Artificial Intelligence and the Future of the World, with Kai-Fu Lee, 10.5.18

[Applause] thank you all for being here great great privilege for me to be sharing a stage with Tyson and Andrews other people I call it sources you know so you guys are like my rabbis and I also have the best crowd of my life once with Keiko it was a chamber making a bigger vision vision so I have about a thousand questions we have a lot of time we have about an hour we're gonna talk amongst ourselves and then we're gonna open it up because I'm sure all of you have a lot of great questions as well and please redirect me people of we all let's take this where we want to go there's a lot of news that I could lead off with right now and it's been a difficult week to say the least in the us-china relationship but let me start with a more personal question because one of the things about your book that's really wonderful I found it to be both a learning experience but also a real personal journey writing a book is not easy and you're never just convention before everything you're about that how hard it is you didn't need to do this so what what prompted you what was the personal thing that really galvanized this creep sure well I'm a venture capitalist in China we have had the fortune of investing a lot of the other courts including 5a arguments and as I look at the success of these AI companies I realized that many of them are not just creating value but they're also displacing and human jobs and I feel there's a social responsibility to share that with people and also to point to some possible solutions so that's when I came up with a few ideas that might work and then I talked and I wrote ahead it is and then the publisher came and said would you like read that will come this and is the one caveat you don't have experience in the US writing about economy and the philosophy not like enter here says you better put the word China on the cover so we managed to find a way so actually when you're hitting it's now two books in one by the book I wanted to write and also get the takeaway from your book and a lot of people try and write about artificial intelligence and you come away feeling like you knew less than you knew before you began your thesis is in some ways I found very simple that you believe that there is a race for these technologies between the US and China and that the winner if I've got this right and correct me if I don't the winner is going to be determined by whether or not you think we are in the innovation phase of AI or the implementation phase they are right and what do you mean by those well I yes that's part of the book I believe if you look at the last sixty two years of AI development there has been only one breakthrough innovation and that was ten years ago the technology known as deep learning since then I don't think we've seen another equal play through so it's kind of hard to project there would be more breakthroughs every year or every five years or even every ten years so we might very well be in the employment face just like the invention of the electricity operating systems the internet that was one big thing and then lots of ways to use it so I think that we know is definitely happening in accommodation phase and if we are China has a lot of advantages with very tenacious entrepreneurs huge amount of data and the very supported government policies on the other hand UX is also implementing so it's clearly doing fine too but also the u.s. is way ahead in research so to the extent there is a breakthrough and that's definitely allow for that possibility u.s. could very well but one final clarifying thought that was an ad is that US and China actually are running in parallel universes Chinese VCS I mean Chinese companies building products for Chinese customers so this is not really at least commercially not a zero-sum game the game of a Chinese company is not let's not come at the expense of in America that's a crucial point actually I want to I want to tease out a lot of what you said in follow-up but Andrew let me ask you the idea of innovation versus implementation G agree without framing and what do you think about the argument that once we first of all want to thank you for coming here in Monterey and refereeing the battle between I think you and I agree from ninety plushes percent of the way so this might be a boring conversation we'll try to make it spicy and first of all your book is amazing thank you for writing made it as quickly as you can I think it's about three or four books and one you tell a very personal story you have educate us about AI you warn us about job and labor displacement and you talk about this horse race between the United States and China says four books in one it's not surprising me that's already on the New York Times bestseller list and as run as you said in your introduction to the stakes are high these days and I think we continue to underestimate just how high the stakes are because what Kyle and I both deeply believe is that artificial intelligence is one of these things that economists are obsessed about called general-purpose technologies and they're properly obsessed about them because these things come along maybe less often than once a century and when they do they quite literally changed the world they put humanity on a different trajectory and they changed the global balance of power so the first one of these that we really saw was the steam engine kicked off the Industrial Revolution made Great Britain the most powerful country of the 19th century the Sun never set on the British Empire great new launched a successful war against China to keep opium markets open at that company power halfway around the world right around the derivative 20th century there is this one-two punch of internal combustion and electricity which America was most successful at developing exclaims took the lead from the UK in terms of the economy in terms of military power became a dinosaur in the 20th century now here we are going into the 21st century and what Kyle and I both believe are deeply is that we have another general-purpose for reasons we might get into called AI called machine learning the country that is that wins and well maybe we talk about what waiting needs the country that means