How Google’s Former China Chief Thinks AI Will Reshape Teaching

Exclusive | EdSurge Podcast

How Google’s Former China Chief Thinks AI Will Reshape Teaching

By Betsy Corcoran     Dec 11, 2018

How Google’s Former China Chief Thinks AI Will Reshape Teaching

Artificial intelligence promises to have a dramatic—and yes, disruptive—effect on U.S. education and jobs in the next decade. But that technology won’t be entirely homegrown: Chinese companies, particularly those building products or services laced with the machine learning algorithms, are increasingly playing a role in the tools that we call “AI.”

There are few that understand what these forces mean for the world—and for education and learning—better than Kai-Fu Lee.

Lee has been an enormously influential researcher, driving forward work on AI. Originally from Taiwan, he came the U.S. at age 11 and went on to earn degrees from Columbia University and Carnegie Mellon University. He then went on to work at Apple, Microsoft and Google, where he served as president of Google China. He started a venture capital firm in 2009 based in Beijing called Sinoventures. He’s written eight top-selling books in China and has more than 50 million followers on social media.

His latest book, “AI Superpowers: China, Silicon Valley and the New World Order,” is almost two books in one: It tells the story of the development of artificial intelligence and why we should pay attention to this work.

In the past 2.5 years, billions of dollars and countless researcher hours have been poured into developing “narrow AI”—specific applications that AI can do better than any human—or even team of humans. All that has led Lee to write this book and to make a stunning prediction: That in the next 15 years, close to half of today’s jobs in the US will be done by machines powered by artificial intelligence.

EdSurge caught up with Lee in California this week in Palo Alto. Here’s why he believes that AI—and particularly AI developed by Chinese companies—is fated to rock our world, and how we learn.

Listen to the discussion on this week’s EdSurge On Air podcast. You can follow the podcast on Apple Podcasts, Stitcher or wherever you listen. The transcript below has been lightly edited and condensed for clarity.

EdSurge: In your book you tell us that in 15 years, close to 50 percent of the jobs in the U.S. will be able to be replaced by AI. That's still a pretty big prediction. Why do you think that’s going to happen?

Lee: Because of AI super-performance. Part of the reason I call the book AI Superpowers is that AI does have superpowers. Not in terms of the science-fiction robot overlords. I mean more that in a single task, with a huge amount of training data, AI is beating humans by unbelievable proportions.

For example, we have an investment in a loan application that uses AI to make loans to people who request $200 or $500, and it's sent to them instantly over the phone. You would think such an app would have a default rate of 70 or 80 percent, but it has a default rate of 3 percent, because it’s able to really use all the data about the user to make a very, very safe decision when the loan is given.

With capabilities like that, why would there ever be another loan officer? Why would there ever be another policeman whose job is to recognize faces? Now, there are other things policemen could do, but part of the tasks is gone, and so is part of the workforce.

Let's talk about education. If you could wave a magic wand, what would be the ideal classroom?

If I could wave a magic wand, I’d make everything go away and start from scratch.

Everything?

Everything. But I can’t. So, that’s just a dream. But everything has to fit within the current expectations and environment. If you think about how everything has progressed, the way an office used to look like, and what it looks like now, and if you think about almost any profession, it’s been completely transformed by technology over the last hundred years. If you look at the picture in the classroom today, versus a hundred years ago, they are the same.

In the average classroom in let's say in the Midwest U.S., or middle China, you will see one teacher and 50 students, or 20 students, each behind a desk, raising their hands, and the teacher lecturing. Now obviously, there are good teachers, there are teachers who go in-depth, who tell good stories and so on. But the lecturing should be done by the masters. It should be getting a physics lecturer from the Nobel prize winner, who also is a great teacher. Everyone should be learning from that teacher. Perhaps in a new form that we are investing in, in China, which is a one-to-1,000, student-teacher ratio.

One-to-1000? Say more about how that could possibly work, because that sounds as rote as possible. One person lecturing to a crowd of 1,000?

Lectures should be given by masters. If you think about Khan Academy, you know what an amazing teacher can do, compared to regular teachers. In the one-to-1000 classroom, we also have teaching assistants. So it’s not like you’re watching a giant video because at the end of, let’s say a 30-minute lecture, there’s another 30 minutes for someone close to you to interactively take questions and help you with points that you didn’t get. So that kind of combination should replace the current type of lectures that people get.

