is mainly concentrated in first-tier cities such as Beijing, Shanghai, Guangzhou and Shenzhen, especially in East China. At the same time, South China and second-tier cities are also gradually rising. He expects that the STEM market may usher in rapid growth like the subject market due to the encouragement of quality education policies, the interaction mode inside and outside schools, the improvement of the average level of parents' cognitive level and other favorable factors.
Then, Lu Shenbiao discussed how to choose entry points based on their own advantages in the STEM education market, and compared the online and offline models.
(Mr. Sai, founder and CEO Lu Shenbiao)
For the offline mode, Lu Shenbiao mentioned its advantages of scenario-based. Offline education can provide richer scenarios and diverse experiences, while also building trust through face-to-face interaction. Physical stores provide more opportunities for offline education to have face-to-face contact with customers, more platforms and channels, so offline education has more diverse ways to acquire customers than online.
For online education, Mr. Lu Shenbiao emphasized convenience and cost advantages. Online education needs to meet the needs of parents for greater convenience, effort and economy, because parents generally believe that online operating costs are low. In addition, the advantages of online educational institutions themselves are that it is easier to manage teams and organizations, and it is easier to organize the supply chain.
Whether online or offline, it will be key to be able to provide a superior feeling that surpasses others in terms of experience or learning outcomes. Invest in the construction of experience scenarios, such as the equipment and space design of offline education venues, to provide immersive experiences that can impress students and parents. In addition, "visible learning effects for parents" should be given to continuously attract parents. In these respects, the driving force of industrial change and technology is indispensable.
As an online education institution, Mr. Sai is actively exploring more new areas to integrate to improve the quality of teaching. First of all, through artificial intelligence technology, the customer service has been partially replaced, replacing some customer service and class teacher work. From the perspective of teaching, Mr. Sai also introduced artificial intelligence teaching assistants to help teachers achieve better teaching results. Secondly, it focuses on the student experience and explores how artificial intelligence technology can become a learning companion for students, allowing them to play the role of students' invention assistants and encyclopedic companions.
When it comes to the market size of STEM education, he believes that it is small compared to subject education. Therefore, it is more important to fully integrate and embrace each other to maximize their respective advantages. Offline, you can showcase products, connect with customers, provide sales services and do some teaching work, as well as host events and events. In addition, since STEM education requires a lot of education and materials, offline can also take on some logistical tasks. Online enables better curriculum delivery, ensures learning outcomes, and builds learning communities that allow more children to have nationwide discussions, not just within classes or campuses.
In such a market environment, STEM education needs to think about how to use technological developments to improve business operation efficiency and user experience. AI replacing teachers is the first step, through AI to achieve better learning results and personalized learning than real teachers. At the same time, we should also give full play to the role of real teachers and empower teachers with science and technology. On the other hand, in terms of hardware and supply chain, upgrades and improvements through technological development can be considered, such as the application of smart hardware in the field of piano and fine arts.
In terms of the hardware market, Mr. Sai is pushing to lower the "threshold for trying" for customers, including the threshold for awareness and purchase, to stimulate the interest of children and parents. Therefore, Mr. Sai chose to enter the market through hardware and use content as a means of service. Hardware products demonstrate learning effects and abilities, while content retains users and stimulates children's interest in self-learning.
Lu Shenbiao believes that compared with the traditional education and training industry, STEM tracks do have huge development potential and growth space. As the degree of standardization increases, so does the degree of market concentration. In the long run, there are several categories of businesses that stand out from the competition.
First, it is a technology-leading enterprise. This type of enterprise has advanced technology or teaching and research capabilities, and can invest the best quality products in the industry. Such enterprises will definitely be able to get a piece of the market.
Second, cost-leading enterprises. It has significant advantages at the end of the supply chain, and the products have a high cost performance. This enables businesses to serve a wider range of markets and users.
Third, it is a user-friendly enterprise. They are able to meet the needs of most people in the industry and provide personalized service. Such businesses have an advantage in building close relationships with users, who are more likely to trust such teachers, brands, or institutions than large brands.
