World renowned AI thought leaders Dr. Kai-Fu Lee and Dr. Olaf Groth discussed how AI is changing the future of economy, government, and society.
Dr. Kai-fu Lee is the CEO and Founder of Sinovation Ventures and author of the New York Times’ best selling book “AI Superpowers: China, Silicon Valley, the New World Order”. Dr. Olaf Groth is the Professor of Global Strategic Management, Innovation & Economics of Hult International Business School, CEO of think tank network Cambrian.ai, and co-author of the just published new book “Solomon’s Code: Humanity in a World of Thinking Machines”.
This event is presented by Silicon Valley Future Academy, with support from One Piece Work, Bay Area Science and Innovation Consortium (BASIC), Corporate Innovators Huddle (CIH), and TalentSeer.
Part 1: Future of AI Talk by Dr. Kai-Fu Lee
Part 2: A Dialogue on What’s the Future of AI with Dr. Kai-Fu Lee and Dr. Olaf Groth. Moderated by Dr. Chenyang Xu, IEEE Fellow and Co-Founding Partner, Silicon Valley Future Academy.
Part 1 Transcript: Future of AI Talk by Dr. Kai-Fu Lee
Thank you very much for so many great welcomes. This is the first time I've had four welcoming speeches, as well as being graced by the mayor of Dublin. So why not add one more introduction? So here is someone else who will introduce my talk - (President Donald Trump’s voice) “(It’s a great thing) to build a better world with artificial intelligence. 人工智能正在改变世界。” And how apt it is that today's sponsor is iFlyTek, because this technology came from iFlyTek. That was not President Trump speaking, but iFlyTek’s speech synthesis based on artificial intelligence. So I think this demonstration shows you the power of artificial intelligence and the advanced states that China is in.
So a lot of the current progress in artificial intelligence is actually in the field of machine learning. And the biggest breakthrough in machine learning is a technology called “deep learning”. Many of you are experts, but maybe some of you are not so familiar. So let me take just thirty seconds to explain. What deep learning is, is taking one single domain, and a huge amount of training data, and teaching a system to do one task much, much better than human beings. But keep in mind, lots of data required, the more the better, and only works in one domain. It is not something that can reason, analyze, conceptualize or create. So for those of you who are not familiar, that's all you need to know for the rest of my talk.
So with this technology, what is it good for? In Sinovation’s portfolio, we think of AI as having four waves so far. And the first wave is clearly in the internet space. Because if you remember what I said, what is single domain? A huge amount of data. That's of course, internet - internet applications from Amazon, Google, Facebook and China's Tencent and Alibaba. And these are companies that have benefited greatly from the data that we willingly contribute, and also the label that we freely label tag as lab rats of these companies. So if you don't remember labeling for these companies, actually every click you make on Amazon and Facebook, every purchase decision you make is contributing to their companies - learning what you like to see, what you like to buy, and also what other people like you like to see and buy. And that allows them to essentially have a knob that helps them determine business metrics.
Many of you are CXOs. So imagine a CEO who has not only a great staff to work with, but an AI engine that can be optimized to maximize user attention, maximize user minutes, maximize total revenue, maximize total profits. So depending on the stage of your company, you just twist the knob, numbers come in, VCs chase after you, people buy your stock, and before you know it, you are a trillion-dollar company. And that is why the biggest powerful AI companies are all internet companies, because internet is obviously the space in which there is the most data, best knob to tune and most single purpose. So that's just the first phase.
Now what about the second phase? Well, many other people have data, businesses have data, banks have data, insurance companies have data, research analysts have data, investment banks have data, hospitals have data and so on and so forth. And all that data used to be just in repository. The person running the data center – it used to be a cost center. But now with AI, that cost center becomes a gold mine, on which you can make predictions and make money or save money. As an example, a bank used to have to record all of our transactions just as a matter of their archives. But now those transactions can be used to help the bank do better credit card fraud detection, target the right products to sell to you, to decide if it makes a loan to you or not, help you with the asset allocation, estimate how much money you have, and try to get you to spend more. And all those things that enabled Amazon to get rich are now also available to banks.
But before you think we're just going to enable all the traditional businesses, I do want to give you one example of a company that we funded that does banking function very differently, so that you're starting to think not just as traditional companies using AI, but disrupters - disrupting traditional companies. So let's think about the process of getting a loan, right? Going to a bank, filling out the form with all the things that they may want to know - your address, name, place you work, how much you make, how much asset you have, where you rent, where you buy, and so on and so forth. And then the loan officer decides whether to give the money to you. So now let's consider a new form of loan done completely as an internet app.
We funded this company that gives small loans, let's say, five hundred dollars loans to people for three to six months, and at a competitive interest rate - competitive with credit card like numbers. But the thing that's different from the bank is, no bank will give you a five-hundred-dollar loan. This app does. And think about the possible default rate. How many of you would be willing to go out on the street and meet the first three thousand people, and pick one thousand of them to give these five hundred dollars to each, and remind them to pay you back at a twenty four percent annual interest rate? What kind of default rate would you have? How many of you are really willing to do that? - Nobody. But what kind of default rate would you likely have? Eighty percent? Seventy percent? - Some huge number, right? Even if you got their ID, they may or may not repay you. So what happens? Why would someone create a business like that in China? Well, because they could use AI to reduce the default rate down to three percent.
So how do you get it down to three percent? Well, first you collect all the data as a loan application does, and you feed it into machine learning, but not just those ten things you fill out on the form, but also things like how long did it take you to type your address? Because if it's a fake address, it might take a long time. And also it asks you, would you be willing to upload data from your phones as a part of the loan consideration? Not private data, not super private data, but data that Facebook, Google, Amazon would ask you, or Snapchat, and you would regularly click “Yes”. So nothing more or less than that permitted by android and ios. So with that information, all of a sudden, there are not ten pieces information but three thousand pieces of information.
You might say, how could there be so many? Well, there is - What apps you have? Do you have games? Do you have apps that allow you to buy marijuana? Do you have gambling apps? Do you have serious apps like Quora or China’s Zhihu? Or what kind of phone you have? What's the model? How long have you had it? What's in your contact list? With the person you call mom and dad, are they related to your mom and dad? Do the phones match? Also what day of the week is it? What day of the month is it? Is it before your payday or after your payday? Before payday? - May be a good loan. After payday? - Not so good. You just got paid. Why do you need a loan? So it has all these information. Just out of curiosity, we looked at the three thousand features and checked the least important feature. So the least important feature is your battery level. That actually matters. It matters very little, but it does matter. Why would your battery level matter as to your trustworthiness? Well, I suppose there's maybe some correlation with people who charge their phone like an OCD (obsessive-compulsive disorder) with people who return loans, and maybe a little tiny bit correlation for people who keep letting their batteries run out with those who default.
So check your battery levels and see if you might get a loan. But it's actually a very tiny factor. So what happens is this deep learning algorithm takes this huge amount of data, including battery level, and decides whether to lend the money to you or not, at a three percent default rate, something that no human being can possibly do. Even if some magical, amazing super genius can do it, it's not worth the money. Five hundred dollar loan - it's only going to make you a hundred dollars on interest over a year. It's just not worth it. But for the machine, it takes one second and it's all profits, right? Because you get the money at a six percent interest rate, you lend it at a twenty four percent interest rate, and you have a three percent default. Do the math. And all you have is a bunch of servers and energy costs. So that is why no human being can ever beat AI on quantitative tasks like that.
