Kai-Fu Lee | AI Era - Leadership and Technology

[Music] good afternoon good afternoon and welcome my name is Kostas Panos I'm a professor of Electrical Engineering Computer Sciences Sciences and I'm also the director of the Center for information technologies in the interest of society I'm here to introduce our speakers and I have to tell you the obvious we live in interesting times you heard the terms of the AI revolution the rise of a new AI superpower the start of a trade war with that superpower the Internet of Things the value of data the threats to privacy the fourth Industrial Revolution so these are all issues that we hear about every day and these are all issues that are defining our era so we live in interesting times and we are not spectators in these times here at Berkeley we certainly make a difference for example in our Institute's and embed learning in general we do multidisciplinary work on the future of work on the face of automation we presented to the world a cute a towel folding robot that showcased the deep reinforcement learning we introduce to the world robots that have the dexterity to pick up pieces almost as fast as humans at random we are talking about robots that cooperate with people we're talking about revolutionary agricultural technology it's aims in the overall health system and of course we're talking about the pioneering work of people trying to put a stop to the idea of weaponizing AI professor Russell is here and he represents that movement and finally there is of course this ultimate argument between the singularity and the multiplicity the idea that AI is going to replace and undermine the human experience or maybe the hope that the AI is going to enhance that experience something that we truly believe in and something that led to this whole notion of not artificial but inclusive intelligence here at Berkeley so interesting times when you have a huge role to play and today's speaker is the best person to shed some light on these issues so today welcome doctor I fully to UC Berkeley who will talk to us about US and China superpowers in an era of artificial intelligence and he certainly has the background to talk about the Dalit he's determining CEO of this innovation vector ventures he's the president of innovation ventures artificial intelligent Institute where more than 200 AI engineers are working side by side with investors for new ideas he's managing two billion dollars in investment fund primarily a finding next generation Chinese high tech companies prior to findings innovation ventures in 2009 dr. Li was vice president of Google and president of Google China previously he called executive positions at Microsoft SDI and Apple he received his Bachelor degree from computer science from Columbia University and his PhD from my own alma mater Carnegie Mellon University in computer science he has several honorary degrees one of them from the City University of Hong Kong and another mechanic Mellon University and he has authored many patents many pet many papers and articles and seven top-selling books in Chinese and has over fifty million followers in social media as luck would have it just yesterday his new book HOSA was released was published just yesterday 9:25 the title is AI superpowers China Silicon Valley and the New World Order and certainly that's an exciting subject so we welcome dr. Lee this is a right place I think for him to visit when he released a new book and this is a very timely occurrence for his visit so I would like to acknowledge that this advice is co-hosted by the husband's speaker series and the seed research research exchange I will bring together these long-standing seminars and communities to examine the future of leadership and technology in a rapidly changing world and this invent by the way was solved within hours we have more than two hundred people on the waiting list and we have people watching not only here but also on and if simultaneous presentation at our headquarters in Citrus across campus in our in our auditorium and of course the broader public over the Internet so welcome I would like to welcome not only on the stage and I would like to welcome and I would like to welcome the in tiesm yeah okay you cannot join us later okay so okay Mike all right wonderful thank you so happy is microphone working you don't hear me okay it's great to have opportunity to come to Berkeley and talk about açai the era of AI and you know probably I'm not popular enough on this campus I want to bring someone who's really popular to tell you at this campus to tell you about AI it's a great thing to build a better world with artificial intelligence and listen to this guy be a TA so so that was a AI synthesized version of President Trump speaking English and Chinese should give you an idea how far advanced this technology is and in my book I talk about four waves of AI it is the singular technology of deep learning and related technologies that is going to take us very very far starting from first in Internet AI basically AI requires a huge amount of data in a single domain and performs extremely well exceeding human performance so the Internet has the most data every day we are contributing data we're tagging we're helping Amazon Google Alibaba to become smarter in their applications and all the big AI Giants today are Internet companies precisely because they were the ones who benefitted and created the wealth from the data that we contributed but we're also beginning the business AI that is using existing business data to make the business process work better so in banking maybe do better targeting of customers better loan approvals better job in determining credit card fraud and better asset allocation so the data is already there and just make use of it could happen in insurance hospitals so on and so forth so all businesses can benefit from it the third wave is adding eyes and ears the perception the computer vision speech recognition digitizing the physical world and using it to create applications that didn't or couldn't exist before for example Amazon echo and Amazon go the autonomous stores are great examples of that you