this one it's very likely to be the big the dominant story of the 21st century and as an American I'm going to play for the home team here my worry is not so much the strengths that China has the - was very very articulate about its the fact that as we start this race I feel like my country is shooting itself in the foot which is not a great way to start off a foot race right so I think maybe we'll talk more about this but I feel like as we're starting this up China is taking very serious steps to in its environment right to succeed I feel like my country is taking some very serious steps to put us in a worse position okay that's a great tip for some of the news of the week which I definitely want to talk to you all about but let me take a step back actually and just yeah I think we have a pretty sophisticated audience here but just just for those that don't really have in their mind yet a definition of what artificial intelligence is I'd like to hear the definition that both of you would give and perhaps how how that's changed in the last you know 10 20 30 years did you mr. Ascher when I first started studying AG through the 8 years ago when I was a sophomore at Columbia I thought hey I was figuring out the human brain and that really didn't happen that's all artificial general intelligence or AGI in short that's still a long term pursuit that may or may not happen over decades what did happen was an amazing pattern recognition technology called deep learning and that is a technology where if you feel it a ton of data in a single domain and then with the outcome that is objective that it will learn to estimate or decide at a much higher accuracy human for example you feel it a lot of fun go games they go better than humans if you feed it a lot of loan data it learns the likelihood that they fall better than humans defeated a lot of Amazon purchase records it has a better idea than humans what things can I show you that you're likely to buy so on single domain large amount of data objective answer type of tasks is providing superhuman actors sounds very limited however if you think about what we humans do on a daily basis or all people's jobs many of the jobs artists just as I described and even those with more complex jobs there are components at the jobs that are routine and can be this is also called artificial narrow intelligence and I and when I talk about AI tonight and in my book I generally refer to a ni because I personally don't see a GI happening in Georgia except this is one of those unfortunate areas where title 9 degrees so this is really uninteresting the AGI quest is not what's going on now but this label you just used is a little bit too conservative of label because it can be narrowing a domain and economically very very very important if there's a big industry based on that domain recognizing fraud and can purchase recommendations and if there are a lot of jobs that happen in that domain diagnosticians are in a very very large category medical industry they're very compensated people I want dr. Watson as soon as possible because the really profound thing that has happened is exactly as kaito says we are no longer the world-champion pattern recognizers even in domains that we invented like playing gold like playing poker this is a terribly subtle difficult pattern matching and we have finally been able to build systems that are demonstrably better than the very best of it one things I've loved into your book wisdom was how you opened it it was in 2017 that the Chinese players played against alphago right tell us why it's such a great story yeah I use that because is more dramatic but the the actual moment where the Chinese people realized how powerful this was was in 2016 when alphago beat these things on the Korean master because Chinese people believe go to be a Chinese invention so there's a sense of ownership there's also a belief that it's not just the high IQ but this experience and wisdom and intuition it's not something machines could easily do all these qualities that we thought were inherent of us were so valuable were not about some stupid machine so happens I happened with yet just yesterday done another interview with Garry Kasparov of all people you know this is a guy that was was 20 30 years ago by by deep blue a lot has changed and he was a little skeptical about AI being all that so what would you say to them about the idea that this was a transform technology that was really going to sort of change everything in the way that we've been talking about even after seeing Africa zero even after after the world I think Gary's appointment and I think both both of us have talked to Gary about this he finds it kind of intellectually narrow which maybe you can make that argument it is not economically yeah okay and so all right so to cut to that let's talk about how artificial technology machine learning is being used in China right now versus the u.s. I mean one of the things that you sketched and I remember how a conversation a similar conversation maybe a year or two ago over dinner with you where you were talking about some of the things that Xiaomi the Chinese company was doing and how China was its own ecosystem now for technology and the penetration of technology in China was so deep and so broad and then when you couple with that the fact that data collection is much more unfettered you don't have for better or so you don't have the same debate about regulation of privacy there as we do here or in Europe that then adds up to something very different than in the US right so can you tell us how that works on the ground sure let's talk about the collection of data for example it's it's not so much that the Chinese users don't care about privacy or are there no laws people just keep using their mobile phones all day and there are all these apps so we checked is a super app in China many how to use it it basically you know most people spend half their time with each other so that captures basically all the things you do socially and think about Facebook if Facebook were to acquire 87 MasterCard and uber and the GrubHub and open table and you know and the list goes on and on that's kind of the WeChat out and that's that's the kind of data Tencent has it doesn't come from you know taking