So moving on to the third task about perception. Perception is about digitizing the physical world with information that was previously not captured. But now we capture it and use it to analyze and do things that couldn't be done before, such as Amazon Echo. We all talk to it. It listens to you and that’s the things that couldn't be done before. That's an example. About Chinese AI, often people think about face recognition. So I'll tell you a recent story that about something that again, no human can possibly ever do. There's a very famous Chinese singer, name is Jacky Cheung, Zhang Xueyou. I see we have an older crowd probably listen to him as do I. One of my favorite singers recently had a concert tour in China, four cities, and he now has a nickname “Policeman Cheung”. Because at his concerts, thirty Most Wanted criminals were arrested. So how did that happen? Well recently, these giant stadiums are wired with cameras to protect people from terrorists and alike. Just as our airports, even in the US, and certainly UK and other countries are. And those cameras are now high definition, connected to face recognition, connected to the Most Wanted database. Well, how large is that database? I don't know - probably a hundred thousand faces or more. And basically, the police apprehended any face that came up as they came in with a ticket, so they knew where they're sitting. So they apprehended about, I don't know - a hundred people? And with seventy of them they showed their ID - “Whoops, sorry, made a mistake. Enjoy the concert.” And the other thirty, “Well, come with me.” So I'm not here to advocate that is a good or bad app. I'm here to tell you the task that it does - recognizing faces out of a hundred thousand most wanted criminals is not a task that any human detective including Sherlock Holmes could ever do. We just don't recognize that many faces, and these face recognizers actually can do three million faces. So again, AI does stuff that human can just not even hope to do.
And beyond face recognition, of course we have things like autonomous stores where the user is tracked as he or she is walking around the store, buying things, looking at things, picking things up. Their expressions are captured. Was it with pleasure? With anger? With disgust? All of that can be captured, and it turns offline shopping experience with even more accurate annotated data than an online shopping experience, right? All Amazon knows is whether you clicked, whether you bought. But now this convenience store knows how you feel about the products. So this kind of “how you feel” will allow convenience stores in the future to become much more profitable, because it now knows who came in, who like our merchandise, who didn't like our merchandise, who bought one, and therefore they know how to deal with procurement inventory sales forecast, and also placement of merchandise to maximize revenue. So that magic tuning thing that Amazon had, is now also available to convenience stores that implement it.
Finally, in the fourth wave, autonomous AI, we’re talking about robots and autonomous vehicles. Robots will go from industrial, such as some inspection at an assembly line, such as a dishwashing robot, such as a robot that can do agriculture, do targeted fertilizing to maximize the growth and equalize the growth of all vegetables, to sort vegetables, to pick vegetables. We're going to have autonomous, unmanned farms. And we're going to have eventually, autonomous manufacturing plants, right? And then we're going to have restaurants with no dishwashers, all done by robots. By the way, you want to buy one? Made in Silicon Valley, a company called Dishcraft (Robotics). So if any of you want to buy one, you can buy one with the book. The book is twenty eight dollars. If you buy a dishwasher, you’ll get the book for free. The washer is three hundred thousand dollars, if you want it. But this is the way - it works, and you solve a very difficult problem. Charge a high cost with volume, the price comes down. In ten years, you can get one for your home. By the way, this washer is really magical. You just take tablecloth - everything, just dump it all in, it cleans everything up. So you can look for the price to come down, at some point we can all by one.
And then autonomous vehicles. I think we all know enough about that, I won't go into a lot of details. But that will disrupt the way we transport ourselves, as well as the future of logistics and delivery, dramatically bring down the costs, making it not only more efficient, but also more convenient, saves our time, and most importantly. They will be safer than human drivers. Now maybe they're only a little bit better, but in five or ten or twenty or thirty years, they'll be so, so much safer that millions of lives will be saved. And that eventually, we won't be allowed to drive, because we're going to be the only ones bring danger to ourselves in these perfectly driving precision machines. That might be thirty years out, but it is inevitable, and that will bring about tremendous cost savings and life savings to humanity.
So each of these four waves will bring about on the order of five to ten percent improvement to our GDP. And there will be more waves coming. Well, how do we know there are more waves? It's like in 1998, if you ask me: What about the internet? I would say, “Oh, there are these waves coming – there’re these website wave, there are these portal wave, search engine wave, and the browser wave. And these are going to create tremendous value.” But what I wouldn't know at that time is that, after that, there would be the e-commerce wave, social wave, mobile wave, and AI wave. So just as internet ended up having tens of waves, the same will happen with AI. So again, the five premises of AI are that, they work only with a large amount of data in a single domain. And if you have an expert, it works even better.
A little bit on US and China. US clearly leads in research - AI was invented in the US, deep learning was invented the US and Canada. And if you look at the top publishings of the top one thousand AI researchers in the world, you can see on the left, the top one thousand researchers as measured by h-index. I'm not sure Chen-Yang is among a thousand, I’m not - once upon a time I was. But if you look among them, US is sixty eight percent, China is six percent.
So what are we talking about? Why would China have any chance when US has such a dominant lead? Well, there are three important things we need to consider. First is that, AI isn't as hard as you think. AI is actually open source, getting easier. There aren't that many breakthroughs - the breakthroughs that have happened are now mastered by hundreds of thousands, if not millions of people. Yesterday I was talking to a prominent VC in the Bay Area, and he said, “We don't have an AI investment group anymore. Every company we invest, they must have an AI group. If they don't, we don't even consider them.” That is the popularity of AI, and the ease of AI. So if you look at the last sixty two years, there has been only one big breakthrough, and that's deep learning. And now the tool kits that are available are making deep learning easier and easier. By the way, these are the tool kits and lots of open source on Kaggle and TensorFlow, on Google Cloud, Amazon Cloud. It is the barrier entries that come down dramatically. And because they come down dramatically, we've gone from an early adopter phase of AI where you need a super expert, to the implementation phase of AI where you just need to know what problem you're solving, have a good business person, and hire a bunch of young people, and go solve the problem.
So actually, the last point where I said to do an application, you needed a super AI expert. This slide is a year and a half old, but I think today, it's no longer true for many applications. You can replace it with young engineers who know enough to solve the problem you have at hand, at least with wave one and wave two of AI, maybe even with wave three. So as an example, we train three hundred students a year with zero industry experience who have taken maybe one or two courses in AI. We take five weeks and have an industry manager help them. And here is an example that they've built. Here are eight students from China, undergraduate students under the leadership of someone who comes from the autonomous vehicle industry. They built this vehicle that is running inside Peking University, from any building to any building in four weeks. So now you see, this is really not rocket science anymore. And this is a very core reason why China has a chance. Because there is a massive number of Chinese AI engineers. And if you look at the top one million people who published in AI - maybe not a million - the top a hundred thousand people who published in AI, Chinese actually account for forty percent. So they're entering the pyramid in amazing numbers with literally, an army of AI researchers and engineers.
Secondly, AI products need to be innovative, and Chinese companies are innovative. This shows you the progression. This is actually the most miraculous page here for those of you who don't know much about China. Your impression may still be: China is a copycat country, which was true only ten years ago, on the left side. All of Chinese companies at that time, they couldn't even get funded unless they said what ’s the US website they were copying. That was the state of Chinese entrepreneurship ten years ago. Of course, this was asked in very nice ways, “Who inspired you?” “Has there been a demonstration?” “Could this model work in a more advanced country?” But what they really meant was, “What website are you copying?” That was ten years ago.
However, copying is frowned upon in Silicon Valley. But copying - there's nothing wrong with copying if you're not violating IP, which they're not. And copying turns out to be the best way to educate a massive number of people. Did we not copy, when we learned music or art? And then some of us became artists and musicians. The rest of us were just copycats. The same is true. Many of the copycats are still just copycats. They amount to nothing. I do agree, if you copy forever, you amount to nothing. But some of the copiers ended up figuring out how to do product market fit. And then they founded and built good products for China, products better than the American companies. Many of you agree from the green side that, WeChat is better than WhatsApp. And Weibo is better than Twitter - maybe not in the diversity of content, maybe not as much as used by the president in each country. However, Weibo is a better product in terms of multimedia ease of use.
So that was actually five years ago, and then today, the Chinese companies have come up with all these innovations. I won't go into them. Those of you who know them, know them; those of you who don't know, feel free to take a picture and look them up. The total valuation of the third column of orange companies, they were purely Chinese innovations. These companies don't exist (here). The concepts don't exist. In fact, maybe you should copy them to the US. And their total valuation - just those eight companies in the third column, the orange - is three hundred billion dollars. So we're talking about five Ubers. So that's that. And now they're being copied everywhere in the world. So the main purpose of this slide is that, Chinese companies can innovate. Therefore it's not going to hold them back in terms of implementing AI.
