A craze for AI large models triggered by ChatGPT continues today.
With the rapid development of artificial intelligence technology, large models, as its core branch, have triggered a craze for AI large models around the world. Since 2023, technological innovations, such as large language models and generative AI, have injected great potential and influence into all walks of life. Especially in the field of education, the deep integration of artificial intelligence and education is promoting education reform and innovation, leading high-quality development.
On January 5, at the AIAED Intelligent Adaptive Education Large Model Academic Seminar held in Kunming, Squirrel AI became the focus of the audience with its world's first all-disciplinary intelligent adaptive education large model.
Education has entered the era of large models.
In the past year, large-scale model technology has developed rapidly, providing infinite possibilities for the expansion of AI applications. The world's major technology giants have laid out the field of large models, and education companies are also actively participating in this wave. For example, Khanmigo, an artificial intelligence teaching assistant launched by Khan Academy, and Duolingo Max, Duolingo's GPT-4 technology product, have added features such as speech recognition and real-time feedback, significantly improving the learning experience.
At the national policy level, the development of artificial intelligence has been supported by industrial policies. In July 2023, seven ministries and commissions, including the Cyberspace Administration of China, the National Development and Reform Commission, and the Ministry of Education, jointly issued the Interim Measures for the Management of Generative AI Services, which provides a strong guarantee for the development of large model technology in the education industry.
* Education companies have increased their competitiveness in the field of large models and intelligent hardware. On November 7, 2023, Good Future announced that its Xueersi large model MATHGPT has completed the relevant filing and successfully passed the review, becoming the first batch of education models that have passed the filing. Xueersi will gradually realize the application of AI capabilities based on self-developed large models on the learning machine, including functions such as "AI Dialogue Learning", "AI Topic Assistant", and "Chinese and English Writing Assistant". In December 2023, Xueersi launched the second-generation learning machine, which incorporates MathGPT technology.
In addition, in September 2023, Job Gang released the self-developed Galaxy model;In October 2023, iFLYTEK released the iFLYTEK Xinghuo cognitive model V30。These large models cover multiple disciplines, multiple school stages, and multiple scenarios, closely integrating AI technology with education to help education innovation and high-quality development.
At this conference, Squirrel AI also brought its latest technological achievements - the world's first all-discipline intelligent adaptive education model, which is also the first vertical field large model in China that combines intelligent adaptation technology with multimodal large models and applies them to the field of education.
It is reported that adaptive learning is the frontier of the development of educational technology around the world, and the learning platform designs learning paths and pushes personalized learning resources for learners by collecting and analyzing learning data. The intelligent adaptive learning technology of Squirrel AI is an intelligent adaptive learning technology, which was combined with education and training through a SaaS system containing a knowledge graph in the early days, and is now applied to hardware such as learning machines.
At the same time as the release of the all-discipline intelligent adaptation model, Squirrel AI also released a new learning machine. Whether it is the technical route of the intelligent adaptation model or the business model of the learning machine, the products of Squirrel AI are quite distinctive, and they also show a certain direction for the development of the educational model.
Why is the education industry rushing to the big model?
Hardware is not new in the world of education. Many years ago, China launched educational hardware such as learning machines, point readers, and point reading pens, and integrated basic AI technologies such as text recognition and speech recognition, providing students with functions such as point reading, translation, and human-computer dialogue. However, these traditional products are essentially more like "think tank" devices that integrate various software functions and learning resources, like electronic dictionaries, providing a wide range of knowledge and resources.
With the rapid development of AI technology, large models, as its core branch, are leading the transformation of the education industry. Although traditional education hardware already has basic AI technologies such as text recognition and speech recognition, it lacks real intelligent functions. The introduction of large models has made the educational hardware have more powerful natural language interaction capabilities, and has become a knowledgeable and capable "cyber teacher" who can teach students according to their aptitude.
It is no exaggeration to say that the large model gives the "soul" to the educational intelligent hardware, making it from a cold learning machine to a warm and intelligent companion to the learning object. This change is disruptive for the entire industry, and will also become a milestone on the road of human exploration of AI + education.