might think Amazon goes way too expensive ten million dollars cost to build it but actually in China we are already building simpler versions of autonomous checkout autonomous stores at thousands of dollars per store so this is happening all around us and of course speech assistance that's gonna change the way we interface and then the way for if wave 3 was adding eyes and ears wave force adding hands and arms and legs and makes machines that are able to walk so robots in manufacturing robots for picking fruits robots for washing dishes these are all happening these are all our investments and of course autonomous vehicle that's the biggest thing that will make huge changes to the way we transport ourselves and also the future of logistics and delivery so each of these four waves III argue could be 10% increments to our GDP over time and each one of these four could also challenge and displace 10% of our workforce and and they're going to happen at the same time now my book AI superpowers argues China's rising but the first thought you have is look at all these brilliant AI scientists they're all American and Canadian how can China possibly say it has a chance not to mention Silicon Valley in which we are today dominated the world the world revolved around a Silicon Valley hedge hegemony right everything from wind tell when I started working first for Apple in 1990 the world was dominated by wind towel there was not a single foreign player so the Silicon Valley frightfully feels entitled that it owns the world but that has changed and the magical thing that happened ten years ago was that China is such a huge market a market that is four four times larger than the US that drew a lot of money and educated a lot of BC's who became smarter over time like myself and we funded entrepreneurs who were incredibly tenacious and built world-class companies like Tencent Alibaba and they build great products that attracted more people and that brought in more money and they build more products and more users came about so that has created really a parallel universe one in which the Chinese companies in the merely eight years went from copycats to starting to leapfrog to pure innovations made in China back eight years ago VC's generally wouldn't fund the Chinese company unless they could say which American company they copied you know because VC's VC's weren't that smart back then and basically if there was already a proven concept in Silicon Valley then it's valid to try it in China but keep in mind China is was a very small market when Google started us the Internet penetration was 30% China was 0.2% so what could the entrepreneurs do but to rely on the smart American teachers and entrepreneurs to teach them but in the last five years that these innovations actually got better today we chat is a better product than whatsapp Weibo is a better product than Twitter maybe not in the completeness of content but in the quality of the product okay and and now there are Chinese innovations I cannot I mean I only have 10 or 15 minutes I can't even explain these companies to you because each one is totally different it's all on my phone it's not on your phone and I use it most of them daily and they're great products and they're not American inspired they're completely Chinese inspired so China has now reached the space where it just pure early Chinese innovation captured on this slide value anywhere from 2 billion to 75 billion and I and when we didn't even include and financial phenomenal amount of value has been created in China on Chinese innovations so this eight years magical change from copycats copied to China into innovators copied from China that is the core basis of why China can become a co-leader with the US in artificial intelligence in building these great companies Chinese entrepreneurs have become incredibly tenacious iterative product iterating on products very fast speed and able to generate data use data and use the data to train AI in China entrepreneurs considered 996 to be a very basic level of working hard and 996 means not working I am to 9:00 p.m. six days a week one company said we have great work-life balance come work for us where 996 so it gives you an idea and also huge capital is again injected into China especially in artificial intelligence China AI funding has exceeded us Chinese stocks on equivalent companies has jumped up based on its growth revenue but also based on expectations and its unique capabilities in China just sign ovation ventures alone not counting the rest of the industry our AI investments have we already have five unicorns I'm not even sure if there are five unicorns in AI in Silicon Valley and these companies are worth 21 billion dollars with one of them about to go public and that's just our investment totally there are probably about 15 unicorns in China this is just in AI not pure AI companies note here that I didn't count you know other companies that used AI these are core pure 100% AI companies so what has happened okay I've told you about the unique parts about Chinese capital entrepreneurship and innovation what is why is a China able to leap ahead with such a deficit in talent it is because we are moving in AI from an era that was research driven expert driven to an era that is application driven and data driven most technologies used in most commercial applications of AI are known technologies like deep learning like versions of reinforcement learning and transfer learning that have become more mature so the AI scientist the worldwide are very generous to share their papers findings in real time and even the open source their code so it is now about who can run the fastest implement the best version iterate the fastest and also have operational excellence that is the country that will develop the greatest value out of AI and also theta is all about data the more data you have the better your AI China has three times as many mobile users as us but that's not the big deal ten times the food delivery 50 times the mobile payment and 300 times shared bicycle rights all of them are generating data you might be questioning shared bicycle rides is