Dai Juan: Hello everyone, I'm Dai Juan from SenseTime. I'm going to share with you today about artificial intelligence.
I don't know if you know about SenseTime, take a little time to briefly introduce SenseTime, an artificial intelligence company.
SenseTime is the world's leading AI technology platform company, and the entire company has many practical cases in the new infrastructure industry of AI and many fields of AI.
Many listeners may have seen, for example, some areas such as the relatively hot driverless driving that is often seen on the road now. Through some technologies of artificial intelligence vision, SenseTime has more landing cooperation with many well-known car manufacturers, such as GAC and Honda, in the field of unmanned driving.
At the same time, artificial intelligence such a technology, but also empower our lives and many industries in all aspects, such as smart medical care, through artificial intelligence image technology, can help doctors better do pathological diagnosis, but also can better do such as three-dimensional reconstruction of bones, with a cheaper and more inclusive way, so that more people can enjoy higher quality medical services.
In addition, artificial intelligence has many applications in the field of remote sensing, such as meteorology, satellites, weather forecasting, etc. How to make weather forecasting more accurate, how to use remote sensing to make images better to help urban planners, better plan the big map of our entire city, and which plots are used for what kind of scheme in urban planning, such directions are actually the empowerment of artificial intelligence in the industry.
In life, you may often use artificial intelligence, including the current voice assistants, like Xiaoai, like Siri, in fact, behind these applications, a lot of speech recognition, natural language processing, such as some basic artificial intelligence technology empowerment.
When we order takeaway, students may not notice that in fact, a large number of intelligent algorithms are involved behind many takeaway platforms. What kind of takeaway delivery people we arrange, how much time we use, and what kind of routes we use. Similarly, intelligent algorithms of artificial intelligence will guide applications such as the taxis we use now. So we see that artificial intelligence is a new technology, which gradually penetrates into all aspects of our lives and empowers many industries.
After spending nearly ten years in the United States, I returned to China in 2018 to join SenseTime as an AI technology platform company like SenseTime, engaged in AI education for teenagers. So during this period, there are actually many friends, including teachers who are asking me: "Dai Juan, you actually did not do work related to the education industry before 2018, why do you want to do artificial intelligence education for teenagers after returning to China?" ”
Here I also share with you, in fact, what touched me very much is that the future era must be the era of technology. My ten years in the United States, including some of my previous work experiences, including the development of many large companies at home and abroad, and the smart technology we just mentioned, are quickly penetrating all aspects of our lives and work. We want to say that the children of the future are the children of ten or twenty years from now, and what kind of basic qualities they need when they enter this world and enter this society. So after thinking, we think that for the children of the future world, whether he enters the investment industry, the advertising industry, the medical industry, the intelligent driving industry, or some retail industries, or even the education industry, and we do history, do the industry of reading books like Fan Deng, in fact, he may need to use the intelligent thinking we are talking about now.
What kind of technology is artificial intelligence? As the name suggests, it is artificial intelligence. What is intelligence? Usually we say that intelligence is a cognitive ability that people have uniquely, which includes human perception, like a child can perceive the world and understand the world by crawling. By learning, we can reason and finally make decisions about a problem or an event. This chain we call human cognitive abilities. What is artificial intelligence? Artificial intelligence is actually the ability to simulate some basic human cognition through machines.
Very early artificial intelligence, in fact, we saw, so that machines can move automatically, in the era of machine automation, is actually part of artificial intelligence. Today's artificial intelligence has actually entered a relatively new era. When we talk about artificial intelligence, we let the machine have human cognitive ability, how to let it have cognitive ability? In fact, when I began to learn artificial intelligence in the early 2000s, it was more through mathematical methods, deduction methods and statistical methods to do the cognitive framework of artificial intelligence, and we actually summarized some human experience into some principled things through statistical methods, and we taught machines.
Since 2010, there has been a growing force called machine learning. Machine learning is actually a branch of artificial intelligence, in the past, the intelligence of machines was defined by people, and it was people who summarized the laws of the world, and then informed the machine of this law, so that the machine could reproduce this law. What is machine learning? Machine learning is what we teach machines not specific knowledge, but the ability to learn. So in this world, there is a group of particularly powerful scientists who have taught machines how to learn through scientific methods and engineering methods in many fields. Through machine learning, machines actually have more flexibility and control.