And I would also argue, it would become more effective because the top teachers are going to be very good in getting the point across. And then, the local teachers can be trained to be teaching assistants, so a job that's much easier for each teacher to learn to do. So it’s a split of responsibility, and the students ought to understand the subject much faster, and therefore have more time for other activities, which really is the core and essence of the future of education.

Another innovation that we’ve seen, particularly in China, has been the use of artificial intelligence to monitor facial expressions of students. Talk a little bit about the pros and cons of using AI to monitor students' reactions.

If you think about the job of a teacher today, we just took away the lecture part and turned it into an assistant job. Teachers also have to take attendance. AI can recognize students and the way in which they participate, and it can certainly save time from taking attendance.

Now, some parents will start to worry, say, "Wait a minute, I don't want my kids captured on video all the time." But the benefit is that the system will know the comprehension level of the students, not just by how they do on exam. By that time, it’s too late. If they're sleeping, or not paying attention in class, there can be gentle ways to deal with that fact. A teaching assistant might offer after-school tutoring. There might be apps that the student can go home and catch up. There might be a recording of the lecture that a student missed. And it’s important that the teacher knows the individual progress and competency of each student so that students can be grouped in the super class, and the class that's doing average, and the class that needs remedial help.

So those are all possible things that can be done. For people who are concerned about the surveillance nature of such an application, [it can be developed] so that there is no such nature.

How would you do that?

Well, don’t rat on the students. Don’t tell their parents. Don’t use it for punishment. Use it to enhance learning.

The output of the AI is up to the designers of the AI software. So only have the software developers target on improved comprehension and participation of the students—not in a punitive way.

Knowing the education systems, whether it's in the U.S. or in China, do you think that we have a gentle enough institutional framework that will use this kind of powerful technology in a way that will protect students?

In U.S., you do. In China, probably not yet, because China’s more rote learning. But interestingly, going back to what else AI can do, in China, we’re using this to currently just to take attendance, nothing more. But [AI] will have to be cautiously expanded.

Let’s switch subjects to turn to the entrepreneurial world. Your book is a fascinating description of the “gladiator capitalism” in China these days. If you were to advise a startup company, maybe even a startup company in education technology, that's thinking about going into China, what would be your advice to them?

I would advise, don’t.

Don’t come to China. Why?

Because the education system is so different. Most U.S.-based edtech companies are focused on some element related to the American system: to how to sell software to schools in a particular city or county, and how to do payment within schools.

The Chinese system is very different. Schools have a very small budget and their approach to teaching is more rigid and not as open to adopting new software unless it immediately leads to better performance, better outcomes. Chinese educators are very outcome focused. If you want to show them something that improves the kids’ grades measurably, they'll look at it, anything else, for productivity and things, it's not clear they'll use it.

The Chinese way to monetize education is through parents. Chinese parents are willing to spend whatever it takes, half of their monthly salary, if the kids can get into better schools. Because China has such a focus on education, the Chinese society depends so much on education performance as a predictor of future success—material success and income.

When you look at the opportunities for edtech and AI in the U.S. market, do you see those two factors as hindering the development of AI here?

I think edtech in the U.S. is not a hot area for investment. The kind of companies we have seen fall into two categories: One is selling stuff to schools—and those tend to be productivity, administrative things that are very hard to monetize because you have to have a dedicated sales force for a modest budget of the schools. The other set of companies we've seen are companies that aim to disrupt education—companies like Minerva and Outschool—which are incredibly aspirational and exciting. But it will take decades to see the outcome and change the way people think.

The positive side of this is that the American education system is so good. So to tweak it is not very lucrative. And it, is very hard to disrupt because it’s already quite good. I know that many Americans feel that American education is no good. It may be no good, but it’s the best in the world.

Your company, Sinovation Ventures, has invested in, on the order of about 10 different U.S. education technology companies. Tell us a little bit about what you've been looking for, and are you investing in them because you believe in their potential in the U.S. market? In the Chinese market? In the worldwide market? All of the above? none of the above?