that AI yes it is because the the 50 50 million rides a day are transmitting data with sensors on the bikes so that that is contributing and of course the Chinese policy is very Pro Tech Pro AI ever since the last year 2017 has become a national priority so infrastructure is being built in China there's a new city the size of Chicago being built for autonomous driving there is a new highway to be enhanced with sensors to help the safety of autonomous driving but in the u.s. I believe autonomous truck testing on highways are not yet permitted because of concerns by the trucker's Union regarding the trucker's future jobs but I think if that's the way that policies go forward while US has a lead in autonomous driving that will reverse if one country pushes and one country polls so today if we look at which countries ahead I think China is slightly behind us still it's only been the last two or three years where China's had a Sputnik moment and decided to go after AI whereas the US has been building for years but China's going very fast and if we project five years later I think China will generally lead in AI but except for business AI and business AI requires structured data storage and data warehousing which Chinese enterprises currently lack so that's not likely to be a Chinese lead lead but the other three areas China will probably do very well so uniquely we have two engines driving the future of AI it's not in the past it's been one engine so I believe AI will go faster because now there are two superpowers driving it not one compared to any of the past tech revolution but rather than talking about AI as a competition between two countries and even in some cases a cold war it really isn't like that AI is really much less like a nuclear weapon than it is like electricity that can enable a lot of things and generate a lot of value PwC estimates that AI will add about 18 trillion dollars to the world GDP by the year two to 2030 and if you if it's hard to visualize how much money that is that's basically a China plus in India and that's how much value it will add but facing a lot of opportunities the AI scientists are really more sharing so I really in my book don't do not want to portray this as a zero-sum game and all this value will be generated and shared however AI will introduce a lot of issues such as privacy security bias from data dominance of the AR Giants increasing wealth and equality and one I want to spend two minutes on is about job displacement job displacement is happening it is happening one on one as robotic chefs are flipping burgers and as pastry check outs cashiers are being this is our investment this machine costs $800 one-time cost and it will checkouts you with the pastries you buy it's a computer vision based and this is another company we funded that is called f5 future store you can order a bowl of top Chinese noodles with your face recognition and WeChat pay with no human completely robotic at $1 at 60 so who's going to eat at McDonald's so forget about those robot burger flippers you know and also white-collar jobs equally challenge C's City has announced 10,000 jobs to be displaced out of 20,000 in operational staff and Facebook and the Chinese company Tokyo are using robot editors not that traditional media would ever use robotic editors but think about it if if most of our eyeballs are on robotically edited news whether it's Facebook newsfeed or total there will be fewer eyeballs on traditional media and therefore editors jobs will be displaced so I am one of the more aggressive predictors of job displacement as a result so what jobs are left for humans and this is a graph I'm trying to summarize from my book if we think about the two things AI cannot do one is about creativity AI currently is trained with a humanly defined objective function and then AI learns AI does not create its own objective function and certainly AI has no ability to invent something whether it's a great story or a great next drug breakthrough so that's one aspect but there is another interesting aspect which is about compassion we humans want many jobs to deal with humans not with robots for example nurses elderly care teachers doctors tour guides concierge bartender masseuse the list goes on so those jobs are going to not be able to be done by AI but some of it might be some combination for example the future doctor might be one where it's largely a I diagnostic but the doctor is there to provide the communication comforting increasing the patient's confidence of survival so if we think about these four quadrants I think the lower left will eventually be totally replaced by AI and that's jobs that have low human touch and low creativity the jobs that are high creativity scientists ask writers of new books and movies they will use AI tool to magnify human creativity so it is a symbiosis of human and AI and then the jobs that are locally a low creativity but high compassion say the future of a doctor's job it will be AI doing the core analytical function with the doctor providing the human warmth for both of which combined to help heal more patients and also dramatically reduce the cost of healthcare and of course on the upper right side it will be where our compassion and creativity continue to shine with the help of AI so it is not a simple question that the jobs being displaced or not it really depends on what type of jobs obviously this is a simplification and in my book I talk also about physical kinds of jobs and also strategic issues and other things that AI cannot do but I think at the top level these are four ways in which human and AI can coexist as a blueprint so I think the whole world really needs to work to work to work together I think job displacement is serious the lower lower left-hand quadrant is large but if we work together I think we can get over the next 15 to 25 years as the lower left-hand quadrant is displaced I would envision just as a hundred years ago we saw a massive shift from agriculture to manufacturing jobs we're going to see the next 20 years a massive shift from routine to compassion the jobs and that will be whether or not we make that transition