away privacy or stealing paper it's just through use it has the data and also anyone on their ecosystem a JD for e-commerce or a mobile for right sharing with have their savings in the data if you have a retailer and you have a hundred stores you would have your segment on that data so everybody has data and the data is legitimate and it's deep so china now has you know three four times more people than the US and each person has three four times more data that's ten times more data and AI gets better with data so that is the depth but also think about the Chinese are actually I think my book is finally one the other thing I talk about is the Chinese entrepreneurial ecosystem has become a parallel it's been this really worried me because I think that people in Silicon Valley work really hard and you think they're slackers and I missed this quote this is another booking into the best description of the Chinese yeah and then asked today I of course which episode Silicon Valley is basically tech centric vision driven dent in the universe Steve Jobs being the typical role model the Chinese system which began as copycats but because there were copycats it forced the Chinese entrepreneurs to respond really fast execute very well and have come up with and the only way they can win is that everybody else loses because as long as there's a few more copycats in the room they could they could raise some more money and copy you again so you know in the case of Groupon there were 5,000 Chinese copycats improvement so today 4999 there's one left is from a time it's a fifty million dollar company and it's so by making sure everybody died but that's that's not a business model what's interesting about the business model is that the Chinese companies in order to avoid any copy ended up with a magical formula to building uncopyable companies right how do you make a copy that's uncopyable well do something that's so hard that drive people crazy to copy so may time for example for for like five hundred million probably sixty seventy percent of Chinese users they can get mei-fun disrupted the way they ate they can get choice' maybe a few hundred restaurants to deliver to their home within 30 minutes of ordering including cooking for about seventy cents per delivery Wow so once you've done that well you've erected the highest wall and how would anyone else do them that the way that's done is not a light bulb it is grinding away a few cents a month from you know two dollars and seventy down to 170 eventually to 70 cents it's by hiring 600,000 people who would work for less than you know for minimum wages essentially and these are not people who would could get a job to Foxconn factory is it they could they would but now it is about training them to be prettiest and good service providers once they are to get another job so the turnover rate is horrible so you need an HR department that completely replenishes the turnover of 600,000 people and processes that can handle that kind of churn exactly on the HR process and then you need to figure out you know Oprah each one solve it because cars and gas are too expensive so it's electrical moped mopeds then the batteries you ran out of battery then is building a battery replacement centers for them and then of course there's some ai ai that figures out you should go pick up ten boxes here and then get a new battery deliver five boxes and then pick up with eight more boxes they deliver for pick up another battery so that's the routing problem so there is AI and there is all the data but now think may time now first they know all these things about the people and now all that knowledge is allows them to do better hard ass marketing secondly they're worth fifty five billion dollars because they are they're doing twenty five million orders a day and thirdly if you dare to challenge them well first get five billion dollars and hire six hundred thousand people feel free to give it a shot who would fund you so that has erected that wall that's so high and think about the risk that washing over took suppose he didn't know he could give the 70 cents suppose he got to a dollar seventy it stops that meant every day he believed in twenty five million dollars so you're actually bringing up a lot this is the point you've all been waiting for we where we actually just an idle mystery of anything you said except your initial agreement that Silicon Valley is mission driven and Chinese market driven here's my here's why I disagree Silicon Valley talks look at what Silicon Valley does HS new project market fit and the baby babe follow they they they are extremely familiar with the type of the clean book on various entry on winner-take-all markets on scale on grinding out operational efficiency this is what so let's take over as an example the Chinese uber didi they they don't just do all the fancy uber permutations of uber like services they're buying off clarifying gas stations insurance companies car leasing companies creating a total silo for the purpose for the purpose of locking in the driver right as the weakness of uber is that driver has two apps on the phone as long as that's the case they can't have a monopoly so didi would like to lock in the driver so if you want the 10% discount gas 20% is coming surance then you better not sign up toward my well I Travis and will so okay so you're bringing up a couple of I think very profound points yes American companies are extremely ruthless they would love to be able to ring-fence as much as their Chinese counterparts unfortunately they're coming up unfortunate for them they're bouncing up against regulation there is already a fracturing original fracturing happening in terms of how the US may regulate AI and Big Data and tech sector in general how Europe might do it how China will do it that's going to have a lot of consequences I mean it's having the practical consequence which right here but every day when I talk to people in the c-suite of just the absolute confusion about what the standards are going to be in which game they can play in I think it's interesting that you see Google right now kind of caught in the cross hairs of politics around some of these things I mean they're going in and depending whose narrative you believe either starting a search engine