In fact, there are many ways to learn machine learning, machines can learn from big data, machines can learn from behavior, and machines can also have reinforcement learning methods. In fact, these are analogous methods of human learning, so after 2000, a group of very powerful scientists defined the method of machine learning with artificial neural networks. So we think about why human heredity does not inherit knowledge. For example, parents understand so much knowledge in this field, give birth to a child, why the child does not inherit the knowledge information in the brain of the father and mother. Maybe not, but human inheritance is actually a very powerful ability, called the ability to learn, that is, the ability to recognize and construct.
So after 2010, there was a group of scientists, who had been in the field of deep neural networks since the 70s and 80s, building a network for machines to simulate the neural networks of the brain. It has neurons, synapses, neural connections and the development and construction of networks. In this way, the machine has a stronger learning ability. This is why we are talking about why such technology can bring about earth-shaking changes in all walks of life and our lives that we see now.
The series I recommend to you today is called "Fundamentals of Artificial Intelligence". This book is very interesting and is called the world's first artificial intelligence book for teenagers. I think that artificial intelligence education, such as the United States, has been relatively good practice and landing. So while we were working on this book, I was in touch with a lot of colleagues in the United States. Because my children were very young at the time, so I didn't go to elementary school or kindergarten in the United States, I called these colleagues or friends I knew and asked what kind of artificial intelligence education your children received in elementary school, junior high school, and high school in the United States. They were stunned on one end of the phone, they said that without artificial intelligence education, why would children learn artificial intelligence in elementary school, junior high school, and high school. So at that time, I realized that we were doing a pioneering work, we were compiling the first artificial intelligence textbook or artificial intelligence reading book, in fact, there was no systematic, youth-oriented artificial intelligence knowledge, skills, tools, cases, projects, such content that we could use as a reference, so this was a pioneering work.
AI takes it as a technology, as a tool, to teach students, to teachers. In fact, in the future, I think artificial intelligence education, or artificial intelligence literacy, has a deeper meaning, what is the reason?
We talk about what knowledge is, knowledge can be obtained by memory, but knowledge obtained by memory is not necessarily completely cognitive. Why is it interesting to talk about artificial intelligence education in this direction, we can see how machines learn, what is the way machines learn, what is the way machines build cognition, because we are in the process of teaching, at least my own learning years, including the way I observe learning now, in fact, I have never or rarely taught children what cognition is, what kind of ability you are building, and what is the way this ability is built. What are the results brought about by its way.
Artificial intelligence education, in fact, to some extent to provide children with an opportunity to observe the process of intelligence construction in vitro, children through what we just said, a lot of teaching machines how to see, teaching machines how to understand, teaching machines how to read poetry, teaching machines how to read this poem, teaching machines how to write texts, and how to write a good composition, in fact, in this process, is the child's ability to build cognition in these fields, has also been greatly improved. So that's a direction that we're very interested in. Speaking of the teaching of artificial intelligence, it is actually in vitro to build cognition such an ability teaching, if children participate in such a capacity construction process, how much help will their cognitive improvement. And in the future world, if there are so many agents in vitro, how can our teenagers or these friends in the world in the next ten or twenty years better cooperate with the agents in vitro to coexist, to symbiosis, and to innovate, we feel that this is a very important ability.
In the process of promoting artificial intelligence education, especially youth artificial intelligence education, everyone will also ask why teenagers? Why not start with college students?
Artificial intelligence actually has different meanings for readers of different school ages, or for teenagers of different school ages. In fact, just like we teach mathematics, if there are kindergarten children at home who can observe, in fact, kindergarten children learn mathematics more called physical calculation, right. When he teaches 1+2=3, he must think that one apple plus two apples equals three apples, which is a very in-depth study in the psychology of cognitive development, because human beings have to go through this stage of entity operation to gradually evolve to this stage of formal operation. In fact, the same is true for artificial intelligence. For children, we may be more to talk about some basic concepts, in fact, for humans, concept is a carrier of knowledge, it can inspire children in the future When encountering such a problem, at least know that there is such a concept, can trigger the corresponding or different solutions, in fact, now it is innovation.