will determine whether we have a great future together but this is a story of 20 years if we think a little bit further at 50 years where most of you are my age and and I think went by that time when we look back we will really see a is serendipity because it came and took away the routine jobs but it really gave us back the time to do the things that we love and think about what it is that makes us human and allows us to follow our hearts it allows us to have more free time allows us to work less and maybe not work and it allows us to drop this workaholism that has shaped us in the Industrial Age the other thing I think we realize is there's a lot of fear about AI I think whether we fear AI or embrace AI it will become a self-fulfilling prophecy as much as there are issues with AI we have to embrace and believe that we have free will that the ending of the story of humans in AI will be one that we we right thank you [Applause] so I'm gonna ask a couple my name's Laura Tyson I'm the interim dean at the Haas School I am delighted to be able to bring this to you as a Dean speaker series I'm gonna start with a couple of questions this is a full house we have limited amount of time I want to make plenty of time for you all to ask questions - are we gonna have microphones how we going to do the questions okay great the microphones are in each side okay so let me start right where you left off before when we were talking before you emphasize the b2b nature of a lot of the AI applications and the fact that the major thing that's happening right now is that in companies are looking for cost efficiencies and the cost efficiencies are eliminating people you're eliminating people they're they're in that quadrant down there yeah most human beings do routine jobs mm-hmm okay yes and so are you concerned about the pace of all of this at society well we don't have any idea how to create more more demand for caring jobs societies don't even know how to do that so are you worried about this transition process the speed at which attach yes I it definitely worries me I think the first step is hopefully there's a critical mass of people who will come to agree with this argument and then just start to work on it so the first step is to publish the book and see what kind of response I get and and I think it is solvable because we can imagine the world where there are ten times more doctors but their job is not like doctors today they don't require 10 years of training they might be more like nurse practitioners but there'll be 10 times more of them because we want the human health care to reach all levels of income right there's so many poor regions and countries that don't have proper health care so if we think broadly there can be 10 times more doctors and there can be 10 times more teachers because a lot of the teacher's job that's repetitive you know giving the lecture grading the homework creating the exam can be done by eye and we're funding companies to do that actually we have a example most Chinese homeworks are not being graded by the teachers but by AI OCR engines and also and actually they can do all kinds of grading including the fill in the blank and even essays can be graded so this is a power of AI and of course the teachers just needs to watch over now what does a teacher do what a teacher becomes life time time mentors right can one on one understand and help each child or student find his or her passion and answer questions so there can be 10 times more teachers but I think you're right that these ten times one doctor attends and Reuters won't happen by AI there are many optimists who say that AI will create so many jobs it will just be okay but hey I isn't going to create those doctorate teachers jobs we're the only ones who can do that right we're the only ones I could do that and if you look around the world and think about how we fund those kinds of jobs education health care care for the elderly they tend to be their social missions and then we fund them socially we don't fund them through b2b models in general we would just don't speak problem I am a VC we wouldn't be able to consider someone who comes and says I want to start the company that hires 10,000 elderly taker care takers a year will grow to a hundred thousand there's no doubt that is their social value there's no doubt there are many of these jobs available today not being filled because people feel that they're paid too little so some will somehow we need to get over that is it to subsidize those jobs is it to subsidize the founding of those companies is it changing our social contract so that our pay is not merely a reflection of the economic value but I saw some combination of the economic and social value these are really big issues and they're issues that you you do talk about in the book so I just I want to change gear a little bit because that's my particular passion but it is in the book and so let's talk a little bit about the us/china because it's so in the news right now is there something in this vision that you see going forward that could be completely changed by massive step backs from globalization by much more protection in the US by Europe essentially adopting totally different privacy and data standards and saying we're having a totally different view of data sorry you can't have it do you how do you see the the geopolitical environment being affected by this or affecting it hmm GDP our ICS a very good thing okay I don't actually like many parts of GDP are but I think a government putting forth standards and putting its values to be reflected by his laws I think it's a good thing you made me want to tell people what GDP are is I show everybody in this room knows but just in case sure it's a European set of European standards that require a consent of users on how their information is used in shares and hopefully eventually giving back user the control of their data and of course US and China don't yet have such laws and I don't think there will be one global set that will work because countries are different but I think countries should put a step forward otherwise I think there would be we could run