At the meeting, Squirrel AI released the world's first all-discipline intelligent adaptation education model. Li Haoyang, founder of Squirrel AI, said that the quality and quantity of training data for AI large models are crucial. Squirrel AI's intelligent adaptive education model has three outstanding advantages: the use of multi-modal large models for application and development, rich technology and industry experience in scenarios, data, algorithms, etc., and based on unique knowledge graphs and MCM graphs, it truly improves teaching efficiency and gives full play to each child's personality, strengths and potential advantages.
With the support of large model technology, Squirrel AI has comprehensive and efficient diagnosis and mapping technology, which can plan the best learning path for students and deeply empower students to teach students according to their aptitude. In addition, through emotional intervention and analysis of children's quality and ability, Squirrel AI can also generate the most effective educational interaction, give positive feedback and motivation to children, inspire learning interest and cultivate learning habits.
Accelerate the implementation of adaptive learning applications.
Traditional adaptive learning generally has two main modules, a mapping test and a preset teaching logic. That is, through the mapping test, the adaptive learning software understands the learning level of students, diagnoses the strengths and weaknesses of knowledge mastery, and then recommends appropriate learning content to student users by teaching logic according to the results of the mapping test. Li Haoyang introduced that unlike the tool-based products based on artificial intelligence technology such as photo search, oral assessment, and essay correction on the current market, the intelligent adaptive learning system can not only provide personalized learning programs and support according to students' learning conditions and needs, but also focus on cultivating students' thinking ability, innovation ability, emotional attitude, etc.
Li Haoyang believes that there are several weaknesses in the current large-scale model based on the general model in the education industry: first, the lack of in-depth understanding of student users, due to the lack of historical data and portrait information, the general model product can not clearly distinguish which age of the user is the user of the product, which limits the accurate grasp of the user's needs of the product;Second, it is difficult to provide personalized learning programs, unable to comprehensively diagnose students' learning conditions, and it is difficult for general model products to provide targeted learning suggestions and programs, so as to prescribe the right medicineThirdly, the general model is not dominated by the education system, which leads to the lack of key information and scenarios required for education, such as subject diagnosis, learning history level, knowledge weaknesses, etc., which makes the general model unable to fully meet the individual needs of the education fieldFourth, due to the characteristics of the general model, it cannot meet all needs, and it is difficult for parents and educators to fully authorize the product to obtain more in-depth student learning, which limits the wide application of the general model product in the field of education.
Compared with the general base of the public data model, Li Haoyang pointed out that the difference of squirrel AI is that it proves the generalization ability of the large model in the vertical field of education, and also realizes the irreplaceable value of the general model. In terms of landing application, Squirrel AI can cover multiple scenarios such as preview, review, test preparation, and homework assistance when providing tutoring to students, giving students more accurate responses and incentives in terms of social, emotional and interpersonal interactions.
For example, Squirrel AI can improve students' emotions by giving them positive feedback and incentives according to their reactions in the learning process, helping to solve some of the children's psychological problems, effectively solving the problems of children's slow academic improvement, difficult improvement, and the formation of good learning habits, helping each child become the best version of themselves, and giving full play to everyone's personality, strengths, and potential advantages. Li Haoyang said.
According to Li Haoyang, at present, Squirrel AI has begun to try to apply large model technology to create virtual teachers. By generating more ** explanations, it provides reference and reference for the landing of large models in more industries.
Li Haoyang said that in the future, the working group will attract more expert members and entrepreneurs, give full play to the advantages of professionals in various fields, and provide professional guidance and suggestions for the development of China's education model technology and the application of the industry. Based on the technological innovation achievements and experience accumulation in the field of educational intelligent adaptation large models, Squirrel AI will also inject a steady stream of power into the construction of a standardized system in the field of Chinese artificial intelligent education and the intelligent upgrading of the global education industry.
*: Global Network.