And today, China has developed into a parallel universe where the Chinese apps and the American apps are probably equally effective. And lots of people want to ask: Well, can Google succeed in China? But actually it's all too late. No American company can easily cross to the Chinese universe, nor can a Chinese company cross the American universe. For example, you all heard that Alibaba did thirty five billion dollars of sales in one day on November 11th, right? So they must have a great product, right? They do. So let's say they landed in America tomorrow. Would any of you buy from them? Of course not. You’ve got your Amazon Prime, right? That is the same answer as to will Google or Facebook succeed in China? So China now has the parallel universe.
Chinese entrepreneurs are incredibly tough. We have experts here studying the Silicon Valley culture in my book. Hopefully we'll get some people interested in studying the Chinese entrepreneurial culture. The Chinese business model is extremely different from Silicon Valley. Not as elegant, not as beautiful, not as technology savvy, not as visionary, but equally or more effective in making money and building a business that can last. I don't have enough time to explain the details. So I would just say that the core thing about the Chinese business model is to build a business that is uncopiable. Think about it - ten years ago, five years ago, when your business is surrounded by copycats. Whatever feature you have, will get copied if it's any good. So the only way you can win is to build a product that can't be copied. Well, features can be copied, technologies can be copied. Once you have an existence proof, things can easily be copied without any IP violations. I remember being shocked by how good Google Map was in 2004, and then two months later, Yahoo came up with the same features. So I think Silicon Valley tends to evaluate features in technology. But once someone has an existence proof, it really isn't that hard to build another one without from the clean room. And that, I think it’s the Silicon Valley being myopic and not seeing that, the only way to build a sustainable differentiation is to build an uncopiable product. That is, the product has so much incredibly ugly, complex details. And the attention to those details makes it so expensive and difficult and lengthy to copy. That is the barrier of entry, not some technology or product. I don't have time to go into details, get my book. So that's the third reason.
The fourth reason is probably not that important, but a lot of money going into China, more money into China in AI than the US, and I think that's somewhat important, and also lifting valuations of companies. Actually on the right is our iFlyTek’s stock. I haven't tracked the last two months - Chinese haven't done them too well. But at the time of this picture, it went from - I mean you would compare iFlyTek with nuance. And this graph shows that iFlyTek used to be half of nuance and now it's a lot more than the nuance. And similarly, with pre-IPO stocks such as Sinovation’s only investments, we have probably created five unicorns, just in the last two to four years. The newest company in this list started two years ago and now a unicorn. These are companies in AI for banking, autonomous vehicles, semiconductors, and on computer vision.
And this is probably the biggest reason, which is that, as I mentioned many times, the more data, the better it works. That's the nature of AI. The right hand side shows you, pick any algorithm you want. Some may perform better than others, but more data helps you more than any other algorithm. Therefore, in the era of AI, if data is the new oil, then China is the new OPEC. And China's data is not only larger in breadth, that is more users, but also more in depth, that is more usage. Because in China, people order take-outs ten times more than America does, that’s lots of data. People read the Facebook news feed equivalent a lot more than you read the Facebook news feed yourself. Even before the recent events, the average use per day, per user of China's Facebook news feed, a company called “Toutiao”, is seventy four minutes for news. And it's massively larger, and that's all data being generated and used. The total number of short video based social network, which doesn't exist in this country, has two hundred and forty million active users per day, not per month or a week, per day. So these massive numbers all become just rocket fuel for AI.
And of course, the most valuable data that is fifty times more than the US, is mobile payment. Many of you think of this as: “Oh, this is China's Apple Pay.” This is not. Apple Pay is tied to credit cards. It doesn't get you away from the two percent of the charge. That is essentially a task on the American economy. In China, the mobile payments are used by seven hundred million people to pay each other with no commission at micropayment levels. So it completely replaces cash and credit cards. And the total amount for last year was eighteen trillion US dollars exchanged. And that has a lot of impact. It will change China from an export economy to a consumption economy. It will change China from a savings economy to a spending economy. It will help entrepreneurs start companies and be able to collect money immediately without building up a million users. And of course, it will create huge data for AI and that is used not only by Tencent and Alibaba, but also the merchants that use the billing. So that is huge benefits.
And the final reason is the techno-utilitarian Chinese policies towards AI, which I understand you already have a discussion. I'll just summarize my top few ideas. First, the Chinese government regulators tend to let technologies go out and figure out how to regulate them as they work, or don't work. That is how the mobile payment took over in three years. In this country, there might be lots of voices against mobile payment by credit card companies and banks that might say things like, “Oh, those software companies, they can't manage your money. There could be hacks and frauds.” and all that. And then governments might have to carefully review things. But in China, the government said, “Well, let's give Alibaba and Tencent a try.” And they proved themselves worthy. So in three years they eradicated cash and credit card. Of course not everything is unregulated. Crypto currency became not only regulated, but illegal. So the government is not unwilling to regulate, but it will kind defaults, let things go.
The other thing is the central government sets a tone, such as the AI document. But the central government doesn't actually fund something. Each local government and state bank might consider to do something. So after the central government came out with the AI plan, we found banks were much more willing to pay for AI software that helped one of our companies become a unicorn. It wasn't through government funding, but it was through government setting a tone that caused the banks to be more open to AI. Another city Nanjing, decided to create the world's largest AI park, because they have good universities. A new city called Xiong’an, the size of Chicago, is being built with autonomous vehicles. And the clever thing they did was for the downtown area. The downtown had two layers of road. Top layer is for pedestrians – people, pets, bicycles. Bottom layer is for cars, autonomous or not. The benefit of that is it will eliminate the possibility of car hitting a pedestrian, as we saw in Phoenix some number of months ago. And therefore, even though US is ahead in autonomous vehicle, China may launch it faster, because the slightly behind technology is enhanced by infrastructure to improve safety.
So this is where I think the numbers are: China used to be nowhere, and it's basically going to catch up and maybe slightly be ahead of the US in implementation in about five years. Now I do want to say that, I don't think this should be, or is a race. Because the Chinese companies are building products only for the Chinese users, companies and people. So the gain of a Chinese AI company never comes at the expense of an American company. But nevertheless, people want to know where things stand, and that's where things stand.
So with a lot of factors, including having two engines of growth driving AI, having seven giants in the internet, hiring training talents, having lots of funds, especially the SoftBank, a hundred billion dollar fund, funding AI companies, and having all those open source pushing AI along, AI will rapidly grow and create a huge amount of value. PWC estimates about sixteen trillion dollars of net increase to the world GDP. McKinsey estimates about thirteen trillion, so huge numbers. These could help reduce poverty and hunger, but they will also raise a lot of issues. These are issues you've probably seen many times and are familiar with. I would just talk about one of the issues, job displacement. So as we talked about AI is capable of taking a single domain and does an amazing job that's superhuman. So what does that tell you? Well, how many jobs in the world are single domain optimization tasks? Lots of data are coming in, making a decision, doing something repetitively. Well, none of your jobs are repetitive or routine, but if you think about all the jobs in the world, there are many.
So these repetitive, routine jobs are going to be displaced by AI on the order of five, ten, fifteen years. Specific numbers may vary. Only the complex and the creative jobs are safe. The complex jobs are safe because (they have) multiple domains, not single domain. Creative jobs are safe because you're innovating, creating something that didn't exist as opposed to optimize it. So that obviously raises a lot of issues because both white collar and blue collar jobs will be displaced. We've seen the city announcement in the back office. They will slash fifty percent of the jobs. We've seen that GM cutting back fifteen percent. You see the trading floors are empty and many other jobs will be displaced. You saw the Google demo, that's going to replace customer service, telemarketing, telesales, and the list goes on.