out of control I on the other issue US China basically a trade situation I think is going to be a lose-lose economic level because the two economies are incredibly intertwined in terms of you know if you take out the phone and look at all the hardware components where there's the IP there's semiconductor the memory the chips it's going to be you know American parts sent to China manufacturers so back to America so who pays tax to whom this is I think it's a I think it's just spending a lot of time on something that will be a lose-lose but there's very little we can do now for us fortunately AI and mobile internet China and us have evolved into parallel universes okay so the software aspect are not really regulatable so the Chinese user with the apps and underpinning of what they use in in the mobile or AI is completely separate stack from the US so there are minimal dependencies and and I think we're one of the fortunate few industries that won't be hurt by this trade war right but you do don't you did talk also about it and it's in the book too about how these technologies can disrupt all of the other industries so you may end up that the way the conflict the trade conflict may play out is in other industries which themselves are changed by AI but yes I can see the verticals here you can't really figure out how to yes do a trade war there can I ask you mentioned that the competition here is not a cold war so think about it from the point of view of the US you gave the u.s. the sort of long-run strength in research so much of basic research in the United States links back to DoD and DARPA if that is just the real situation we have funded the basic breakthroughs in knowledge that fund all of this stuff through DoD so do do you think that there I I didn't worry about the danger that some however the technology can be used can be supported and used actively for geopolitical military purposes by two Defense Department's who are who want to build their power yes I I think there are sort of two branches in both US and China AI technologies in the u.s. are kind of code driven by universities funded by DoD NSF yes yeah and then the industries the powerful Google Microsoft Facebook right they have very powerful branches there in China actually universities are quite weak in and they're generally not funded by the Defense Department okay orange board government and well certainly the the Chinese NSF and I think the government in the military do their own research in both countries both countries as well it as well which we we have no idea what they're doing yeah hopefully that brain trust is currently with the industry and academia but in China the industry is expanded so fast as you can imagine in this age of implementation the Chinese industrial giants are really good at AI Chinese universities haven't yet caught up so it's not quite symmetrical I my PhD work was funded by DARPA I I'm personally incredibly impressed by the visionary leaders at DARPA and NSF who are the ones who set out this speech recognition challenge machine translation autonomous vehicle and at the time they didn't seem very military all right well that's one other question then I would urge you to if you have a question to kind of line up at the microphones my question is thinking about the rest of the world here because right now you would say that if you look at world adoption there's much more of world adoption of the u.s. AI driven or AI related companies like Google or Facebook they have global presence whereas a lot of the companies you mentioned up here which are very big but they're basically in China for China so does that what why is there that difference in the world and do you see the Chinese companies going out into the world to basically compete with US companies and other companies in the rest of the world ok I see Chinese companies going out but I don't think they'll compete for that quite a long time with the US for the following reasons well first why didn't Chaney's companies go abroad earlier well first they've only gotten good the last three or four years very short history secondly China such a large market that's the opportunity cost to go to another country just doesn't seem worth it why should Tencent hire hundred people for an Indonesian product when that hundred people could build another Chinese product which is a ready market there already familiar with that's been the reason in the past but at this point the Chinese inner market is not growing as fast as before so these giants have become they're worth 500 billion dollars and they're obviously thinking about global expansion so 10 cent Alibaba bytes dance and didi and others are have very strong global expansion plans they're also startups in China that are starting to do copy from China work so one of those bubbles that are Chinese innovation they're a star that might take one of such Chinese innovation and then copy from China and have it available in say Brazil or Indonesia or India and those companies are reasonably successful and if you think another really interesting thing that I only recently came up with mafia in the book is that China is not one demographics actually China can be roughly cut into three demographics one is kind of with quality of education thinking more like us the second maybe closer to Southeast Asia the third maybe closer to Africa more advanced countries in Africa so now you're seeing these Chinese products coming out they're not all for the first-tier some many of them are for the second and third tier like pin dodo is really a third tier I mean and and tow-in are more for maybe second here second plus third tier so China can be very powerful in using its unique knowledge of the second third tier right to go to Southeast Asia in Africa so my projection is with the Giants and the startups and the know-how of the copy from China China will make inroads to Southeast Asia Africa and India and Islamic countries so it actually coincidentally matches with the bellroad initiative and and I think there's so much so many people and such so such large markets this will keep China very busy the Chinese entrepreneurs overview