On Nov. 3, 2018, ODSC West ended in San Francisco. As the world’s largest AI application and data science conference, it invited 4,500+ professionals from fields like machine learning, deep learning, AI research, data visualization, data scientific management, etc. In addition, core contributors to open source tools, libraries and languages also expressed their views on AI and big data. The exchanges of ideas between professionals from various fields will undoubtedly become an accelerator for the development of the whole industry. This year, China’s AI and big data leader Squirrel AI Learning by Yixue Group founder Derek Haoyang Li was a keynote speaker at ODSC West.
Guests invited to this conference included Daphne Koller, a currently leading AI scholar, professor at the Stanford Computer Science Department and founder of the online education platform Coursera; Andrej Karpathy, a former Tesla OpenAI fellow, one of the top experts in computer vision and deep learning, who developed the CNN/RNN with Feifei Liand now is Tesla AI and autopilot vision director; Reza Zadeh, Matroid founder and CEO, who focuses on machine learning, distributed computing and discrete applied mathematics and began to work in Google AI team from the age of 18; etc.
At the conference, SUNY associate professor of political science and famous writer Virginia Eubanks expounded her views on social equality in the era of AI and presented her systematic research on the influence of AI on poor and working-class people in America through data mining, policy algorithms and predictive risk models. Hernan Asorey, chief data scientist of Salesforce, described how to build a net-score-driven application and how to use the program to automatically identify and successfully serve customers through real cases of Salesforce.
At the conference, Derek Li delivered a keynote speech titled: “How Squirrel AI Outperforms Human Teachers and Redefines Education”. It not only was well received by the audience, but also attracted attention from Orange County’sformer Education CFO Darren Dang and ODSC founder Sheamus McGovern.
A summary of Derek Li’s keynote speech is presented below:
From Oct. 2017, Squirrel AI Learning began to verify the feasibility of machine teaching outperforming human teaching. Four human vs machine teaching competitions were organized and the results were surprisingly consistent: machine teaching outperformed human teaching. In the fourth human vs machine teaching competition, which took place in 100 cities, students with the same knowledge level were divided into two groups and respectively received teaching by human teachers and Squirrel AI adaptive system. After 8 hours of teaching, students in the machine-taught group learnt 42 knowledge points on average, while students in the human-taught group learnt 28 knowledge points on average. In terms of score improvement, the machine-taught group’s score increased 5.4 points on average, while the human-taught group’s score increased only 0.7 point. The impact of the result was no less than the sensation caused by Alpha Go’s successive defeats of top players Sedol Lee and Jie Ke.
Squirrel AI Learning founder Derek Li explained that the reason why AI can revolutionize traditional education, which has not changed for hundreds of years, is its innovation of traditional education in four dimensions: versatile and know-it-all, tell big stories from small things, infinite computing power and self-evolution.
First, compared with humans, AI is more versatile and know-it-all. Considering the cost, China’s education model is usually one-to-many. Such general cramming is uniform. It seems that every student is taught the same knowledge, but their knowledge levels are quite different because every student varies in their basic ability, concentration, weakness and range of understanding.
Squirrel AI adaptive learning system is versatile and know-it-all. It can truly teach students according to their aptitude. The teaching robot detects each student’s user profile and scans their knowledge points, so the system can accurately know students’ knowledge mastery in a short time, and then implement different teaching programs for different students.
Next, it can tell big stories from small things. On the basis of determining students’ mastery of knowledge points, it matches and consolidates knowledge points targeting each student’s weak points. Using machine teaching, Squirrel AI can truly reconstruct the knowledge map based on the causes of mistakes. On the basis of accurately finding the causes of making mistakes, the AI adaptive system develops personalized learning programs for students, and finally, helps solve the student’s learning difficulties.
It’s worth noting that in the process, Squirrel AI realized super nanoscale knowledge splitting and put an end to the rough description of knowledge points in the traditional knowledge graph. At present, most education companies can only achieve three-level splitting of knowledge points, but Squirrel AI can realize nine-level splitting. For example, for theaddition or subtraction of fractions, Squirrel AI can split it into 100 knowledge points. It can effectively detect students’ weaknesses and help enhance their divergent thinking.
Third, it has infinite computing power. Nanoscale knowledge point splitting is naturally inseparable from strong computing power. Based on big data and with the help of powerful computing power, Squirrel AI can quickly locate the knowledge that students have mastered and have not mastered. It can put forward effective solutions in a short time and make up for students’ knowledge gaps.
This short and fast approach solves the common root problem in current education. The effect of AI adaptive education is influencing more and more students and parents. Perhaps that’s why Squirrel AI Learning’s paying rate reached 80% in six months. Obviously, the value created by AI education is being recognized by users.
Last, it’s self-evolving. After all, teachers’ energy is limited. It’s impossible to realize one-to-one tutoring. It is also difficult to find knowledge gaps and understanding deviations from students’ exercises and classroom performance. Even the students themselves cannot tell which knowledge points they haven’t understood completely. However, AI teachers can not only find the shortcomings of students in class, but also continuously help them to check and fill in gaps and make more optimized learning plans in the process of self-evolution.
The four traits of AI not only make Squirrel AI become the best in the education sector, but also lead the whole sector to redefine education. When asked whether AI will replace human teachers, Squirrel AI Learning founder Derek Li said human teachers cannot be replaced because of the essence of education.
“Teachers are needed for planning teaching and determining the direction. But we don’t need famous teachers. The cost of training famous teachers is too high and they cannot be personalized for every student. In classroom teaching, famous teachers cannot put much energy in every student. The algorithm matching in our classroom can customize the experience of 100 teachers in our system, so every student actually learns from the essence of 100 teachers.”
The teaching system of Squirrel AI seems to be uniform, but actually is personalized. It can not only realize rational and average distribution of teaching resources, but also truly solves the problem of treating symptoms rather than root causes found in traditional education. It traces the root of every student’s difficulties and blind spots in learning, so as to form a complete knowledge chain.
At present, AI is in a period of rapid development in dense data collection. In the future, its data integrating ability will become stronger and stronger. In terms of knowledge transmission, through continuous data collection and adaptive optimization, AI will be able to deeply understand the student’s needs and requirements. The role played by teachers will no longer be a “carrier”, repeating knowledge every day. Focusing on the emotional temperature of students, teachers will be leaders who awaken, encourage and inspire students, truly fulfilling their educating function.
AI will not really replace human teachers, but teachers who do not use AI will be replaced by teachers who use AI. In the future, teachers and AI will achieve synergistic development. Only in this way can we maximize the utilization of education resources and improve the productivity of the whole education sector. Teachers will not be replaced, but education will be redefined.
The arrival of AI is both an opportunity and challenge for the education sector. In the future, only enterprises that can use AI to separate and coordinate teaching and education can truly lead the development of education. Squirrel AI Learning will continue to work on this. Haoyang Li also believes that in the next 7-15 years, Squirrel AI can truly integrate the wisdom of Socrates, Da Vinci and Einstein, so that every student can receive perfect education.
Read More: Interview with Megan Kvamme, CEO at FactGem