other search engine in China or thinking about starting one researching one at the same time that they're being pretty difficult around certain requests and in Washington I mean how do we see these sorts of politics laying out well I think that two parallel universes will stay parallel so any efforts to cross will be incredibly difficult even if there were no trade dispute or regulatory issues if I could brag a little bit I would say that when I ran Google China we brought nine percent market share up to 24% half Google stayed it could be intertwined in this parallel ecosystem perhaps can hold on to maybe you wouldn't even increase a little bit its market share but once you exit and go down to zero re-entry is extremely difficult this is not the statement to say it was right or wrong to exit but it's just a fact that we had that 24% why do you think they're trying to re-enter now okay since I really know nothing I would think if I were at Google executive I would think the Android ecosystem is giving Google nothing right I mean it was giving us so much because you're in Europe from the requirement to put Google Play YouTube Gmail Google search Google Maps and those monetize for Android is free in China Android is still free and those five services aren't in the phones so I would be wanting to find some way to modify somehow those Android phones and there's so many of them that would be one yeah what do you think ended well I love Thai foods image of these different parallel spheres I just don't think that they're going to maintain the quarter between China what we know about history is the one of these powerful technologies comes along the countries that tried to development also try to develop it on a global basis UK became rich by selling steam engines to the entire world the u.s. got rich in the 20th century by selling cars and General Electric to the world I doubt the China that the great companies of China are going to be contented even with that very huge market for a long time I think the countries ambitions of global I think the company's ambitions are too low it's going to be fascinating when we watch the battles outside the Great Firewall of China so I personally agree with that I think the Chinese companies are starting to expand in mobile already and certainly in AI if you see the way it actually expands it's interesting deedie is not expanding into Southeast Asia and a lot of places what it's doing is its some partnering with local partners that compete with uber but when you see them in invest money in the Southeast Asia or South America or other places it comes with tech because suddenly the ride-sharing services are good are competitively pulled over if the AI and the look and feel is injected but I guess you would come back as expansion because it's just because it's through investment and technology injection there is value to be realized maybe be aggregated so in that sense I would see China's expansion in AI mobile to probably roughly appalling order very strong in Southeast Asia very strongly in Africa when the time comes possible and even likely in India and probable in but over time in Middle East and that kind of but almost have no chance in developed in the currently developed countries because of the large I acceptance of the intertwined American technologies and probably concerns about using Chinese technologies just a couple more questions then we're gonna open it up to the audience um let me let me challenge you on an interesting point that you make in the book you talk about how the ability of the state in China to orchestrate economic development have a kind of coherent industrial policy can be an advantage when it comes to AI there are some you know in the u.s. that might argue well if you look at how innovation typically happens it happens in a more decentralized way now you also make the point that some of the companies themselves are very entrepreneurial in decentralized but if we if we assume that there is state control that there might be greater state control the future is that how should we think about that in terms of innovation and an ecosystem that can really you know take AI to the next level in China at this point in time we don't see those signs right at this point in time AI really much left leap forward without government participation until about a year in Africa so we were an early investor five years ago and the series B and C D de pavo were done without any government support and I think governments coming in right now in a very wise public infrastructure kind of way so they're building infrastructure to make autonomous vehicles safer near cities for autonomous vehicle so I think right now it sure looks like a win-win they're doing everything private companies can't do they are forming some investment vehicles taking on late stage funds but I think they compete with the private funds I don't think there's much difference so I think the AI companies and the whole ball companies currently are do not have a high degree of government involvement obviously its own agencies public security bureau military they're developing their a I am sure we don't really know what they are we currently it's kind of two separate well I think in the West assuming there's a strong commingling but but there actually isn't now that's not to say it can't happen in future because one of the things happening in China in traditional industries is one cluster which is like a hybrid hybrid format some public companies and private companies and traditional industries are basically combined together to hopefully get the best of both worlds and what they get typically is the worst of all we look at the actual performance of Chinese state-owned enterprise which is the largest part of the economy it's absolutely dismal and the Chinese high-tech sector has been this one person exception to this for a lot of the reasons that guy who is r2 the question is what's going to happen going forward The Closer these come to SEO ease the more worried everybody was China should be that there's another factor though there which is that the experiment that China's running is so fascinating because this you know communist country started as you say with didn't shopping opened up a large internal market and allow