Europe and english-speaking countries very high barrier of entry what's it going to take for an American to switch from you know Amazon to JD that's like impossible the same is true with the reverse so I think the world will actually become two parallel universes in the Chinese German Chinese mainly Chinese Internet and mainly American Internet not in regulatory or governance sense but just in what apps you use kind of sense okay look I'll copy asking questions that I would encourage so I know you've talked about up there a multiplicity versus singularity and here I'm gonna reveal that that is those kinds of terms are discussed a lot you say you're not well why don't you talk about your of you on multiplicity versus singularity okay well singularity is the idea that things are moving so fast that machines will be so smart that there will be existence your risk and and in particular with super intelligence right I'm not a believer in super intelligence I can see possible extension existential risks but I cannot see super intelligence basically extrapolation of the current algorithms don't I don't see any roadmap towards that I think we do see for those of you who are not into AI you probably see more and more headlines among the eye being able to do cancer and automat autonomous driving and go and it's all those great things you think singularity must be coming but that's actually deceptive because it is applications that are growing exponentially they're all still based on a single breakthrough so that actually validates my point that we're in the phase of implementation is one set of technologies that have matured and we these seasoned entrepreneurs are going to really milk that technology for all it's worth and get it implemented but those things that you read you know are not breakthroughs in technology and if we look back on the 62 years of AI history that I would say there's only one gigantic breakthrough which is deep learning and and that was 10 years ago that was 10 years and we have yet to see another in the last nine years their small break but unless we see more breakthroughs I would be hesitant to project any date for super intelligence and singularity of course when breakthroughs come you know I reserve the right to change my mind I think you mentioned Sputnik moment in your talk and I believe in your book the Sputnik moment in China was was go was the game yeah it was it's actually quite unbelievable that China two years ago a I wasn't even understood by a lot of people including the government and it was the moment that alphago defeated léa Seydoux the Korean super player that really woke China up I know most of America didn't watch the match but a hundred million people in China did and the shock was that hey we Chinese invented this game go and you US UK company and and all the computer scientists say computers can't beat humans it shows human wisdom thoughtfulness zen and all those things and no computer brute force can possibly solve it but then all of a sudden you guys did there must be something magical so that's what pushed the Chinese government VCS entrepreneurs to do AI of course we've been in we've been quietly investing AI for some time but but it was with our proprietary knowledge that's how we have the unicorns and but other VCS were not but once the goal became a household thing then everybody trying to that's just an amazing story because here it was when I when I was a child it really was the Sputnik and that was the moment when you actually it was a totally different world but yes yes question Oh could you please introduce yourself and are you from you know where you're from in terms of a university you don't have to tell me here the University fee well thank you dr. Lee for the presentation my name is Quinn I'm a first year ha student at Berkeley so my question is that you touched a little bit on your point on the singularity is that as a venture capitalist you mentioned deep learning has been technology has you seen a brick through 10 years ago now I see a lot of implementation different scenarios as a venture capitalist what do you what for for the next breakthrough out what's wrong your mind that you you think bring out a new game changing technology that could you see a new instrumentation in different scenarios the first question the second question is I'm a entrepreneur working on a deep learning enable just a recognition enabled education technology company really improving the user experience for the kids and also overall teaching efficiency for the teachers love to have an opportunity to see if that's possible sure you know how to reach me I gave you the way to reach me on the next breakthrough see as a venture capitalists we see so much opportunity in deep learning so we are funding new applications for us we we actually as venture capitalists we each company we fund there is already inherent market risk competition risk talent risk economy risk and that's a lot of risks and we don't really want to take a research risk so we fundamentally don't invest about technologies in that are not at least to some extent proven at least from one domain then maybe for not to apply to another domain but but they're speculate what research we would like to see I I think I'll be interested in I mean look at the weaknesses of deep learning right it requires huge amounts of data so who might invent something that requires less data maybe advances in transfer learning maybe advances in the unsupervised learning those might be advances would like to see the one thing that we really think where we will see is the platform ization of AI that is let any non engineer be able to program AI that is the moment this will truly probably right and you see that you know when you program the iOS you don't have to learn how it's based recognition the fingerprint recognition works right it's all encapsulated in their API and modules so you just need to be a good programmer so when AI reaches that stage I think it's going to be a huge moment and we are watching that carefully as far as true research breakthroughs I'm no longer in research so so that's probably all I can cover