And then the blue collar jobs turn out to be a little bit more difficult, because building an iPhone is not something robots can do yet, but washing dishes, picking fruits, inspecting assembly line, they can. Replacing cashiers - on the right hand side is a company we fund. It's a pastry shop that has no cashier. Basically, you take all the pastries in a tray, and the computer vision sees what you bought and charges your WeChat Wallet directly. So it's a one-for-one displacement. The other example to the left of the pastry is F5 future store, another one of our investments. That one is displacing fast food. It's an autonomous fast food. You can see part of the store is actually a giant robot. That is, you can call it a robot. It has no arms and legs. It's just a giant cooking machine with microwave elements, oven elements, moving arms, etc. And what it will do is you can just go in and get a meal. A wonderful Chinese beef noodle, fish ball soup, salad and so on can all be had for less than two dollars per person. So who's going to eat at McDonald's in China, when Chinese food is available for half or a third of the price? Now that is not displacing human workers, because they started with no autonomous. However, to the extent that they gain fifty percent market share in the fast food space, well, there will be fifty percent layoffs in the McDonald's, KFC, and the other human operated fast food. So these displacements are everywhere. They're one-for-one, and therefore through industry disruptions.
So all these people coming off routine jobs displaced by AI, what will they do? Some economists will tell us, AI will create many jobs, which I tend to intuitively agree with. But it might take a long time. We don't know what those jobs are, and those are probably non-routine jobs. So even if there are more jobs available in five or ten years, those jobs are probably non-routine jobs. So routine workers who lose their jobs probably cannot easily get a non-routine job. So that's not a solution. Have all of them become CEOs and scientists? That's not likely to happen either. So we have a problem.
But if we think about it, it's not just creativity that is one dimensional. There is another element that AI cannot do, which is compassion, empathy, connecting with humans that trust the connectivity. There are many jobs that are held today that we would not want a robot to displace. So therein lies the hope. So we should actually look at the two dimensional graph, where the x axis shows the creativity and the y axis shows the compassion, empathy or human connection. And we can plot those jobs in various places. And then we're actually going to see many jobs could become available in the upper left corner, which is compassionate jobs that may not have a lot of creativity, but they are very much needed. And actually many of them are high growth jobs. For example, US in the next five years will need 2.3 million more people in health care for nursing at home care, elderly care. Because more people are living longer, and people over eighty need five times as much care as those between sixty and eighty. And people don't want to be cared by robots, and robots can't really care for people yet. So this is an area that's a very high growth area. But of course, they're not paid enough. So policies and training will be needed to train people were displaced on the lower left to become jobs on the upper left. As an example, Amazon is providing training for its employees, not just within Amazon, but also for jobs like aeronautic repair, something robots can’t do, or nursing, something robots can’t do. So I think there are enough jobs if we think about the upper left and how we can increase the pay and social status, and the training for the lower left to go to the upper left, because it's unlikely that the lower left can go to the right. So there is a possible answer.
So in terms of AI and people coexistence, I think the lower left will be displaced by AI, it's inevitable. The lower right will be - AI as tools to help scientists become more creative, help them come up with more drugs faster, thereby creating value and saving lives. And on the upper left, we can see AI as an analytical core that will help - let's say doctors - make their diagnosis. The diagnosis might be made by AI, but the doctor provides the human warmth and touch. So that can really make the cost of health care come down. The same goes with teachers. So it changes doctors and teachers to a more humanistic job - less training, lower costs, may be also reaching out to more people who can't currently afford the best education and health care. And of course on the upper right is where humans celebrate our ability to create and become compassionate. So there you have it, the coexistence blueprint for human and AI. So that covers the top in terms of the opportunities and challenges for the next ten or twenty years.
But if we look a little further, maybe thirty years, maybe for many of your grandchildren, when they go back and read the history book, what they are not likely to see are all the opportunities and challenges, all the value and job displacement and compassionate jobs. I think what they will remember AI, is something much more profound. I think they will remember two things. The first thing is that they will remember that AI is serendipity, because it has come to liberate us from routine jobs. This is the first time in the history of mankind that our grandchildren don't have to do routine jobs anymore, so they can do what they love and they can have time to think about what it means to be human. The second thing I think they will realize is that AI is not some (thing) scary, robotic, taking over mankind. But it's simply a tool that we control. And AI has no free will. We, and only we, have free will. So we will control the AI, make AI works for us and we're the ones will write the end of the story in the story of AI. Thank you.
Part 2 Transcript: A Dialogue on What’s the Future of AI with Dr. Kai-Fu Lee and Dr. Olaf Groth
Chen-Yang Xu: Thank you very much for coming, and I really appreciate that. I've seen a lot of familiar faces and those are my long-term friends and collaborators. But I also see your faces, so I'm eager to get to know you. Today in this room, we have talked about brilliant mind on the stage, but I have to acknowledge that there’re lots of brilliant minds in the audience, and I wish we have multiple days to have many of you on stage talking about this important topic.
So in the next session, we will have a dialogue and kind of piggy back on - Kai-Fu’s talk. This is my third time hearing Kai-Fu’s talk. I never get bored, it gets better every time. I've been in AI for about twenty eight years, Kai-Fu has been in for more than thirty five years and started in Columbia, then CMU and then became a professor there. Before AlphaGo beat human, Kai-Fu built a computer game I think called “Othello”, and was on New York Times back in the late eighties. And then he came to Silicon Valley, worked for Apple, Microsoft and then China, and come back today and share with us. On one little detail, I think you may not notice: in the reason one of the five reasons, there was a picture on the lower right, Kai-Fu didn’t mention. That was a picture of a stadium. The stadium must have more than ten thousand people, and the person on the stage was Kai-Fu. So in China he is a rock star. He really packed the stadium. But I'm glad today we have this very intimate setting for you to really hear him up close, and also from Olaf, which is my long-term team friend and collaborator. One thing interesting is with the thirty five years of experience, we talked about the AI rapid development - I call it “AI explosion”. I really mean it, this is not an exaggeration. Those who know me know I don't exaggerate.
I thought every one of us as an AI expert become so small in this explosion, and because there're so much frontiers get break open, because of the advance of a new technology, as a result, it's very hard to find any AI expert that know so much of the scope of AI universe, from policy, to the future of living, to jobs and technology, to company. So I'm just glad that we are hitchhiking Kai-Fu’s brilliant mind in the office, thirty five years travel in this AI universe, and both in US and in China, and in just last thirty minutes it’s scaled down. So obviously, read his book. I read his book, it's brilliant.
So with that said, I do want to devote the first question to Olaf. Because Olaf has been in corporate for many years and consulting for many large companies, and seeing the challenge in the cockpit, and the transform and innovation. And then he also is a co-author with Sean on the Bay Area Innovation Ecosystem years ago, and really looking at how Bay Area stays competitive and still as the major innovation spots connecting to all the innovation centers in the world, including China, like Zhongguancun. So he is really a world renowned expert on innovation, AI, new technology, futures. So I want to ask Olaf, what prompt you to write this book, and with of course, Mark - the co-author from Berkeley is also in the audience - and why do you believe the message in the book is so important right now?
Olaf Groth: Thank you for the question, Chen-Yang. First of all, let me express my gratitude for being here today. It's a real pleasure and a privilege to be with you and to be on the stage with Dr. Lee. I have read his book very carefully and I thought it was incredibly well written, and very, very educational, especially on the dynamics in China. So thank you for having me today.
The answer to your question is that, about two years ago, my very good friend Mark Nitzberg, who is sitting over here on the right, who is my co-author, who is a serial social entrepreneur had just sold his latest venture to Amazon, was taking a new role at Berkeley, getting back to his roots as an AI scientist. I had just taken a role of heading a research center at Hult International Business School and we were both looking at artificial intelligence, and found that the narrative on artificial intelligence was hopelessly over-hyped, and primarily on the dystopian sort of Terminator and Skynet side of things.
And while that's incredibly entertaining at times, it also really jaded our view to the tremendous opportunity that exists. And I think there's a lot of overlap bet ween Kai-Fu and myself on pointing that opportunity out. So Mark and I set out to craft a book that really could open our eyes on the tremendous opportunity, not just in the United States and China, but really globally that AI brings, not just for economic development and business development, but rather for human growth and development as well.
The second reason is inherent in what I just told you that is that, we felt that there is really no global look. So teaching in Hult International Business School, we have some students in the audience here, welcome! We have students from about a hundred and forty countries on each campus. And so we said “what a great asset to have to do some global research”, because there is a lot more going on in AI than just in China and the United States. So it's a net optimistic book, but we are also believing that by looking out very far to terminator scenario on autonomous weapons, etc., we are also preventing ourselves from looking at the risk that is already inherent today. In order for us to harness that short-to-be in term opportunity, we also have to be clear-eyed on the risk. So we believe ethics need to be front and center and parallel to the economic opportunity exploration, if we want to harness that opportunity.