entrepreneurship to flourish and like you point out I love it's just competition to flourish this is a great idea in many America revealed a high tech industry that you have at the same time I think it's probably fair to believe that every Chinese high tech company might not have government investment it certainly has government oversight government influence in a way that's probably a good deal deeper than were used to in the West and we saw the cover story in Bloomberg this week was about how we discovered a tiny microchip and a lot of Chinese of circuit boards which we're about to get installed in worldwide data centers which was almost certainly installed by the People's Liberation Army that is not a great recipe for getting your products why would he accept it well you stole my next question which was going to be the geopolitics of all this so not only did we see that Bloomberg story this week but you have Department of Defense coming out today actually with a new white paper on US supply chains and there's certainly a push interestingly not just from the White House but within the kind of security Hawk community within the far right wing but also the far left interestingly to have reshoring and to really put more pressure at a time when us-china relations are fractions to put pressure on you companies to to insource there's just for many different quarters right now vectors that are kind of pushing at this fight over AI and the technology the future there's worries that the trade war becomes a real war what what do you think about that how you know what how worried are you is a blinking red for you right now for the trade war I think it's it's a problem for both countries I think nobody wins trade wars and when I wrote the book one of my hope foresight benefits would be that we will get to invest a lot more in the US and combine them exceptionally great research in the US with the great implementation and market of China I have given up that hope to be selling more books and have fun with all of you and I'll go back to China and just invest in China the I I don't think I've had much opportunity to invest in the US but I accept that and it doesn't matter too much 95% of our past investments were in China we can't make a 100% it's not in the world about the real war I certainly hope not but I do want to wear you know global chat go global supply chains are very very fluid things and we're going to respond to these pressures like trade wars like national security threats China has built up this amazing industrial base relied or all those global supply chains of technology products that's really that that's like it's a cruise ship it doesn't it's not going to turn slowly but it can turn time okay a lot more to teach now but let's open it up and give you all a chance to comment ask questions go to the entirely new topics don't people introduce themselves first be sure and if you want to introduce yourself we have my exertion people just stand up okay yeah so pass it raise your hand and we're gonna we'll get a mic to you and I don't oh there we go first question you could have cries my name is Jack gonna want to talk about some of the dangers of AI and I know that China has a lot of expertise obviously as a large country has problems dealing with the large population as you know us does as well but probably a very different animal I guess I've been curious about you know some of the dangers of AI and how how we can avoid you know a dystopian future in both countries okay sci-fi future Terminator don't worry about because I don't think we'll get that far I don't believe in singularity I don't believe in this though in assigning a robot overlords about super intelligence I don't think there's any benefit as for that however there are lots of dangers we talked about the job job displacement issue that's one of them also I think security is another one used to talking about PC and long security we're usually the worst cases we install Windows we throw away your PC but in the case of AI it's actually much more complex because viruses are not cold anymore they could just go in and tweak the numbers I don't know how it would it would be very hard to detect because deep learning is just a bunch of numbers which are unreadable anyway and being updated all the time anyway because it's continuously learning and you just tweak a few more numbers and suddenly you know your loan application money could be sucked out and face recognition you could terrorists may not be recognizable anymore and autonomous vehicles be turned into weapons to run into people so I think there needs to be a whole new discipline of security around and I think in either in that burden or error or bugs or malicious are all potential dangers in security I also think we don't understand the side effects of AI some people call it a the bias but for example I don't think Facebook was had a malicious intent to manipulate people's minds and increase prejudice they were just saying I want more peoples I wants to read my newsfeed and let that AI algorithm do whatever it took but it turns out if you were if you had a little bit of prejudice if they might show you more prejudice prejudice prejudice to articles that might make you more prejudiced so these kinds of unintended butterfly effect type of things some books have argued could lead to very big danger and I wouldn't be I wouldn't throw that out yeah unfortunately I couldn't completely agree with that so I apologize to everyone here this notion - I both get involved on these singularity superintelligence discussions way too often he's working chromatic than I am I really just started beating my head against the table in frustration because it's it's so far in the future it is so hypothetical our colleague and roommate has a great way to talk about this risk he says worried about killer robots and big scary AI is like worried about overpopulation on Mars but all the risk that - we talked about in particular he and I are both kind of sounding the alarm about labor displacement that we think is kind of likely to happen the line that I always use is that if current trends continue the people are going to rise up way before the machines do so hi I'm Dan Elliot programming here China Institute so president provoked I flew