On November 23rd, In NIUDAY Beijing station with the theme of “The penetration and integration of AI Industry technology”, Xie Hualiang, technical director of Smart Education Division of Hamirai SEG, shared the “application of AI in education industry” with everyone.
This article is a record of the shared content.
Hello everyone, thank you very much for inviting me to share this with you. What I’m going to bring to you today is the application of AI in the whole field of education.
My topic today is new era, new technology and new education.
We used to have teachers speaking on the blackboard and students listening below, but that’s no longer the case. We are now facing an era in which, due to the popularization of AI and new technology, the old way of education is overturned. I’m going to share with you the whole process.
First of all, a simple popular science, why will this year’s ZTE incident intensified? In fact, the most critical logic is the competition between China and the United States in core technologies. The same is true for AI technology. For example, we established four leading enterprises in 2017, including BAT and IFLYtek, as well as Facebook in the United States and Apple, which is preparing to push driverless driving recently. Behind this is the trend of AI fever sweeping the world.
In AI, there is a global contest called Kaggle Contest, where you are given questions and data to work with algorithms and models. Kaggle has a lot of real-world questions, such as airport security, housing price prediction, lung cancer prediction, airline optimization.
Take airport security, for example. If you go to the security gate at the end of Terminal 2 in Beijing, you will find that it is not artificial but fully intelligent. There will be a camera in each luggage storage area, and each camera corresponds to the person and the luggage. Then the whole security process is semi-automatic, unlike T3 or T1, which has to be manually checked, which is troublesome. If you have the opportunity to go to T2, you can experience it.
Kaggle also did well in predicting lung cancer. Generally speaking, now medical resources are not rich, through a good algorithm, can judge the probability of lung cancer through big data, give it the film, it can identify the probability of lung cancer.
Next is route optimization. Maybe we all fly, but we don’t know that the most troublesome part of a plane is flight scheduling. For example, I came back from Nanchang last week. There were only about eight flights a day from Beijing to Nanchang, but for example, to Shanghai or Shenzhen, there might be more than 50 flights a day. What’s the difference? You have to predict how many people are going to take the flight based on the daily flow of people. If manual scheduling is very painful, the country has to row more than 2,000 routes every day, must be completed by machine operation.
There’s an app in Spain called Pay Per Laugh, or “pay-per-smile theatre.” Each person has a screen in front of them that charges you for every laugh you make. For example, a spectator who laughs 38 times will pay €11.40. This is a very typical application of face recognition, and some theaters in Europe are doing this. After paying for the smile, the number of viewers increased by 35%, the unit price increased by 6 euros, and the total revenue per game was 28,000 euros. This is a use of life scenarios by AI.
Now let’s briefly talk about the development of AI industry, such as data, algorithms and calculation examples. In recent years, we can see that AI is a big trend in search, e-commerce, social networking, virtual assistant companies, autonomous driving, education and other fields.
All AI landing has four elements, the first is the landing scene, is the steering wheel. For example, today we seven niuyun, can reduce labor costs, otherwise there are countless excellent algorithms are worthless. The second is data, which is similar to our gasoline. The core of a car cannot run without data. At least the direction of AI is still based on the logic of big data. The third is algorithms, like car engines. Good algorithms can make a huge difference in efficiency. And then finally, we have an example, which is the wheel. Whether it is neural network, ordinary machine learning, or deep learning, the core is to rely on GPU, now all are TPU, the earliest CPU can not keep up with this era, now all need TPU, with a large number of deformation calculation operation unit processing.
Education is a traditional direction for us, but what if AI plus education?
Let me begin with a brief overview of the development of education. The earliest education was elite education, such as the imperial examination system in China and the elite education of the western aristocracy. Now why is the socialist system better than the capitalist system? The core is that the socialist system provides collectivized mass education, while the core logic of the capitalist system is still personalized elite education. European and American TV dramas will shock your vision with countless entertainment content to persuade people not to study. In fact, they are doing social stratification, which is the “nipple music theory”. As a result, the good people, the elite group, continue to dominate the Western countries, while the ordinary people you don’t want to learn and go down.
But to Our country, the socialist system is a collective mass education, and everyone has to receive nine years of compulsory education or twelve years of education. Nine-year compulsory education was made compulsory by the state to give more ordinary people opportunities. Education and medical care are the bottom line of a country. After a few years, the same group of people will still be standing at the top of the pyramid in Europe and the United States, while in China, every ordinary person will be given the opportunity to change their destiny through learning. This is the embodiment of the superiority of socialism.