Chen-Yang Xu: Okay, thank you Olaf. It was interesting, why I think this combination is a very potent combination, because when I read Kai-Fu’s book, it's actually better than reading even a science fiction, which is always my favorite genre. And it was fantastic, I read everything and in the end I said - because I worked for Siemens for seventeen years, I traveled to Europe a lot - and I said, but what about the rest of the world? In that book you obviously focused on the US and China - that premise is they are leading. And this is why I thought, to really give a global view - Silicon Valley is more global than anywhere in the world. We have actually one statistics that I always want to share, is that more than fifty percent of the startup founders in Silicon Valley are foreign born, and that's always striking my head when I share that with many of the CEOs visiting us, and they were always shocked. So diversity, that's why I thought (should) bring Olaf in, as he has a much global (background) in Europe and also other regions that we can combine very well to give a more comprehensive view. One thing is, I actually did my undergraduate in neural network, parallel computing back in the late eighties and early nineties. And when I graduated, I was very fascinated about the neural network. Everything is great, but I didn't - that was a bad timing, an AI winter, and I didn't solve any practical problem. And the only thing I could do was really some toy problems plus some handwriting recognition. So I decided to pivot to the computer vision and it was more pragmatic. But over the time, also there was a parallel on speech recognition as well. The PhDs, scientists got very frustrated because it improved very little, and so we got this up and down in an AI wave. Because AI is not new, even though suddenly showed up, it's actually since 1956.
So every time when they went up and then went down, and next time they went up, people said it may go down again. But I do share - when I read both of your books, and myself included, I was looking at where AI is going - I do think we have a shared belief, AI is now - also like Mayor Haubert talked about - it's real, it's no longer artificial. So I just want to ask both of you, what's new and different this time round, from the past, all the past up and down? Kai-Fu, why don't you take on this first?
Kai-Fu Lee: Sure. I think the last two times were based purely on faith or belief that AI would take off. There was some small trajectory that looked big, and then it hit a plateau. But this time, AI is really beating human in everything, not just in data-centric tasks, which we are not very good at, but also in human perception, AI is beating human speech recognition, image recognition. And I think in the not too distant future, even complex tasks like autonomous vehicle - it will drive better than people. It arguably already does for certain tasks like parking and (driving) on highway. So I think when it's truly exceeding human performance and by this much, it's for real.
Secondly, we're seeing the smartest people be really pouring into AI. The smartest kids are all going into AI. When I talked to a company like Pinterest, half of their engineers are working on AI. This is not even what you think of as an AI company. When you go to Stanford or Tsinghua, everybody studying AI is almost sad that many other important computer science areas are not being studied, because it's so hot. But when you have so many brilliant minds pouring into this area, problems will get solved. Also so much money has been poured into AI. Now that also will create a bubble. But (there are) lots of money and lots of brilliant minds, and I think traditional companies have capitulated. If you talk to any large car manufacturer, they've already accepted someday this will be autonomous driving, even though no one's demonstrated it. And if you talk to any ride sharing company, Uber or Didi, if you talk to any car parts manufacturer in Europe or Germany, they've all capitulated. So I think in a place where traditional companies have capitulated, smart people go in, and technologies are good enough, and lots of money is flowing in, I think it has to be real.
Chen-Yang Xu: Thank you. So let's turn the round to Olaf. What are the reasons that make you believe AI is real down? Why people in this audience and beyond, readers should take it very seriously, no matter what jobs they are having right now, if we really, really take it seriously about AI, this time it’s real?
Olaf Groth: Yeah. I think the reason for why this moment is special is essentially for three reasons. The first one is scale. The scale of available data is just tremendous, and we can talk a bit more about data and what it means. The second part is that, I think for the first time in our lives, we are now encountering a technology that could make decisions. They can actually make cognitive decisions for us. And with that, comes really the third, which is that AI is now making us think more about ourselves. And to me frankly, that is the most important part, because we are having to renegotiate our power, our trust with the people that innovate on AI, and with the tools themselves. Because with the ability to make decisions, comes a certain amount of leverage and power. So the companies that are running these AI systems are actually now making more and more decisions for us. They understand us in ways that we often don't see ourselves, and that can be tremendously valuable. But it is also a bit treacherous. So we need more transparency on that, but those are the three reasons for why I think this time is different, and certainly very different from the industrial revolution.
Chen-Yang Xu: Ok. Thank you. Yeah. Actually it's interesting - you actually almost answered my next question. Because we're looking at why this time it’s real - AI is around sixty years, a little bit over sixty. And Olaf just mentioned about one of the dimensions that he was looking into with Mark is on the power, right? That AI can bring to the world and to all sorts of jobs. So if we zoom out the timeline, let's not look at sixty years - the humankind has been on the planet for tens of thousands years, and let's look at maybe five hundred years. I think both of you talked about why this time AI is not just another big trend. It's so big, right? It's a megatrend that it actually rivaled the last industrial revolution, which was actually many, many years ago. So Kai-Fu, as you said in the book, leading AI deployment will translate into productivity gains on a scale not seen since industrial revolution, you said that just minute ago. What's new about the AI technology revolution in this longer time horizon, and why it is different from industrial revolution, for instance? Maybe Kai-Fu you can kind of expand on that because you’ve been talking about that in the talk.
Kai-Fu Lee: Sure. The industrial revolution took a long time because the infrastructure had to be built. Old factories had to be transformed. And take electricity as an example. It took over a hundred years to build out the electrical grid. So other innovations can be built on top of assuming there is electricity, and that was the serial slow process. I mean even today, cars are not yet electrified, right? - So the electrifying of the world takes forever. But AI is largely purely - for many of the applications - is purely software and can run on the clouds, and can be harnessed by people who can just write software, so it will come a lot faster.
The other thing that is different related to jobs is, industrial revolution actually easily created more jobs, because the very fact is instead of having crafts people - a small number of crafts people created a car in a long time - assembly line made many cars with many lower skilled workers much faster. But in AI, it is actually doing the work of the routine work at the assembly line. So it's both a benefit in a sense that it liberates us from routine work. But there's also a challenge, because it's not automatically creating jobs in as it improved productivity, the job creation will come later and it's hard to predict. So industrial revolution has been great for job creation. AI, I think it's more questionable and may take longer.
Chen-Yang Xu: Ok. So Olaf, you also mentioned industrial revolution. How do you see - why do you compare AI revolution with industrial revolution? I mean, that's - we know it's a massive scale, and what's your point of view on that?
Olaf Groth: Yeah. So first of all, I agree with Kai-Fu on this speed. This is coming at us much faster than we are likely able to adapt. And the industrial revolution took some time. It was very capital intensive and capital came in the form of machinery in property and factories. This is coming in the form of software in bits and bytes, and it's coming in the form of new ventures and applications that you can proliferate that you can market and scale up at near zero marginal cost. I thought the banking example was a beautiful example. So whether I market that up to a thousand people, or a billion people, the marginal cost of getting to a billion people is compared to the industrial revolution scale on economics, minute and zero marginal cost, or close to zero marginal cost. So it's much faster, and our human lives - the way we live our lives and we may think of ourselves as very agile as entrepreneurial executives. The Chinese are frankly in my experience, even faster and even more agile, but even at that we are not able to keep up with the speed of change to business models, to structures in our lives, our decision processes - how we organize our lives, and the institutions in our societies that are very ill-equipped. Just think about government institutions - no offense to the mayor or anybody else - but how much sophisticated AI talent or even medium skilled AI talent you have in these institutions? And yet, every area of life, even political life as we witness certainly in the United States, is being touched. So that's number one.