yes hi dr. Lee thank you so much for being here my name is Joyce and I'm a first year MBA at the Haas School Business also my question for you is that you'd mentioned that creativity is one of the few things that humans will be able to supply you also mentioned that in China the education system is facilitated by robots like papers are graded by robots and my first instinct for that is that it could potentially really quash creativity so how do you reconcile the potential for AI to actually reduce out-of-the-box thinking yeah I think the fact that pay papers are graded by robots are not completely contrary to enhancing creative thinking because I can see what you mean you know your robotic Lea Grier grading things that are gradable by robots but you're also freeing the teacher with more time to teach creativity so I think it's a double-edged sword potentially I think the bigger problem really is that the the Asian learning education in general is too much rote learning that is the fundamental problem that stifles creativity I think the three C's that are really important our curiosity critical thinking and creativity and those happen are lacking in the Asian societies and education and I think that is an advantage of the American education and so I see their what creativity a I'm not being able to do creativity is an advantage for American society at large yes hello dr. Lee my name is Amy and the first student double majoring in eats and business and you talked a lot about how Chinese tech firms have been growing and developing a lot of new technology I think something interesting about how the tech industry in China has grown has also been the rapid rate of adoption widespread adoption by the Chinese people like for example with Ali pay and WeChat smade China largely cashless society whereas in America like venmo hasn't had the same traction so I was wondering what are your opinions on the differences and I guess attitudes towards technology and how Americans can better adopt technology faster yeah okay that's a really great question yeah I think there are many reasons I think Chinese people like to try new things and are and and I think they've been rewarded for it right so actually the adoption of mobile phone was very early and then the China and then every time new things came about Chinese people tried then got benefit from it so I think it's kind of Pavlovian the fact that all the new things seem good so I'll keep using new things that's coming to be one one one aspect but probably a bigger issue a bigger aspect is that China is coming from behind and sometimes there is a late mover advantage because if you adopt something that's great for the 90s like credit cards that blocks you from not using it the fact that China didn't have a lot see China was mostly cashed with a little bit of credit cards so jumping from cash to mobile was much easier because credit cards kind of gave you 50% of the convenience so if you already have it you may be reluctant to switch plus I think many of you may feel you have you know more trust in Visa MasterCard then let's say Facebook have maybe because it has an it's an entrenched element I mean these I mess with our brainwash you with all their ads and brands right we security I guess I would guess if we did a survey most of you would rather use Visa MasterCard and Facebook even though I think Facebook can implement easily a more secure system in mobile payment so I think it's about entrenched market similarly you know Chawan when US had all the landline phones and china just didn't have a lot of phones Dennett leapfrogged with mobile and I think that will continue to give more advantages to China because u.s. went through multiple steps of change take another example retail write us started from these mom-and-pop shops in the small town downtown then the Walmart's and Costco's came and then the Amazon came so there's a very gradual transition in China before there were Walmart's from Costco equivalents ecommerce took over so it kind of stifled all the retail so retailing in China is quite behind in all kinds of ways in their IT systems and everything but that means it's ready for disruption again so we're very actively in investing in retail and that includes using IT AI to track it better using it to leapfrog further and actually it's kind of good that China didn't copy the US super model concept which as you know is kind of vacant China doesn't have them so there's no baggage so now the real estate companies in China are thinking about how do we construct a mall of the future and we're actually advising them right we're telling them look we think the future mall should be the following whatever you can buy on e-commerce just do ecommerce however physical and online should be connected so you should have sensors in your stores so you kind of know what your users are doing and buying that may be a little scary here but in China it's acceptable except for one in the UK yeah cameras already anywhere so you have online offline fully connected in China we advise them that in the future most services were bifurcate into fully autonomous mid to low end products such as the future fast food will be autonomous because we're not in fast food stores to enjoy the wonderful service and the gourmet food and the convenience stores will be autonomous but there will be luxury stores that will be fancy beautiful big there will be Michelin restaurants so the future of the Chinese mall will be this combination of online plus offline very autonomous low to mid end and then very high-end luxury that's still human service so that allows with China to potentially jump further link frog again in retail and hospital would be another one that one is a lot tougher but China's going through a hospital reform and yes this is another area we can leapfrog I mean the China hospital experience is horrible right now I mean you have to line up get a number wait three hours and doctor sees you for three minutes and then you're out so something of that here well here are the differences you make you have to make an appointment a month early but the doctor will give you all the time you