We have invested in quite a few edtech companies. And we thought there were linkages we could draw and perhaps bring the technologies to China. In some cases, we just wanted to invest to learn about what's happening here. So we’ve given our U.S. team a pretty small budget.

Artificial intelligence promises to have a dramatic—and yes, disruptive—effect on U.S. education and jobs in the next decade. But that technology won’t be entirely homegrown: Chinese companies, particularly those building products or services laced with the machine learning algorithms, are increasingly playing a role in the tools that we call “AI.”

There are few that understand what these forces mean for the world—and for education and learning—better than Kai-Fu Lee.

Lee has been an enormously influential researcher, driving forward work on AI. Originally from Taiwan, he came the U.S. at age 11 and went on to earn degrees from Columbia University and Carnegie Mellon University. He then went on to work at Apple, Microsoft and Google, where he served as president of Google China. He started a venture capital firm in 2009 based in Beijing called Sinoventures. He’s written eight top-selling books in China and has more than 50 million followers on social media.

His latest book, “AI Superpowers: China, Silicon Valley and the New World Order,” is almost two books in one: It tells the story of the development of artificial intelligence and why we should pay attention to this work.

In the past 2.5 years, billions of dollars and countless researcher hours have been poured into developing “narrow AI”—specific applications that AI can do better than any human—or even team of humans. All that has led Lee to write this book and to make a stunning prediction: That in the next 15 years, close to half of today’s jobs in the US will be done by machines powered by artificial intelligence.

EdSurge caught up with Lee in California this week in Palo Alto. Here’s why he believes that AI—and particularly AI developed by Chinese companies—is fated to rock our world, and how we learn.

Listen to the discussion on this week’s EdSurge On Air podcast. You can follow the podcast on Apple Podcasts, Stitcher or wherever you listen. The transcript below has been lightly edited and condensed for clarity.

EdSurge: In your book you tell us that in 15 years, close to 50 percent of the jobs in the U.S. will be able to be replaced by AI. That's still a pretty big prediction. Why do you think that’s going to happen?

Lee: Because of AI super-performance. Part of the reason I call the book AI Superpowers is that AI does have superpowers. Not in terms of the science-fiction robot overlords. I mean more that in a single task, with a huge amount of training data, AI is beating humans by unbelievable proportions.

For example, we have an investment in a loan application that uses AI to make loans to people who request $200 or $500, and it's sent to them instantly over the phone. You would think such an app would have a default rate of 70 or 80 percent, but it has a default rate of 3 percent, because it’s able to really use all the data about the user to make a very, very safe decision when the loan is given.

With capabilities like that, why would there ever be another loan officer? Why would there ever be another policeman whose job is to recognize faces? Now, there are other things policemen could do, but part of the tasks is gone, and so is part of the workforce.

Let's talk about education. If you could wave a magic wand, what would be the ideal classroom?

If I could wave a magic wand, I’d make everything go away and start from scratch.

Everything?

Everything. But I can’t. So, that’s just a dream. But everything has to fit within the current expectations and environment. If you think about how everything has progressed, the way an office used to look like, and what it looks like now, and if you think about almost any profession, it’s been completely transformed by technology over the last hundred years. If you look at the picture in the classroom today, versus a hundred years ago, they are the same.

In the average classroom in let's say in the Midwest U.S., or middle China, you will see one teacher and 50 students, or 20 students, each behind a desk, raising their hands, and the teacher lecturing. Now obviously, there are good teachers, there are teachers who go in-depth, who tell good stories and so on. But the lecturing should be done by the masters. It should be getting a physics lecturer from the Nobel prize winner, who also is a great teacher. Everyone should be learning from that teacher. Perhaps in a new form that we are investing in, in China, which is a one-to-1,000, student-teacher ratio.

One-to-1000? Say more about how that could possibly work, because that sounds as rote as possible. One person lecturing to a crowd of 1,000?

Lectures should be given by masters. If you think about Khan Academy, you know what an amazing teacher can do, compared to regular teachers. In the one-to-1000 classroom, we also have teaching assistants. So it’s not like you’re watching a giant video because at the end of, let’s say a 30-minute lecture, there’s another 30 minutes for someone close to you to interactively take questions and help you with points that you didn’t get. So that kind of combination should replace the current type of lectures that people get.