and Andrew both and lots of thinking about the jobless future and I wonder if you can shed some light on what that looks like and what we're all going to be doing with ourselves okay all right so in my book I go into this in some detail because so we're actually funding companies and we're seeing the companies are going after the routine jobs so the jobs that are routine both white collar and blue collar white collar leaders because blue collar there's mechanical issues in being able to build an iPhone that's not something or bodies can do yet but there is you know almost draw paths we're increasing jobs of routine we'll be displaced over the next I'm saying 15 years in reality the technically feasibility may be 10 to 15 years in practice various things may take longer but these jobs will be displaced by AI that's marginally very low cost doesn't complain doesn't sleep so what to do about that I think there are we can go back and think about what AI cannot do I think the one is creativity strategy conceptual thinking etc but there aren't that many jobs in that category and Retraining isn't good that it be very easy at all but there are probably enough social jobs jobs that require human interaction and compassion and empathy jumps from nannies nurses elderly care teachers managers concierge and so on one could imagine over the future world where there are 10 times as many doctors that rely on an AI engine that has a four-year medical degree from college and maybe more like nurse practitioners and that provide comforts and visit you at home and makes patients feel better the patients are the most vulnerable and the teachers could have AI take care of waiting exams and writing grading homework and giving lectures even the famous people can do that and the teachers can be more than what mentors so if we sort of are a little bit um optimistic and extrapolate a little bit those job categories might be large enough and we train them all for the routine job workers to to be retrained to that category this sounds like crazy thing to worry about because the unemployment rate in the United States just hit me what we eat yearly 49 year low the last time the unemployment rate was this low was 1969 in America so he died something like we're just vastly ignorant talk about this threat but we would look at the job creation in the US are my colleague Bob word is a very good economist says it beautifully said we don't have a job quantity problem we have a job quality problem we used to turn out lots of good old-fashioned middle-class jobs in this country now the job creation engine has down shift and it straining out kind of lower of middle class jobs less well-paid less secure fewer benefits and they'll healthcare all of those kinds of things cut Hughes is braver than me because he's giving kind of a time frame when that situation might tip and we might have a job quantity problem because of his relentless advances in technology I'm not prepared to argue with that you know who knows a lot more of this stuff than we do than I do but if we play our cards right this is good news not bad news everybody wants more wealth and action that's the remote we're talking about if we get this right in most very big changes they're going to think about life of job a career or community how value it involves I completely agreed may be subsidizing some of these caring professions that we will need in large numbers in the future for me the good news is that the world that we're describing you can look at it and kind of a scary to stone in way or by utopian way the optimistic thing to keep in mind is the world were describing is incredibly wealthy world because we're getting all these services done by machines at zero marginal cost holy cow our prosperity is amazing it's a matter of sharing it I mean before I take a next question let me ask you just two quick follow-ups on that what you're both saying implies we need wholesale reform of education not just in the US but probably in China too would you agree with that I mean that to focus more on say liberal arts sort of creative thinking as opposed to just churning out stem graduates yeah you've seen liberal arts these days it's an indoctrination school absolutely what what part of stem is AI not gonna be able to do I want to push back a little bit creativity right but does that argue for a different kind of education than you know what we have now in either country I think the u.s. education is actually closer to what we need to be but still very far the Chinese education is much more alluring and I think we need more empathy why not one mentoring I think we need more attention and social skills and I think we need I personally think we need to discover the gifted and talented early and get them into if they're willing to get them into the economic value creation chain as soon as possible by the way a little more advertising this Sunday on 60 minutes I'll be talking about the education in China a very short summary Wow sometimes when you're behind is you have a leg over advantage so China is now caught up in an environment where there are not enough good teachers lots of mediocre teachers their villages with very very inexperienced teachers so we're now helping to experiment with using a combination of milk face recognition and automatic homework grading complimenting a home test exam rating and also automatic English pronunciation error correction and automatic math drilling to take over the parts of jobs that some teachers may not be good at and then have the teachers do more one our mentor we're doing a great job I think in both of our countries of training out the kinds of workers we needed 70 years ago when the mismatch is just absolutely unbelievable when I think about the educational system that I think both of us were part of were kids how many was sat in an orderly grid of desks and listened to the voice of authority all day every day and did a different thing every hour that's fantastic it'll work in factoring they want you to be the lead in assembly line worker it the point then the actual point was a drive out creativity and to make sure you couldn't talk to your neighbor and develop social skills that was what it was for we really don't need that educational approach anymore and I wish