The second one is pervasiveness, which just dovetails the first one, as I explain. And then it's really about business models and industry being completely disrupted, and taking apart. And the industrial revolution built, and of course it changed the structure of the economy in long term, but there wasn't an established industry to be disrupted. Here we're now dealing with a very smart set of tools that are taking every industry apart. I have a lot of clients in automotive and industrial and infrastructure and engineering, in the chemicals and consumer goods, and they're all looking at this now because they're completely seeing these seven to nine - as we say, the global digital barons, Kai-Fu called them “oligarchs”, essentially absorbing these other industries. That is hugely disruptive, right?
Now on the positive side - because I told you our book is not positive - I think there is an amazing opportunity for human growth to actually get to the next stage. So as part of the book, you'll see examples of how AI is enabling human beings to learn better, or to be taught better, how to diagnose our health better, how to keep loved ones from having recurrences of diseases coming to their aid much more quickly, how to solve things, big things like climate change and food crisis, etc. So while we're looking at this very big disruption in large scale, there's also a phenomenal opportunity - you finally live lives that are much more purpose-bound and healthy than we have before.
Chen-Yang Xu: Yeah. Let's expand on that. Because I read your book, the central theme of the book by you and Mark is really on this notion: Human values and trust and power relationships are the foundations of our society. And you basically, together with mark, actually in that book outline how AI will change that fundamentally. And so to help the audience, can you give them sense what do you really mean? Can you give some examples on how AI will shape these three dimensions in a way that we really have to think about the world in a different way? Just help us understand that.
Olaf Groth: I'd be happy to. So first one on the concept of power which sounds very market billion, and it can be. But it's really just our personal agency, our ability to make decisions and choices, and to have leverage over the institutions and forces in our lives, right? It’s very, very important that we think very carefully about how we use AI, and how our relationship with AI and with institutions and other people and commercial interests is changing. For instance, the big one is of course that these seven to nine big, global internet companies are tremendously powerful, and they're gaining power through the deployment of AI. Google has something on the order of twenty supercomputers running, essentially it's Google world, that's an order of magnitude more than most small countries. And essentially they know you extremely well, they don't necessarily tell you how well they know you, because they don't want to scare you; Nor do they have any interests to scare you or to meta-label you in ways that doesn't align with their commercial interest - they want to sell you stuff. But we have to know that Amazon and Google know everything about you. And I think Kai-Fu gave some examples as well, but they know not just your income, your social status and who you're connected with. They also know your basic demographics, your race, your ethnicity, your family status, what relationship you're in - not because you told them, but because they can infer from statistical correlation, all of which is being turbocharged by artificial intelligence, including very sensitive things like your sexuality. And so that's a problem if you yourself may not have explored in depth who you are, right? So we have to understand that there's tremendous power in data.
And that gets me to the second point, which is: it's not just about our relationship with the big internet companies. And look, I'm a business professor. I like business. I love entrepreneurs; we have many in the room. And yet we also need to get to responsible entrepreneurship and in the era of AI, ethics is unquestionably tied at the hip with innovation. There's no more option to split those two. But the other one where this is true is that, different people in societies are empowered by AI very differently. I think Kai-Fu called this data “colonialism”, and we agree there are countries that are not yet as digital as the United States, China, or even Europe. And who is empowered there when people don't have a digital data footprint? Even within the United States, we are so - you are not just ethnically but socioeconomically very diverse. People who don't have a healthy, strong data footprint - let's say, not the stereotype but somebody from South Central LA, or Central Detroit, or even frankly, parts of East Palo Alto - who doesn't have a good data footprint, will be less empowered in the future than somebody who does, because advertisers are simply not interested in them, and the data is often corrupted and biased.
And we've already seen some examples with Uber and other companies. So we need to be very careful on how we exercise power on values. I mean, just look at the United States elections, the trouble that we're in with Facebook now. I mean with all due respect to Facebook, but that has to change, right? I mean, literally they're deploying algorithms that feed us information that's pushing us further into certain camps. So our views get more extreme, they get more elevated, our sentiments rise, and all of a sudden, we start hating each other when we really should start liking and loving and understanding each other better, right? So that has a lot to do with values. And then trust is really what comes out at the bottom. How much do you trust the machine and how much do you trust each other, based on that smart machine, that AI. And if you think about it, that is the most valuable currency in societies – it’s not money, we can always get more money - it's trust. Once that's gone, it's really hard to earn that back. And so I think we need to design AI that is trustworthy. So we don't take our societies apart.
Chen-Yang Xu: When things are interesting about AI, one of the dark sides of the current deep learning revolution is actually not explainable. Because I worked in the health care for many years, one of the biggest hurdles for doctors and patients to accept AI to begin diagnosis is: What the heck is going on there? You just cannot explain, because all are stored in those millions of weights in that deep neural network. So that's definitely one area. Trust is just one of many examples of why this trust, the value, and power is very important. Now of course, we talked so far about AI is real, so we should pay attention; and AI actually evolves the fundamental shape of society, power, value, and trust as well as what make us human. And then we also established that AI - we all believe, and the scale of the change is massive. It's as big, or even bigger than industrial revolution. But we know, when changes come, there’re not always positive. There's the other side of the coin, its dark side, and changes inevitably bring a lot of risk, even damage or destruction. And unless we actually take care of those pieces, AI would be more like a bull in the china shop, it runs full speed, but all the china will be – the beautiful things that we have built in society - destroyed. So I want to ask you: You talked about Facebook, and this is highly concerning for us, the investigation is still going on. And we also see basically on AI - so how much does AI, and what does AI do scare both of you? And we know also, AI brought Google, Amazon personal data breaches and the manipulation and so on. What kind of safeguards can we have? And one thing is, the data breaches just now become routine. I'm a Marriott member, I think just last week Marriott had five hundred million of their loyal customers data was breached. And so what scares you, and then what are some safeguards that you would advise, that all you think that we should have in place to really realize the benefits of the AI, not kind of sow the seeds for the destruction of AI?
Kai-Fu Lee: Well, the first one, as I mentioned in my talk, is the job displacement. Not unemployment, but people displaced from the jobs and not having the prospects of finding another. We're seeing some of the effects in the US already, in the “rust belt” and so on. But that's going to be exacerbated because more and more jobs will be displaced. And that loss of job is not going to be made up by something like universal basic income - just give them money. People have attached the meaning of their lives to their jobs. So I think, I'm very concerned that when governments look at historically low unemployment number, and therefore assume that job displacements aren't coming or can be fixed, I think are going to be in for a very rude surprise when - as we all know, when the technology works, it really tips. So unemployment numbers may be low now, but when it comes up, it may be unstoppable. So I think that's a big issue that will bring social stability issues. It will increase the rate of depression, the likelihood of suicide, increase the substance abuse. So I still think that's the big one. And that's number one.
Number two, I think it's “haves” and “have nots”, and that's both within each country, we covered that, but also between countries. While US and China, you can talk about the race all we want, but what really is the outcome is both US and China will be winners, is the other countries that have aspirations of going from developing to developed countries, countries that wanted to follow China's model of outsourcing manufacturing, taking on that work with low-cost, low-wage people and then gradually crawl out of poverty; or the India model, using its English competency to do IT and Call Center outsourcing. But both of those sets of jobs are going to be gone on the order of five to ten years, because the very fact - they're outsourceable and routine, means AI will do them. So the developing countries are also not aware they're in for a very rude surprise, and their hope may not be realizable.
Now, coming back to the things we normally talk about - privacy, security, data bias, the power of monopolies - I'm worried about all of them. If I were to pick one, I would say I'm most concerned about security, because AI security is not like PC or phone security. Because you're not just hacking, putting in software virus. Viruses can be detected; they’re running on your phone and PC. You can trace where they come from, you can find the patterns, you can remove them, (in the) worst case you throw away your phone and PC, or you basically reinstall the operating system, you lose some data. But AI could go in to steal money from the bank. All they do is switch some numbers around. Terrorists can get away from face recognition by putting things on their faces that fool the recognizer, if they actually knew how the recognizer works, how it was trained. Vehicles can become invisible by doing things to the vehicle itself or by hacking into the data, and because you're hacking into the data, the discovery is not nearly as easy as discovering code being inserted. Imagine tens of millions of numbers in the deep learning algorithm that's updating itself anyway, because it's getting more data, retraining itself. Some bad people just tweak a few numbers and all of a sudden, it might be out of control. It might make mistakes and might send money to bad people. It might disclose your private data. It might hurt your health, your safety. And also imagine, in the world of drones. What about the assassination drone? What about the autonomous vehicles that start running over people, not avoiding them? So I think the massive number of security issues is the one that concerns me most.