And I would also argue, it would become more effective because the top teachers are going to be very good in getting the point across. And then, the local teachers can be trained to be teaching assistants, so a job that's much easier for each teacher to learn to do. So it’s a split of responsibility, and the students ought to understand the subject much faster, and therefore have more time for other activities, which really is the core and essence of the future of education.

Another innovation that we’ve seen, particularly in China, has been the use of artificial intelligence to monitor facial expressions of students. Talk a little bit about the pros and cons of using AI to monitor students' reactions.

If you think about the job of a teacher today, we just took away the lecture part and turned it into an assistant job. Teachers also have to take attendance. AI can recognize students and the way in which they participate, and it can certainly save time from taking attendance.

Now, some parents will start to worry, say, "Wait a minute, I don't want my kids captured on video all the time." But the benefit is that the system will know the comprehension level of the students, not just by how they do on exam. By that time, it’s too late. If they're sleeping, or not paying attention in class, there can be gentle ways to deal with that fact. A teaching assistant might offer after-school tutoring. There might be apps that the student can go home and catch up. There might be a recording of the lecture that a student missed. And it’s important that the teacher knows the individual progress and competency of each student so that students can be grouped in the super class, and the class that's doing average, and the class that needs remedial help.

So those are all possible things that can be done. For people who are concerned about the surveillance nature of such an application, [it can be developed] so that there is no such nature.

How would you do that?

Well, don’t rat on the students. Don’t tell their parents. Don’t use it for punishment. Use it to enhance learning.

The output of the AI is up to the designers of the AI software. So only have the software developers target on improved comprehension and participation of the students—not in a punitive way.

Knowing the education systems, whether it's in the U.S. or in China, do you think that we have a gentle enough institutional framework that will use this kind of powerful technology in a way that will protect students?

In U.S., you do. In China, probably not yet, because China’s more rote learning. But interestingly, going back to what else AI can do, in China, we’re using this to currently just to take attendance, nothing more. But [AI] will have to be cautiously expanded.

Let’s switch subjects to turn to the entrepreneurial world. Your book is a fascinating description of the “gladiator capitalism” in China these days. If you were to advise a startup company, maybe even a startup company in education technology, that's thinking about going into China, what would be your advice to them?

I would advise, don’t.

Don’t come to China. Why?

Because the education system is so different. Most U.S.-based edtech companies are focused on some element related to the American system: to how to sell software to schools in a particular city or county, and how to do payment within schools.

The Chinese system is very different. Schools have a very small budget and their approach to teaching is more rigid and not as open to adopting new software unless it immediately leads to better performance, better outcomes. Chinese educators are very outcome focused. If you want to show them something that improves the kids’ grades measurably, they'll look at it, anything else, for productivity and things, it's not clear they'll use it.

The Chinese way to monetize education is through parents. Chinese parents are willing to spend whatever it takes, half of their monthly salary, if the kids can get into better schools. Because China has such a focus on education, the Chinese society depends so much on education performance as a predictor of future success—material success and income.

When you look at the opportunities for edtech and AI in the U.S. market, do you see those two factors as hindering the development of AI here?

I think edtech in the U.S. is not a hot area for investment. The kind of companies we have seen fall into two categories: One is selling stuff to schools—and those tend to be productivity, administrative things that are very hard to monetize because you have to have a dedicated sales force for a modest budget of the schools. The other set of companies we've seen are companies that aim to disrupt education—companies like Minerva and Outschool—which are incredibly aspirational and exciting. But it will take decades to see the outcome and change the way people think.

The positive side of this is that the American education system is so good. So to tweak it is not very lucrative. And it, is very hard to disrupt because it’s already quite good. I know that many Americans feel that American education is no good. It may be no good, but it’s the best in the world.

Your company, Sinovation Ventures, has invested in, on the order of about 10 different U.S. education technology companies. Tell us a little bit about what you've been looking for, and are you investing in them because you believe in their potential in the U.S. market? In the Chinese market? In the worldwide market? All of the above? none of the above?

We have invested in quite a few edtech companies. And we thought there were linkages we could draw and perhaps bring the technologies to China. In some cases, we just wanted to invest to learn about what's happening here. So we’ve given our U.S. team a pretty small budget.

Trending

Get our email newsletterSign me up
Keep up to date with our email newsletterSign me up