want but but time for leapfrog anyway right time for the car it's time for the front yes over there yeah hi thank you I'm Brian I'm a full-time MBA at Haas and I just wanted to touch on you mentioned bias in AI and I'm wondering how you factor bias and AI into your investment thesis maybe not for pastry recognition but for other applications how do you think about that right right no I think social responsibility is a part of everything we do so we don't invest in AI for weapons we don't we don't invest in in AI that's um that's predatory to to users and customers we don't invest in you know ái that could be used for fraud and we don't we would not invest in the company that that's it that obviously enhances prejudice based on data bias but the way I would think about it is sometimes non-technical people might think of bias in different ways so let me clarify what bias means from a technical standpoint first a lot of people think hey I can be biased because the programmer may have bias that's actually largely untrue because AI basically works from data yes programmers can't have bias is possible but it's not as much as you think okay so theta can have bias but where does the bias come from some people think you throw too many features at it then it has bias actually not because a machine learning can learn some features don't make a difference and they throw them away and actually tagging is something that's important so if you tack someone as likely to default or not default on the loan if that's not properly tagged that could cause trouble but more features does not so how do you sort of so given it's largely data-driven how do you minimize the problem of bias so suppose you trained it with too much data from one gender the other gender may suffer too much data from one race another may suffer and so on well the best way is really to ensure a proper balance of your data in a way that reflects the demographics you could further say that's what if you could further undo the effects that that past data have has accumulated yes by doing something for example right if you thought that a I'll use myself as an example so I don't insult anyone so hypothetically if you have data where let's say the Chinese are being prejudiced against and and that it forced them to do things that may cause them to have what looks like bad data that you don't want to give loans to them okay but you want to be fair to the Chinese hypothetically what what you could do is to remove race from consideration right from the from the data so you could do that but you have to be aware that a I just optimizes on the function so if you want to have it to minimize default rate and lonely more to Chinese actually increases a default rate you're intentionally making it less accurate by making it less prejudicial to perhaps a sect a sector that you've found in the past discriminate against so that's that's I think a balance you have and the last thing I'll say is that while we can all say their worries about data bias I I will I believe humans have much more bias if we do what I said above I think it will be much less bias than humans to give you an example a university in Israel did a study on judges right you would think judges are fair right that judges before I guess this is a study yeah the study found that judges before lunch were meaner to give harsher sentences and presumably because they were hungry you know those kind of phenomenon will not happen with what they are so I think we have one more okay one more question and I I think I'm gonna go over here because they've been I haven't paid attention to this side enough so here we go hi thank you very much dr. Lee I'm wearing two national students from China and now I'm a junior in hospital business with UC Berkeley so you mentioned that the AI development china-us are two parallel universe so what kind of advice would you give for future entrepreneurs who want to work between this two countries so given your own professional experience globally okay great I think cross-border will be very difficult you know there's been that interesting theme for the past ten years you would think such two large markets something cross-border would be great so given their parallel universes I wouldn't advise you to work for an American company trying to do business in China or Chinese company trying to do business in America that company anyway I wouldn't I I mean I'm giving a generic statement it's obviously exceptions but everything else being equal I would say be careful about taking those jobs because those companies will have a hard time crossing the parallel universe I would say that's what might be interesting is learn what's so great in America and then move to China and use that knowledge or the reverse for example if I could be a venture capitalists in Silicon Valley I can use I know all the tricks of these entrepreneurs what they've done and no competitor in Silicon Valley knows I think I could do something really exciting however sophia's seems to not make that like Cynthia Cesare control for on foreign direct investment in the United States which we have just made more more restrictive as there is as active policy so it's now much more difficult for a Chinese investor to get a small investment in a tech startup in the US right so it's to summarize I think given there are two parallel universes your best strategy is to fill your brain with one of the parallel universes and have things in your brain that people in the other universe don't know and then go move there and show them how powerful they are there you go that is like some of the best down-to-earth career advice I've ever so we have caillou lee has a number of other speaking engagements we are so honored that we have the opportunity to be part of the book launch and to have this great opportunity to discuss and get your insights thank you so much [Applause]

Channel: Berkeley Haas

Views: 18,964

Keywords: Kai-Fu Lee, Dean's Speaker Series, UC Berkeley, Berkeley Haas, Haas School of Business, Laura Tyson, CITRIS Research Exchange, AI

Id: 5rU4cnbJti8

Length: 58min 34sec (3514 seconds)

Published: Thu Sep 27 2018