Based on the above three points, Lu Shenbiao suggested that when choosing the future development direction, STEM entrepreneurs can choose the most suitable direction according to their own advantages and market demand.
With the popularity of ChatGPT, this technology also brings new opportunities for education, which is gradually changing the education landscape, promoting positive changes in knowledge production methods, and promoting the sustainable construction of digital transformation of education. The deep integration of ChatGPT and the education industry, human-computer collaboration helps the intelligent transformation of English learning methods. English Fun Dubbing is an English learning software developed by Hangzhou Fisuke Technology Co., Ltd., which makes English learning fun by dubbing videos of different types and times.
(Tan Meihong, founder and CEO of English Fun Dubbing)
Tan Meihong, founder and CEO of English Fun Dubbing, pointed out that the product design concept of English Fun Dubbing is "interest is the best teacher". She pointed out that now the post-70s, post-80s, and even post-90s and post-00s are experiencing difficulties in learning English, especially in pronunciation and speaking. Tan Meihong refers to the theory of second language acquisition to enhance the learning experience by integrating sound, scene and text. She believes that listening and speaking are the foundation of learning and should be mastered before moving on to more advanced literacy learning. By combining "scene + sound + text", learners will have a deeper memory of the content. Multimedia content such as cartoons, movies, and anime is highly comprehensible in language learning, helping to create a learning environment that is pure English.
Tan Meihong mentioned that the overall positioning of her English dubbing is to practice speaking, and the team has developed an intelligent learning ecosystem for this. This system uses AI intelligence assessment to provide learners with instant voice over scoring and feedback, as well as suggestions for words and sentences that need improvement. Learners end up with a comprehensive learning report with an accuracy and fluency assessment.
Tan Meihong also discussed the possibility of combining GPT technology with English dubbing. She shared three main features of GPT: First, GPT integrates a wealth of knowledge, especially basic and general knowledge, and can serve as a personal learning assistant. Second, GPT has strong learning and reasoning abilities, such as GPT-4's ability to be comparable to Stanford's top students. Finally, GPT has a wide range of application areas and a deep knowledge base, and is easy to use and extend.
In terms of language learning, ChatGPT is very suitable as a learning assistant, which can be embedded in the application scenario of English dubbing. ChatGPT is capable of producing high-quality picture books of various types, which are loved by users. In addition, ChatGPT also has role-playing and conversational capabilities, which English dubbing uses to launch AI foreign teachers. AI foreign teachers can provide foreign language dialogue and gradually introduce thematic courses, including business speaking and test preparation such as TOEFL and IELTS.
Tan Meihong further pointed out that from the perspective of providing customized AI foreign teachers for B2B institutions, corresponding types of teaching content can be tailored according to the specific application scenarios of the institution. She firmly believes that compared with hiring a real foreign teacher, the use of AI foreign teacher has obvious advantages in cost performance. AI foreign teachers have unlimited patience and can continuously carry out various types of training with learners. Moreover, the addition of AI can make the learning process more interesting, but also promote educational equity and give more people access to quality language training.
ChatGPT has ushered in a new era of human-robot collaboration at the level of talent and organizational thinking. Tan Meihong pointed out that although the technical ability of current entrepreneurs is crucial, this factor will become increasingly irrelevant over time. This is because advanced technologies such as ChatGPT will be able to assist people with a variety of technical tasks. As a founder, deep insight into the future, strong belief, and strong adaptability and perseverance will become more central and precious. This so-called willpower and mental strength will be the key to the success of entrepreneurs.
For start-up teams, it is important to have the right tools. Tools can help teams guide where to innovate, accelerate the optimization of ideas and product improvements, and facilitate access to resources.
As for the development of future talents, it is necessary to learn and master various tools, but also to actively think and explore new opportunities to lay the foundation for their career development. Also, at the right time, invest time and energy in grooming yourself to become a prompt engineer.