Chen-Yang Xu: One thing when I listened to Kai-Fu’s talk about moving from this one dimension - from optimization to creativity, to this two dimension - having the compassion and love, it actually resonated deeply with me. Because seven years ago, actually I lost my mom and then four years ago, I lost my dad. It was the first time you might - and then about at the same time, the center I was managing, the people I loved dearly, who worked with me - we went through a major reorganization. So there was a lot of pain and emotional challenge, as me and my team adjusting to this new reality. And it was around that time, because almost like three tsunami hit me, and I began to get in touch of my compassion side. I know you actually got into that side of the world, because you're a cancer survivor. So I basically, for the first time of my life, I felt great pain and more in my heart and then get kind of into touch with my emotions. And I realized, jeez, I've been living my whole life more than forty years in my brain. Good education, good job and seems as adequate. But until that moment I realized how inadequate I was. And how much bigger the space of the emotional space, of the compassion space, the love space I've been missing. And I thought I would not be living if I haven't gotten into that round, but I really love the way you draw and really give the future of what defines us as human. AI will actually force us, even force our hand moving to that round, and to continue to adapt and evolve. So, Olaf, tell us your point on this.
Olaf Groth: I think the point you're raising is a great one, and it ties back into jobs because we need to get much better at figuring out what makes us human, and what makes us good at being human. There's a wonderful play on word - I could do in German, I can't quite do in English - that says, “Just because you're human, doesn't mean you're being good at being human”. And it's a bit more elegant and interesting in German, but I won't bore you with German language. I agree with Kai-Fu Lee on the jobs front. And I think the key to that is that, we are not currently training humans to be good at being human. And we're then not pairing that with education broadly based - I think China is better than the United States, and the United States is better than most other parts of the world in doing this - but broadly based education on data science. And I'm not talking about coding. I actually don't know if coding is the future of our digital education. But a critical reflection on data science: What does it mean machine do - How is it designed? What does it spit out? How does it relate to our values by that intersection between the human side of us - love, empathy, compassion? I mean, we did crawl out of the primordial slime together millions of years ago, and we have that bond, that connection, but we were too often forced to be very mechanical. And the established education systems are in part to blame for that.
And so we need to get much better at educating people at the human side of things than just the mechanical, analytical, quantitative and very linear skills. And that goes for me as a business professor as well. So we definitely need to take that seriously and I am scared, especially I might add - I'm scared of the United States. I think there are countries that are realizing this is a big issue. Germany just released a new AI strategy that is much better than I thought it would be. And then the Nordic countries are getting pretty good about understanding the conflicts between human development, economic development. But the United States is the big laggard. Under the current administration, I don't see how we will be prepared. I literally and I don't mean to be party political, but literally, I want to shake Mr. Trump and say it's not Mexicans, its machines, right? I mean, let's be very honest about this. It's the machines that will create poverty, not immigration, right? That's a total ruins. So we need to prepare for that. And we are not, I haven't seen anything. And the United States is big, it is still an eight hundred pound gorilla in the world economy. And if this country goes down the tubes, then it'll have implications in China and in the rest of the world.
The second concern I have is, frankly, how we educate coders. For the first time now, Berkeley and MIT, almost at the same time, announced programs where they will teach AI and advanced data science in conjunction with things like anthropology, philosophy, sociology, psychology. Our coders, our software designers need to get an education in humanists, and in natural and human sciences. They need to understand what the implications are of the code that they put out there, just like the rest of us need to get data science and coding classes, right? So that's number two.
Number three that I am worried about is that, we will make - I'm a big fan of this mentality in the United States, that's why I came here. I'm an immigrant from Europe, and Europe tends to regulate before it seizes opportunity. In the United States, we have a slightly different proposition, certainly in China there's a different proposition. But what scares me this time more than anything else, is that we will get it wrong at a massive scale with massive fallout - on the commercial side, allow Facebook taking to the next power. And then we will forgo all this wonderful opportunity that there is a transform in our societies, because people will get scared. They'll stop trusting. That's what I'm scared about.
Chen-Yang Xu: Thank you. Because of time - actually I can stay here, keep asking because I have lots of questions. And one thing of course in the news, even I read the book, both Kai-Fu and Olaf, and Mark, believe it's not about AI race, right? It's really - AI is not between countries, because AI changes the society, people's jobs, and life - so big that we need to create sort of collective solutions from around the world. We can learn from each other, how to handle this. But we see in the news every day, it's AI race among all these different countries, they're fighting. We don't have time to expand on that unfortunately. And also we talked about the future of health care a little bit, government, universal basic income, but it also was about future of education, future of transportation. There was a little bit data there. So there's a lot more, it's really touching every fabric of our life, right? And we have a CEO and President, Bill Diamond from SETI Institute, he's sitting right there. And I thought how about using AI to help us speed up discovery of alien signals? When I was watching star wars years ago, I thought planet travel - there's no planet outside, at least we haven’t observed it. And look at now, we know thousands of planets all in the last twenty years. So all these can be accelerated, unfortunately we run out of time.
One thing as an engineer - we don't bring a problem or a manager unless you have a solution. And the good news is both of them offer a solution in the book. And I think it's a quite similar shared vision. I think in Kai-Fu’s book, blueprint for how human and machine can coexist - it's more something called “human - machine symbiosis”. And then in Olaf and Mark’s book, it’s more talking about the “symbio-intelligence”. So I think this is something I wish I have time to ask questions, but I do want to respect time. And I want to give a few questions to the audience to ask you guys before we move to the book signing. But maybe you can touch on those in future, also when you answer the questions. So with that, I'm opening to some questions from the audience. I do want the first question to come from the back.
Audience One: I'm referring to the map that you showed that essentially China, essentially being like a “two X”. You do mark an opportunity versus the US. One big void is India, where is India? About the same size as China; pretty well educated people; actually has democracy - some people would say it's an advantage; speak English. Why is India a void on this map?
Kai-Fu Lee: Well, first just empirically, there are no unicorns in India in AI. I think if you look at the reasons for success in Silicon Valley and China, a very core part is one generation teaches the other in a VC entrepreneur ecosystem. In Silicon Valley, we know the semiconductor people taught, the VC people taught, the software people taught, the internet people taught, the mobile people taught the AI people. And each generation is actually very related to the next, and then the wisdom builds up in the VCs and entrepreneurs who teach and talk to each other. And capital increases since it's a virtuous cycle of people getting smarter and better and wiser, and methodologies are worked out. The ecosystem grows stronger, makes money and more people invest. And the same thing has happened in China. So I would argue this is the very core of any other third country that wants to have that must have a very strong VC entrepreneur ecosystem. That's the biggest issue.
Now we could go into a lot of the other aspects. The Chinese came up with a winner-take-all methodology, and make a system uncopiable. Silicon Valley has drawn all the immigrants from all over the world, the smartest IQ, and the visionaries change the world, make a dent in the universe. I think those kind of methodologies also are important. And then you can go into issues like, many of Chinese study America and most go back, and that is a very powerful force for China. Most Indians, smart Indians come and study, and many more of them stay here. So you kept theirs - could be billionaires, startups’ / companies’ CEOs.
Olaf Groth: India is behind. We did a comparative study of different countries, not just for the book, but for a client on national strategy. India is clearly behind. It's a bit of a tragedy of the cognitive era that they shouldn’t be behind, because they have this great asset of being software based, and have great IT education. But that is exactly the problem. It's part of what's called the “developing country trap”. They have a lot of installed labor that is good at IT, and therefore they're protecting that labor. So Modi has made a mistake that, he has not invested years ago in artificial intelligence, the way the Chinese started doing it, or the way we have done it, because he's seeking to protect thirty million coders. And many of these coders aren't sophisticated neural network coders, they are run of the mill - no offense, but undifferentiated. That's number one.
