Recently, IEEE ICDM 2018 was hosted in Singapore. Experts, professors and scholars from all over the world in the field of data mining gathered together. As one of the important guests to the conference, Squirrel AI Learning’s chief scientist Wei Cui delivered a speech with the company’s self-developed Squirrel AI intelligent adaptive system. He introduced the practical application and development prospects of big data, AI and other technologies in education to the public, which was unanimously appreciated by the attendees.
As one of the top three international conferences on global data mining, IEEE ICDM has been dedicated to in-depth data mining in statistics, machine learning, pattern recognition, database and data warehouse, data visualization, knowledge-based systems, high-performance computing and other fields. The conference also invited UBTECH’s chief scientist Prof. Dacheng Tao; IBM Almaden fellow C. Mohan; Ramamohanarao (Rao) Kotagiri, dean of the School of Computing and Information Systems, Melbourne School of Engineering, the University of Melbourne; Graham William, Microsoft Asia Pacific R&D Group’s director of cloud computing, AI and data science; Steve Miller, SMU Vice Provost (Research) of Information Systems; and other tech giants. They displayed their latest technologies and achievements in computer vision, blockchain, machine learning and other related fields. In addition, the attendees also discussed a wide range of common concerns about the promotion and application of cutting-edge technologies such as database and machine learning, as well as their future challenges.
World-renowned data mining and AI scholar Prof. Xindong Wu delivered an opening keynote speech titled “Great Wisdom”. He proposed combining human intelligence (HI), artificial intelligence (AI) and organization/business intelligence (O/BI) with big data analysis for industrial intelligence in organizational activities. UBTECH’s chief scientist Prof. Dacheng Tao introduced his team’s achievements in machine vision, including breakthroughs in object detection, scenario analysis, depth recovery from single color images, target tracking and other aspects. IBM Almaden fellow C. Mohan shared a robust, effective and accurate automatic 3D segmentation algorithm for OCT imaging of retinal tissue layer and choroid.
As one of the pioneers of the application of AI and big data in education scenarios in China, Squirrel AI Learning has opened more than 1,600 learning centers in more than 300 cites in China, which have served more than 1 million students. The contract renewal rate is about 80%. Now it has become an important force in revolutionizing traditional education in China. Squirrel AI Learning’s chief scientist Dr. Wei Cui pointed out in his speech that these achievements rely on Squirrel AI’s self-built algorithm core. By collecting and analyzing learning data, Squirrel AI uses a nanoscale knowledge graph to detect knowledge points related to the targets in the least time and create personalized dynamic student portraits, forming self-learning and feedback on the prediction ability of AI and the effect of learning content.
At the technical level, Squirrel AI has integrated the most cutting-edge AI, big data technology, psychology, pedagogy and other relevant theories, forming a set of adaptive learning strategic algorithms running through the whole teaching process.
First, Squirrel AI has rebuilt the knowledge graph, benefiting from the improvement in AI and algorithm technology. In education, knowledge graph and graph theory are usually used to describe and represent each subject’s knowledge system. However, there were always two problems in past knowledge graphs. First, the description of students’ knowledge points was very crude; second, only strong correlations between knowledge points were marked, while weak correlations in the majority were not taken into account.
Squirrel AI adopts nanoscale separation to mark the four key points, i.e. difficulty level, importance, mastery and status with different icons and colors. In addition, the formerly universal 4-6 dimensions of student knowledge portraits in global intelligent adaptative learning have been upgraded to more than 30, so that each student’s mastery of knowledge points can be clearly displayed in one knowledge graph. Taking middle school math as an example, in the adaptive system of Squirrel AI, the original 300 knowledge points have been refined to 30,000 knowledge points. In this process, Squirrel AI uses the theory of association probability of non-correlated knowledge points to build a network structure between knowledge points, so that knowledge points can be deduced from each other. At the operational level, Squirrel AI can adjust test questions with the largest amount of information in real time according to different feedback on each question from each student, so as to test the most knowledge points with the least questions.
Second, Squirrel AI can track students’ learning conditions through data analysis and grasp their learning curves in real time. According to the human forgetting curve in educational psychology, students always accumulate some knowledge loopholes in their daily learning process. However, such knowledge loopholes will not disappear because students advance to the next stage, but will affect students’ knowledge learning in the next stage. Using the Bayesian network and probabilistic graphical model, Squirrel AI can clearly complete students’ knowledge portraits and understand students’ overall mastery of knowledge more comprehensively. In addition, Squirrel AI adopts Bayesian knowledge tracking theory, which can detect students’ past knowledge loopholes.
Third, based on the dynamic tracking of students’ knowledge points, Squirrel AI can help students build personalized learning paths. With the genetic algorithm, neural network and machine learning technology, Squirrel AI can push appropriate learning content to students, get feedback and keep drawing multi-dimensional student portraits. According to the degree and state of students’ knowledge mastery, the system automatically plans the most suitable learning difficulty and order for students, helps them check their omissions and fill in the gaps, so as to ensure that students can use the least time to grasp the knowledge should be grasped.
Read More: Security Tips for Online Marketing
To sum up, Squirrel AI can go deep into the links of teaching, learning, testing and practicing through data collection and analysis, and truly realize accurate control of the whole chain of students’ learning, turning education concepts such as “individualized teaching” and “teaching students in accordance with their aptitude” into reality.
In addition, Squirrel AI Learning’s chief scientist Wei Cui said: “Although Squirrel AI already has a very precise question pushing system, which can continue to improve through the improvement of Squirrel AI’s knowledge detection system, we still hope to improve interactivity. In the future, students’ real-time heart rate, brain wave and facial expression recognition during learning will be added for comprehensive analysis. Each student will be equipped with a virtual personal assistant to provide better learning services for them.”
In fact, Squirrel AI Learning has always attached great importance to technology research and development. Since its establishment, the company has gathered three of the world’s leading experts in intelligent adaptive learning, namely Wei Cui, Richard Tong and Dan Bindman, as chief scientist, chief architect and chief data scientist of Squirrel AI Learning. They respectively came from three world famous AI adaptive education enterprises RealizeIT, Knewton and ALEKS. Integrating their nearly ten years of first-hand experience in the application and R&D of AI adaptive education technology with China’s education and teaching habits, they successfully developed Squirrel AI, China’s first AI adaptive learning engine centered on advanced algorithms with proprietary intellectual property rights. At the AIAED on Nov. 16, Prof. Tom Mitchell, the godfather of global machine learning, dean of CMU School of Computer Science, a member of the American Academy of Arts and Sciences and the National Academy of Engineering, AAAS fellow and AAAI fellow, officially accepted Squirrel AI Learning’s offer of the position of Chief AI Officer. As Squirrel AI Learning’s first person in charge in the field of AI, Mitchell will lead a team of more than 10 AI scientists and hundreds of AI application engineers and technical teams to conduct basic AI research in the field of intelligent adaptative education, as well as the development and application of related products.
Moreover, Squirrel AI Learning has established a joint AI Lab with Stanford Research Institute (SRI) and a joint AI adaptive education lab with Chinese Academy of Sciences (CAS) to enable a customizable, measurable, and teachable personalized education mode. In the past two years, Squirrel AI Learning has made great achievements in AI and big data. Its scientific research papers have been taken by EDM, CSEDU, AIED, AERA and other top international academic conferences. The company has won many international scientific research awards, such as EdTechX innovation award, which has established its leading position in the field of AI+ education.
In the future, Squirrel AI Learning will continue to increase investment in AI, big data, machine learning, education theory and other related fields, and promote the practical application of AI in education scenarios.