When we encounter a lot of problems, such as we do scientific innovation activities for teenagers, I find it very interesting. We talk about picture recognition, it actually has many applications, you may often see the bus, or the subway turnstile, or the access control of the face of the entrance, you will feel that this is an application we often see, including like Alipay we often say that face verification is also one of them.
Like our Youth Future Artificial Intelligence Summer Camp, there are actually some students who I think they want to use artificial intelligence technology to solve, and they may use the current technology or the method they can think of at the moment cannot be solved. For example, there is a very good case, when hanging water, in fact, some nurses may not be very experienced, especially for children whose blood vessels are relatively small, the needle will sometimes be inserted inaccurately, which may involve inserting it several times. So we have a kid, and I think it's very good, he designed a system, there's a camera, it's able to shoot blood vessels, and then through image enhancement, the blood vessels can be more clearly displayed to the nurse, and then the nurse has more data and information to be able to pin the needle. So some such projects, in fact, through some of the disciplines we are learning now, it is difficult to think of, but if children know that the machine can actually see the world, the machine can actually enhance, help us use the machine method to enhance some information in the world, and even this machine can analyze, the human brain is difficult to analyze or capture some information, based on such a scene, they can actually complete such projects.
In addition, there are some concepts about machine learning. Most of our current learning, like chemistry and biology, is probably more of an inductive method. Let's talk about Mendel's pea experiment, as a scientist, in order to understand the principle of genetic pairing, he planted a variety of peas in a monastery for a lifetime, did a lot of data collection, and finally came up with a theory that I think will be learned later in biology.
In fact, in a deductive discipline like physics and mathematics, we talk about what the theorem is and what the formula is, and then deduce and solve more problems in this way. In fact, this is in the thinking of artificial intelligence, we say, it is another new perspective, it is based on the current data of a perspective, many solutions to the problem, may be because the machine has the ability to capture and observe, analyze larger data, and then can get, it may be different from our traditional induction methods and concealment methods.
Therefore, in the entire learning process of artificial intelligence, from enlightenment to entry to foundation, it may be more suitable for people with some basic science and technology and information technology literacy to learn. In fact, like the spiraling process, from the basic idea, and then finally we can use such a core thinking, use such a core skill, to solve or innovate to solve practical problems in life or work, even on the basis of some thinking paradigms that did not exist before, we use such some thinking paradigms, and then do innovation.
I am Dai Juan, I am now doing education at SenseTime, and I am also doing some public welfare undertakings of artificial intelligence science education for teenagers in the intelligent education center of Shanghai Artificial Intelligence Lab. I actually began to engage in this direction of artificial intelligence for a long time. I studied at the University of Science and Technology of China from 2001 to 2004, my undergraduate was in electronic information engineering, freshman, sophomore, and junior years are actually my scientific research direction is signal processing, the direction of communication as everyone knows. When I was doing my internship, I was doing some radar signal satellite processing at the Institute of Electronics of the Chinese Academy of Sciences. In my senior year, that is, in 2004, I was fortunate to meet Professor Tang Xiaoou, who was at Microsoft Research Asia as the head of the CV group - Computer and Vision Group, and I got an opportunity to change from the major of signal processing. In 2004 and 2005, it was really a purely academic direction, and at that time it was not called artificial intelligence, it was called pattern recognition and CV, and it was called computer vision.
In 2005, after I graduated from the bachelor's degree, I went to the multimedia laboratory of Hong Kong Chinese University, which is also the laboratory of Mr. Tang, and continued to do artificial intelligence related scientific research, my scientific research direction is 2D, 3D face classification and face recognition, but in the era there is actually no deep learning this way, nor the way of neural networks, mainly in some ways of statistics.
In 2007, when I graduated from graduate school, I reached an inflection point in my life to decide whether I was going to do a PhD or do something else. I was fascinated by the products. The reason? Because in fact, I think whether it is to do research, or to go to the industry or the company to do products, in fact, there is a very essential point, what we do can bring some changes to the world to a certain extent, big or small.