Secondly, he said early on one of the autonomous cars in the country, because cab drivers are going to lose their job. So again, protecting established jobs which then prevents you from innovating to get to that next level of economic development. India is also a democracy, like you rightly pointed out. Democracies are messy. India is a very messy democracy. You think this is messy, try India. Honestly, India to me is the most diverse country on earth, both in terms of political entities, but also in terms of cultures and ethnicities. And so it is at once beautiful, because in that diversity lies innovation power potentially. But it is also incredibly chaotic.
Lastly, too many illiterate people. And that's the problem that China is not facing to the same degree, or that we're not facing that we've overcome earlier. Right now what's starting to catch on is this Indian ID card system. But you literally have to send government workers into villages where there are a lot of illiterate people that don't speak - they can't read or write, much less speak English, and have to guide them through sign-up for the ID card system. It's an incredibly difficult, tedious process. And we're not talking about a couple of tens of millions. We're talking about hundreds of millions of people, right? So that's the difference.
Chen-Yang Xu: Thank you. So let's do two more questions.
Audience Two: Thank you very much. I enjoyed all three of your comments. Dr. Lee, I wonder if you could just make a few extra comments about the parallel universe you described. Is that compatible forever? Are the geopolitical repercussions, is that sort of thing? Would you mind expanding on that just a bit?
Kai-Fu Lee: Sure. What I really mean is that the users’ habits have changed, so there's not a one-for-one anymore. In a copycat world, you have a search engine, we have a search engine; you have an e-commerce, we have an e-commerce. They largely look alike. Therefore they're more interchangeable. The fabric of the Chinese software ecosystem is now different. The WeChat as a social network is actually extremely different from Facebook. The way, in which you promote, the power of the company, the types of capabilities there are. So essentially, you don't have two cylinders that can be cut, and says, this is one of those, this is one of those. So, the challenge for any American company to go to China would mean, you have to learn that completely different way to build your app - your compatibility, your interoperability with other apps, the assumption of how users are, how the users might think in terms of when I want to buy something, I do this; when I want to use a service, I do that. It's all changed. So it's consumer behavior, it’s brands, it’s different classes of products. So the interoperability is very low. So I think the challenge this brings about is that, crossing over would be extremely difficult. Whether there are geopolitical implications, possibly.
So in terms of the future, one thing often people criticize about the Chinese companies is, well, it's just in China, and US is global. But I think the Chinese companies are actually going quickly global. Just that you don't see it because when the Chinese companies go global, they don't come to America, nor do they come to developed countries. So you don't see that.
And one of the interesting power of china is that, it's not all the same. There are actually many tiers of Chinese users. There're the sophisticated users, there're the countryside young users, there're the countryside older users and their habits actually are different. So these new apps that are being creative, that are innovative are often targeting non- so called “mainstream American” type of users, of which there are maybe three hundred million in China, but there are also six hundred million different users. And the Chinese new innovative apps often target them first, and then go mainstream. And then a side effect of that is that, these new users that are non-mainstream in America are actually demographically very similar to Southeast Asia, to African users over time. So my prediction is, the Chinese ecosystems will gain increasing acceptance in Southeast Asia and Africa for sure. India, possibly; South America, possibly; Middle East, probably. So this will end up with the American ecosystem really just becomes remaining strong in the developed world. So this means the Silicon Valley old mentality in the Wintel days that the whole world revolves around Silicon Valley innovation will bifurcate into a world, where maybe half of it is US primary, half of it might be Chinese primary.
There are a lot of details that I can't go into right now. So I do think that requires a different thinking, in terms of Silicon Valley entrepreneurs and the view of the global technological dominance. I guess the final interpretation, maybe what Eric Schmidt has said, which is there will be two Internets with different regulations, and data protection and nationalism. I think that assumption is still speculative right now. I don't see the evidence that will happen, but I guess we also can't rule it out.
Chen-Yang Xu: So maybe both of you want to give a quick comment.
Olaf Groth: I'll try to be really quick. Because I know we have to close out. In fact, the rise of the two models is what we hope to write the second book on. Because that is exactly I think, where the world is going and it is at once tremendously powerful in terms of economic development. And it is also very scary, because where this could go is essentially two political models competing with each other. I, like Kai-Fu, don't like the term “race” because it's very zero sum. But the way this could play out due economically, due geopolitically is that China's model replicates very well, or can be adapted very well throughout, let's say its “belt and road initiative” of sixty five countries, forty six percent of global GDP, that's massive, right? And you probably know the “belt and road initiative” and then countries like Venezuela and elsewhere. And the American model in adaptation - because we obviously have seen abuses recently - will work better for democracies. And then you have two evolving camps, and I think that's a tragedy. I personally have “chat envy”. When I go to China, and I go frequently to teach and research. What WeChat does is amazing. I mean, literally I'm envious. I'd love more of that. That isn't going to happen if we have this competition of two models happening, one along the non-democratic and the other one along the one that has a lot of democratic sensibilities, right? I wish we had more Chinese applications here, but that's not currently the way it's playing out. And that's really unfortunate.
Chen-Yang Xu: Ok. Thank you. Despite the time I wish I can - there are more questions, but we're running out of time. So I do want to close the panel and before we really close, I want you each have sixty seconds, give the audience a key takeaway they want from the book, or yourselves about AI of the future. And then we'll close for real.
Kai-Fu Lee: Sure. I think we heard a lot about the opportunities from AI and the dangers from AI. Olaf and I are both taking optimistic view, and obviously we're also very cognizant of the dangers. And I think my closing remark is that, I fully recognize dangers and I think we have to choose to be optimistic, because the view in which we view the world will likely become a self-fulfilling prophecy. If we all tell ourselves, AI's going to take our jobs away, humans have no future which just hide in our homes and forget about getting a job, then humanity will have no future. But if we become optimistic and seek ways in which to overcome the problems of security, job displacement, think about the value of compassion, think about the opportunities for humanity, and think about the opportunities AI could bring about, and think about all the dividend from AI and how could be used, then that belief will too become a self-fulfilling prophecy. So it's not that I don't see the dangers, but I think as thought leaders, we really have to take the high road, because if more people believe in the optimistic future, then that is the outcome that we will have.
Chen-Yang Xu: Thank you Kai-Fu. Olaf, what's your key take-away for the audience?
Olaf Groth: Yeah, I would say: Definitely focus on the symbiosis between you and the machine, and ask yourself where is your human strength? And if you're a leader in industry, in a company, think about what this means for how you lead, how you develop people, what kinds of skills do you teach them that are innately human that we need to sharpen and get much better at, rather than making humans objective viable machines. That is not what we're good at; we will never succeed in that. And AI can be used either way. So make that choice, understand that symbiosis really well.
Second recommendation is: Get involved, meaning look at AI as a tremendous and huge opportunity with a positive lens in front of your eyes. At the same time, don't be blind, don't be naive. Look at the interests behind AI, avoid the objectification, and get involved on safeguarding it. Make it safe so we can get to the new horizon. In the book we're talking about we need to have a multi-stakeholder congress across borders. It's not an AI scientists led, government led, corporate led, all of these need to be at the table. But we need to sort of not go bottom up or top down, but middle up and down, meaning we need to work with foundations, work with academia and pulling all these other interests, and make that a global multi-stakeholder congress, including China.
Many people in the west say, “Oh, we need to get our act together as the west first before we take on a dialogue with China.” And I disagree. Because I think then we are in this world of the two worlds, and that well be really hard to stomach. So at the end of that needs to be what we have dubbed a digital Magna Carta that respects different value systems and safeguards AI, right? But now we need to do these too and intend opportunity and governance. You can't have one without the other.
Chen-Yang Xu: Bravo. With that I cannot have a better stopping point, and I really appreciate all of you coming this morning, make the time worthwhile. And I do hope all of you can take something away, that will go back to your work and life and view the future of your work and life differently. And then prepare yourself, and also people you can influence and you lead, into this era of AI, and then do it right our own script as both of you have said. So with that, I really appreciate, and of course hopefully it's peaked your interest for the books, because we do have the - it's not the end of the event, and so there's a book signing, this photo opportunity, and so you can get the book, read it. And so in any case, let's give a round applause to the panelists.