there were large scale experimentation have you the states along exactly the ones that you described there for you hi my name is Joan I mileage for having my license called a tie I have a start up which is social and pet star which is part of the NYU Steinhardt net technically bitter and it's just recently invited to join the coalition for digital intelligence is announced last week at the World Economic Forum its collaboration between an organization called EQ Institute Ocoee CD and I Triple E and it's the goal is to provide a global Sanders nation around digital intelligence to prepare people for digital skills and digital literacies in future and I thought I was fortunate enough to hear he speak earlier this morning and I and you had touch upon a subject around how a mass scale how would you be able to educate your thoughts on this as well in terms of helping to spread the digital intelligence movement that is getting visual literacy and you know well I think one of the good things about about the AI is is that when applied to education it can be more broad right even though it creates wealth inequality from a job sense the accessibility to education can be lowered by using AI and technology to assist on that I think some of the way some of them don't require AI for example one of the things we're trying is a really good super good Beijing teacher each 800 kids at the same time within large TV and then with the modest amount it's better than more because there's interactions with clickers so that's one way to brought them and reduce the cost of education and an another way to think about this we're doing an experiment this is brand new no one's heard that before so it'll be the first where we're trying to make a simulated teacher basically someone to teach you English because natural language is incredibly hard that's why these chat bots are always silly over one after a while but the traditional teacher-student relationship is that the teacher is directed so when the teachers directive and the student is responsive you know let me tell you about fruits what fruit would you like to learn do you know the difference between an apple and an orange which one is this that kind of teacher-student interaction it is possible with now basically 3d graphics motion capture plus natural language understanding it is possible to make a simulated teacher that is directive at least or entry-level learning if that is possible I think then education can be much more broader than that went to 800 nation amen right now the world's most popular way to learn a second language is a free app called duolingo so you can achieve astonishing scale with this as country says when you put cutting-edge AI capabilities on top of that and combine it with the supercomputers but we all carry around these days and really have the opportunity to do have people interact with an entity that will help them apply our very advanced very difficult skills learning a second language is an incredibly involved difficult thing I think we can make progress Ari definitely a lot about what you guys are talking about with preparing a new workforces starting at a younger age and education what about the workers and Willie they're already there how do we kind of address that problem either - great question cuz I hear this from every single CEO I talked to that there are positive well I think there's a new movement for CEOs and companies to take greater responsibility for employees not just to quote employed process but really to ensure that if they exit exit with the skill set that they need I think that's a fairly well defined way I don't have all publishers are willing to do it I recently heard Jeff Bezos talk about this and he talked about how he would hidden its building specifics but he talked about his employees are able to through Amazon subsidy take classes to become a nurse so I've got to be thinking that he has a very clear roadmap as to when his warehouse is getting automated and he won and as one of the beneficiaries read the first trillion dollar company he feels a responsibility to take care of them and I think he sees maybe a similar roadmap so I think corporations if there could be some really good examples we all spread the word I rather bet on corporations taking the first step what about the zero-sum problem I mean you certainly might have some individuals some wealthy companies taking taking those steps 18 T's doing this IBM is doing this but do I mean and remember use your thoughts should we have a depreciation of investment in human resources the same way you do when you buy machinery I mean are there any policy levers that you could see that would help this is a tough one I'm not sure bunking with the corporate tax code is going to achieve the desired goals and your question it is super vexing because the track record of career retraining is not a happy track record as the programs don't look very good especially people farther on in time technology offers some hope but it right now it feels to me like especially with adults trying to pivot and do something else with their careers technology work me is kind of a diamond today detector and what I mean by that is people who have the tenacity the skills the and then the curiosity man they can go scale up and retrain but really interesting things with it like what technology is not so far in the middle of their career is that a rising tide is going to float all the bows I don't know how to get it so it's amplifying the superstar effect with individuals but also companies - in sectors yeah these amplifications all what you guys have talked about is that China has been really able to take apply AI and you know make it very effective in the marketplace but if you look at the basic they are research at my conferences like this ICML those still seem to be dominated by you know like research institutions and companies in the West do you think what is trying to do to bridge that gap do you think that will you know that will converge in the next few years or is that something that's gonna take a lot longer oh that won't converge the US will be in the league for a long time China is getting better right I mean 10 20 years ago China probably priced one percent of the publications now that might have 15 or 20 