SHANGHAI, June 5, 2019 /PRNewswire/ -- With the rapid development and wide application of AI in all walks of life, online education has surfed the tide of AI applications. At present, an AI -based adaptive learning system is being rapidly developed and delivered to both China and internationally. What is adaptive learning? What are the application mechanisms of adaptive learning? How will AI reform and evolve the education industry? Squirrel AI Learning by Yixue Group presented detailed answers to the above questions at the World Summit AI Americas, which had around 1,000 participants.
The World Summit AI, an international AI research and industry application conference, was held in North America this year. World Summit AI Global Series is known for the quality of speakers. Last year, more than 140 experts from Apple, Facebook, Amazon, Alibaba, Tencent, Google, Intel, NASA, Uber and IBM Watson delivered speeches at the WSAI in Amsterdam. From April 9th to 10th of this year, the Summit was held in Montreal, and speakers included Yoshua Bengio, winner of the recent Turing Award; Gary Marcus, Famous Author and professor at New York University (NYU); Hassan Sawaf, Director of AI at Amazon AWS, and long list of well-known professionals from the AI industry.
Joleen Liang, partner at Squirrel AI Learning, also gave a speech on "Squirrel AI Learning system", an advanced algorithm-based AI adaptive learning engine. According to her presentation, the learning engine has broken down knowledge points of a subject into more than 30,000 sub-points, compared to only 500 points in general textbooks. In addition, the system will adjust itself with the change in the ability of students in real time, the corresponding knowledge points will become increasingly practical and useful. Besides, Squirrel AI Learning can also identify specific learning characteristics of different students to provide them with personalised learning path and materials, thus allowing students to master knowledge points and overcome difficulties in an efficient and effective way.
Beginning with the four major difficulties faced by traditional education, Joleen Liang made a presentation: How AI Adaptive Learning Can Provide Solutions to the Difficulties Respectively. Firstly, educational resources are unevenly distributed, and there is a relative scarcity of top teachers; however, AI can break the time and space limits and enable the integrated application of top teachers' rich teaching experience, even if there are a small number of such teachers. Secondly, students differ from one another in their own learning situation, whereas the traditional "one-size-fits-all" education has difficulties teaching students in accordance with their aptitude, however, AI can provide personalized teaching for students according to their learning needs. Thirdly, the same teaching schedule is applied to all students in traditional education, whereas students differ from one another in internalizing knowledge in terms of speed, AI can help students overcome their weaknesses and provide unique teaching schedules. Finally, the traditional cramming method of teaching highlights knowledge acquisition only. It ignores the training of learning capacity, independent thinking and learning habits; AI teaches a man fishing then to give him a fish.
She added, "Squirrel AI Learning has set up a nanoscale knowledge map, by which a knowledge point can be broken down at nine levels, and an MCM system can be established, which can quickly detect learners' learning capacity and quality. With these advantages, it can accurately draw a portrait for learners to predict their learning path and time, so as to finally recommend personalized learning contents to them. In addition, Squirrel AI Learning conducts a multimodal behavior analysis (MIBA) to monitor students' facial expression and behavior, evaluate their attentiveness, and detect and analyze their study motion to accurately understand their learning status and help teachers identify real difficult problems."
The MCM system, which can quickly detect learners' learning capacity and quality
MCM system of the Squirrel AI Learning can set the factors of learning thinking apart to detect the learners' model of thinking, learning capacity and methodology. Joleen took physics as an example and pointed out that in physics study, the model of thinking could be split into symmetrical thinking, creative thinking, classified discussion thinking, analogical thinking, etc. After evaluating and testing students, the MCM system can identify differences in learning capacity, knowledge point comprehending capacity and related weaknesses among students that score equal points, so as to provide them with a more specific learning program.
The Squirrel AI Learning system assesses the students' knowledge level using the knowledge map technology and knowledge space theory. It helps to gain a clear idea of students' real-time learning process using Bayesian networks, Bayesian inferences, Bayesian knowledge tracking and item response theory (IRT) so as to evaluate each student's mastery of knowledge and predict their future learning skills. Based on its adaptive learning engine as well as algorithms, content, data and teaching design, Squirrel AI Learning dynamically recommends the most suitable learning content to students.
Last November, Squirrel AI Learning engaged Professor Tom Mitchell, a top AI scientist, global luminary on machine learning and Dean of School of Computer Science, Carnegie Mellon University (CMU), as the company's Chief AI Officer. He concluded that Squirrel AI Learning aims to build a super-AI teacher that is as knowledgeable as Socrates, Leonardo da Vinci and Einstein together. With a wide range of knowledge, excellent curiosity, divergent thinking mode and logical inductive ability, the AI teacher can teach a wide array of students in accordance with their aptitude.
World Summit AI Americas
Highlights of World Summit AI Americas
As a large-scale AI and industry application conference, the World Summit AI is famous for being addressed by AI industry elites.
Yoshua Bengio delivered a speech titled "Deep Learning beyond Supervision". According to his presentation, the thing that machine learning is lacking at present is the understanding and generalization outside the training distribution. Currently, the learning theory only involves the generalization in the identical distribution; the model can learn, but does not summarize well (or has high sample complexity when adapting), and modify the adjusted distribution, etc. In addition, the reuse effect is not good, the poor degree of the knowledge module is also the "short board" of the current machine. Yoshua believes that our machine learning model remains susceptible to examples of distribution and antagonism if it still runs based on superficial statistics rules.
Daphne Koller, one of the founders of Coursea, the largest platform of massive open online courses (MOOC), discussed machine learning as a new method for drug discovery in her speech. As a research scholar, her treatises have been published in more than 200 publications such as Science, Cell and Nature Genetics. Daphne has a new role as the CEO and founder of Insitro, an enterprise that specializes in drug development with machine learning. She mentioned that the technological progress of machine learning can help to build a comprehensive high-throughput platform for biological data to make it possible that extensive research might be carried out based on the cell system. Moreover, with time going by, the system content will be enriched to re-summarize correlations among human diseases. Machine learning plays a role in allowing researchers to interpret massive data generated by it, and takes intervention measures to improve human health.
Gary Marcus is a successful scientist, best-selling author, entrepreneur, and CEO and founder of Geometric Intelligence (a machine learning startup that was later taken over by Uber). As a writer, he often writes for The New Yorker and New York Times. Besides, he is the author of four books. As a professor of psychology and neuroscience at the New York University, he has published a lot of papers on human and animal behaviors, neuroscience, and genetics and AI in Science and Nature, etc. In his speech, he revealed the shortcomings in the development and application of AI again and again. He believes that to create trustworthy AI, the following four points must be considered: Firstly, the cognition of mankind is very complicated. So, when researching AI, it is not suitable to search for the so-called "tiger balm". Secondly, there is no substitute for common knowledge. Thirdly, the Gary Marcus human learning process is not merely about big data or digital processing, but also involved in more complicated processes, such as touch and exploration explained with the example of his daughter climbing a chair. Finally, common knowledge may be derived from an inherent understanding of time, space and objects.
World Summit AI Americas aims to explore the important influence of AI on business, scientific research and technological development. The two-day summit has been related to AI4Good, application solutions for enterprises, discussion on successful practice, and formulation of a plan for promoting the application of AI. Next year in March, the World Summit AI Americas will be held again in Montreal, the city of research and implementation of AI.
View original content to download multimedia:https://www.prnewswire.com/news-releases/squirrel-ai-learning-by-yixue-group-at-world-summit-ai-americas-wsai-how-does-ai-upgrade-the-education-industry-300862396.html
Copyright ⓒ Acrofan All Right Reserved