But that is the past, and the needs of the future have changed. In April this year, the country put forward education informatization 2.0, which requires that in addition to the informatization upgrade, more in the transformation of technical literacy, to build the “Internet + education” platform, to personalized education, large-scale teaching in accordance with their aptitude. This is reflected in the new college entrance examination reform policy, which has been changed nationwide since 2018. Now the new college entrance examination is 3+3, in addition to the number of languages outside the three courses, the remaining three courses by your choice, such as the future want to be a reporter to learn liberal arts, is geography, history, politics, if the future want to be a programmer that you may learn is physics, chemistry, biology. After the new college entrance examination reform in 2018, the future examination will be more suitable for everyone than before, each person is unique, this is the whole Of China’s current education reform. Under the big goal and direction, it means that the core of individuation must be in the future, and the bottom of individuation must rely on artificial intelligence to do the support. The core of the new college entrance examination is the change of the whole national strategy. We hope that in the next 15 or 25 years, our country’s talent can compete with the United States head-to-head. We are no longer the same subject for everyone, but more innovative and cultivate more excellent talents.
In the direction of daily education, routine has at least four links: teaching, learning, examination and management. The first three parts are carried out jointly by teachers and students, and management is generally the relationship between the school and teachers and students. Let’s take a look at what applications are being made to support these situations.
01 AI speech recognition
Speech recognition is common, but what does it combine with educational learning? For example, if I say a word “desk” and repeat it in English according to the standard pronunciation, the software will give me a score. We know that it is very hard to recite words in daily life. The most powerful person may recite one thousand words a day, while the average person may recite one hundred. If I do this pronunciation training, IT will be very painful for me to let the teacher teach me. It is impossible for the teacher to listen to you recite a word ten times every day. And with a machine that takes the trouble to recognize speech, I can use the machine to score my pronunciation, to judge my pronunciation. If I were A child, I could carry out stage division from A to Z, and I could keep training with my accent, avoiding chinglish. This is a scene of speech recognition, part of self-study.
02 Magic Mirror system
Let me explain the magic Mirror system in one sentence: it is a visual system that collects students’ real-time status and presents it. Finally, the data is summarized to reverse teachers’ teaching and promote students’ self-discipline.
We launched a scene in the “Dual Teacher class” called “Star of Concentration”. First, explain the “double teacher classroom” : “double teacher” is a new mode of education. For example, in the High School affiliated to Renmin University of China in Haidian District, there are excellent teachers, but some areas with relatively poor education resources want to absorb high-quality education resources. In this case, teachers in Beijing could be seen live on camera, in life-size, in other areas. There are offline teachers on site, only responsible for receiving homework on site, correcting everyone’s classroom discipline and so on. The combination of these two teachers is called “double teacher”. But there was a drawback, because the teacher wasn’t there, which meant that the students would get bored and get distracted, so we used the star of attention to solve this problem. There is a camera in the classroom that can take pictures of everyone and give them points based on their posture. This is a very small improvement, and it’s an application scenario for AI.
Another application scenario is the small class of our offline training institution Xueersi, with about 15 students in each class. Education you may have anxiety, especially do the parents, in the face of such problems, learning and thought of young children will have a “box”, it can identify the children in the classroom, such as whether or not he’s got his hand up, expression, pencil or chat with children beside, all can record is very clear. And can generate statistical reports. In this way, parents and children can be urged to communicate in time for problems, and teachers can also be urged according to the classroom effect. This, like Focus Star, can make the class fun.
The above two products are very important for the three points I mentioned before: the state of students, the state of teachers, and the independent improvement of students’ state in class. These artificial intelligence products, we have a technical term for them called “academic assessments,” which look at the whole student. In addition, the most core is the learning process, the learning process if not artificial intelligence, is meaningless.
First, self-study, then academic assessment, all monitoring of daily behavior. There are very specific ai implementation methods involved in this environment.
03 Knowledge Graph
When we went to school in the past, in fact, if you look carefully at the outline of the Ministry of Education, you will find that no matter what version of the textbook, the corresponding knowledge is finally sorted out. For instance, we say the high school mathematics can be dissected into big mesh structure, each dot form what knowledge contents, which can correspond to the knowledge, knowledge what are the questions below, this is equivalent to a very large library, this is after our final knowledge map, each content of knowledge map with numerous extension.
So what does it do? For example, we can set questions automatically. If we find that a certain knowledge point is not well mastered by a student according to daily learning habits, we can completely rely on the knowledge map to recommend corresponding frequently wrong topics to them. Another scenario is that according to artificial intelligence, according to the difficulty of knowledge automatically group papers, define the difficulty coefficient, you can very quickly generate the corresponding papers, can greatly reduce the burden of teachers, increase efficiency. It accords with the educational policy of “increasing efficiency and reducing burden” advocated by the whole country.
Using science and technology to promote the progress of education is our good future mission. We are not only an education company, but also a technology company. We look forward to using science and technology to support China’s education informatization to a higher level in the future.
That’s all for my general share today, thank you!