I was actually very touched at that time, I came into contact with some very early product managers at Microsoft at that time, and I felt that doing products could change the world on a larger scale with some of our own ideas. In fact, I heard a sentence at that time, a product manager of Office said, we make something called Touch millions of lives, which can touch many, many lives, so I heard it at that time. After graduating from graduate school, I joined Microsoft's Academy of Engineering in China and became a founding member of the first group of Microsoft Engineering Academy to do education. So when I joined Microsoft, I was doing education at Microsoft, which was very interesting, and I did a lot of education-related work at that time.
It wasn't until 2009 that I moved from Microsoft China to Microsoft's headquarters, Redmond City, in the Greater Seattle area of Seattle, and actually turned around to do the input plan. In fact, to put it simply, the mouse and keyboard we are talking about are on a touch screen. In fact, the artificial intelligence we are talking about today, a touch keyboard that everyone looks ordinary, in fact, there is a lot of technology behind it, and its entire design may be the most complex design on the mobile phone in addition to games.
So at that time, it was also with my team, from 2009 to 2014, it was a system of input methods for all languages on Windows and Phone, and in 2014, our English system was actually quite interesting, breaking the Guinness World Record for the fastest text input on the touch screen, so this is also very interesting. It turns out that we can do engineering and products, and we can also do the world's top such a way.
In 2014, for various reasons, our family moved from the Seattle area to the Bay Area, and as you may know, in a place like Silicon Valley, I joined Apple's Siri team and did a lot of things that I thought was very interesting with the Siri team. Siri from 2014, we just talked about artificial intelligence at some point may still be this kind of speech recognition defined by rules, and then include a way of natural language understanding, and we converted it into a way called machine learning. At the same time, I also manage the entire Siri data science team in Siri, there are a large number of people will use Siri, its implementation in our call content operation or not, Internet companies have such a team, what kind of people, what kind of functions, based on the actual usage data of these users, we can better plan some of the functions of Siri in the future.
In 2018, my family moved back to China from the United States for family reasons, and I joined SenseTime. Since the beginning of 2018, SenseTime has created this team of SenseTime Education from zero to one; We began to focus on youth AI education and the improvement of teenagers' AI literacy, until now.
When asked why I did artificial intelligence education for teenagers, SenseTime did not educate this division, which I founded after returning to SenseTime.
First, I think SenseTime has this foundation to do artificial intelligence education for teenagers, and the way we do it is to transform the cases in all aspects of the industry that we see now into some courses and cases that teachers can teach and students can learn. We are telling children from the perspective of technology, what do you see in the application of artificial intelligence, what is the logic of thinking behind this application.
Second, we see that many industries have been infiltrated by artificial intelligence, so we are thinking that in the future, children, such as he to be doctors, teachers, and content operations of Internet companies, have no data thinking behind them, without this kind of data analysis, intelligent analysis of this thinking, without such a model, children can be better qualified for such work.
Based on this thinking, I think that there is actually a big problem now, especially in Chinese education, not necessarily a little problem in his cognition, the biggest problem is that I think it is still more traditional, maybe more encouragement, let him try more. In fact, frustration is also an accumulation of experience.
So I think from this point of view, we can make everyone relax, encourage the children more, and they may grow up healthier. And the children's current anxiety has been around since elementary school, which is bad. I have always told all the parents around me not to have chicken blood, and your chicken blood will affect a group of parents. Our children are very relaxed, so I always feel that advocating more encouragement.
The future is the era of technology, which I think is a very certain thing. Technology itself is a fast-moving thing, so its uncertainty is actually very high. The future must be an uncertain future, under such a future, what kind of literacy do our children, classmates and teachers need to have? In the face of the future, what way can we better face such uncertainty? A good way, in my opinion, to read widely, to understand widely, with an open mind, to learn more about some information in different knowledge areas of the industry, to understand some of their paradigms, in the future in the face of uncertainty will be more calm, will be more comfortable, so that children are not afraid of the future. Thank you.
【Fan Deng Reading】 column manuscripts, unauthorized handling, prohibited transportation. Paper books, please pay attention to [Jingguan Mall]