Talking about the investment in English fun dubbing teaching and research, Tan Meihong said that she will continue to introduce corresponding types of picture books, and will also use ChatGPT to generate and produce picture books, with the blessing of AI, about four to five picture books can be completed a day, which greatly improves production efficiency. The addition of AI foreign teachers will reduce labor costs and reduce user purchase prices, which will effectively promote educational equity.
With the rapid development of the AI industry, it has become crucial to develop and reserve highly skilled talent. The integration of industry and education is deepening, and the cooperation between schools and enterprises has become a key talent training strategy. A typical example is Deep Blue Academy, which was established in June 2017 with the support of the Institute of Automation of the Chinese Academy of Sciences, and is committed to cultivating cutting-edge scientific and technological talents that meet the needs of enterprises. Deep Blue Academy focuses on the three frontier areas of artificial intelligence, robotics and autonomous driving, and has launched more than 40 systematic courses. In the course design process, Deep Blue College first analyzes the company's talent demand and capability models, then extracts key elements from these models, and combines the characteristics of artificial intelligence disciplines to build the final course content.
(Deep Blue Academy co-founder Zhao Song)
Zhao Song, co-founder of Deep Blue Academy, emphasized two core characteristics of AI technology in practical applications: accuracy and fault tolerance of scenes. He said that although the accuracy of AI technology is gradually improving, it has not yet reached extremely high accuracy in many application scenarios. However, this does not prevent AI from being applied in certain scenarios with high fault tolerance for technology. For example, in e-commerce platforms, even if the accuracy of the AI recommendation system is not very high, e-commerce companies are still willing to invest heavily in AI technology, because even a 1% increase in accuracy may bring billions or even tens of billions of revenue growth to the company.
Zhao Song further shared that applications with high scene tolerance rate mainly include product recommendation, creative design and hotel face recognition, while applications with low scene fault tolerance rate include autonomous driving, medical and industrial manufacturing. In the case of medical robots, for example, if the accuracy is not high, then the patient's life may be endangered when performing surgery.
However, these characteristics of AI technology landing also make some problems in the actual deployment process. Zhao Song pointed out that the problems that AI algorithms may face in practical applications include scene adaptation, insufficient device computing power and limited hardware storage space. These limitations are especially noticeable when applied on phones, cars, and other devices. Therefore, when companies hire AI talent, they often look for talents with algorithm design capabilities and expect them to have engineering skills, including coding, debugging, and optimization.
Zhao Song is concerned that there is a gap between the current demand for AI talents and the actual capabilities of candidates. Companies seek talents with strong practical skills and direct problem-solving, and candidates often lack theoretical foundation and engineering ability. This is partly because faculty at colleges tend to have less exposure to AI technology and lack practical experience, and their teaching is more theoretical.
To this end, Zhao Song pointed out three breakthrough paths for AI talent training, the first is to strengthen engineering capabilities, including emphasizing that AI algorithm engineers are also engineers, focusing on cultivating their engineering capabilities; The second is to focus on the combination of theory and practice, so as to "know what is true and know why it is so"; Finally, it is necessary to increase the internship experience and let students practice more to understand the actual needs of enterprises for talents.
In this regard, the practice of Deep Blue Academy is to implement these strategies by adopting two models.
On the one hand, the college offers courses jointly with enterprises, and two-thirds of the lecturers are senior engineers and managers of enterprises with rich theoretical and practical experience. At the same time, Deep Blue College's full-time engineers work closely with enterprises to design corresponding projects according to the company's talent needs model and integrate these projects into the curriculum. This approach develops talents that are more in line with the needs of the company by combining experienced lecturers and well-designed practical projects.
On the other hand, the importance of maintaining frequent interactions with businesses. Deep Blue College frequently communicates with the human resources and technical departments of enterprises to understand their talent needs, and adjust and update courses based on the information collected and the characteristics of AI disciplines. This closed-loop feedback mechanism ensures that Deep Blue Academy's courses can be adapted to the needs of enterprises in a timely manner, and courses are basically updated once a year.