it depends on the caliber of the conference I guess in the mid to high-level conference China the Chinese publication might be 40 or 50 percent so there's been good improvement but I think the fact is that US has two giant advantages one is just better universities with better research and two is a giant funnel that attracts the whole world's top IQ people to at least one consider studying here and some percentage state so those identities cannot be reversed and China cannot build one you know China does not attract the top students from you know Korea and Nigeria and Italy to study their US and what we're doing right now is pouring cement in what else would you do short of an actually dropping laws a seven to try to cut off the spigot of the world's most you know ambitious talented tenacious people who desperately want to come here to build their lives wow what a great idea unfair for me to grab a second question but thank you to follow up to that which is you said that okay so China doesn't have us has better universities and better research so again back to the news we've been hurt hearing about an ideological crackdown in China and the tightening of intellectual debate discussion whatever at university isn't it sort of rejection of Western thinking etc does that matter is it going to make a difference to China well I think the pursuit of technology and excellence is always a top priority so I I don't think there would be any significant impact but creative thinking we've been talking about the mood for creative thinking that's what I'm curious about you know the creative thinkers in technology in AI in China usually came to the u.s. at a younger age get fully trained here and they go back the education system are just a vast gap that China is improving but it remains you read a lot about AI and so people like you Bal Ferrari best seller of the days and also dr. Bailey of Stanford and Louisville have been talking about this is that you know utopian or dystopian of the are lacking I think at this time whether in China or us a representative force that is being employed or that is creating these startups or companies so whether it's females or whether it's underrepresented groups just like you said you know if it's arts humanities if it's bad or pizza what do you see coming to you are those because I think whatever because really it will be we have a plan to get to fifty fifty percent men and women in my partnership and the plan is to hire one more man so we are actually doing that in our company with out of it we don't think such a VC exists in the world and it's not the top tier of VC so thank you and I think it is about the mindset the environment that you create and the effort that you put in to attract both great people and to enforce adversity I'm absolutely believer in that and I think just more I think that that just needs to be more broadly understood and I think all the work that Sheryl Sandberg Feifei and others are doing that's just wonderful I would say one thing about if we believe in what I said about the transition from the teamwork to compassionate work that's going to really tip the scale a lot because I think you can imagine many of the people will become I think the employee percentage the women will exceed men because of the new follow-up to that um in a couple of weeks in Europe there's going to be a big brussels like conference on bioethics and big data ethics do you I mean to your point who's designing the algorithms do we need more transparency in general and what would be the process for that in order to really understand what's governing these systems that are going to be governing governing our entire economy all of us well I think it's the same issue as the diversity I think you wonder data to be very balanced that's the most important thing obviously the workforce is important a lot of the bias and the data is caused by just a lack of balance I think that is the first step and I think also we need reasonable quality data because garbage in garbage out and I think the the explainable AI part is always because as an AI scientist I actually think AI has to be some sense though itself down to its way to us because the laxity always worries me as somebody that cover the financial crisis I know but at the same time how the 3000 features interacting the 3000 dimensional space we can't comprehend so I think explainable AI would just be asking an AI system to pick out the most prominent three or five things in a way that we can understand but it doesn't truly reflect that all the details it's sort of a make-work to sort of let people feel better but it doesn't address your deep nice earnest and I don't know how to do that yeah yeah let me see I'll show us on this first of all the really good news is the pipeline is getting a lot better very quickly with diversity with gender representation a majority I believe either in near majority or a majority of the advanced degrees in the Natural Sciences and physical sciences I'm never going to the males in America Carnegie Mellon's computer science department now I think is either close to or at 50/50 representation so the pipeline is changing remarkably quickly and those already over the workforce so I'm pretty enthusiastic about that run-up Wow am i nervous a massive governmental oversight of innovation and research activities for all kinds of reasons not the least of which is government sometimes gets a very very raw the nutritional guidelines and all of us Americans we've given the government stamped and approved and super bad this notion that there's this crop of super in London brilliant unbiased unbiased regulators that you need to parachute in the industry that's like in Santa Claus what about FICO scores for your data what about something that gives people citizens an opportunity to correct mistakes to be part of the process you know the EU put in place a right to be forgotten it was just a make-work project for the big tech companies the big American tech they're lobbying pretty hard against it but yeah well it's law across the lobby against it it's a stupid make more project right well what's interesting about that I think that's a really nice test case the right to be forgotten did not give an EU citizen the right to tell the spanish newspaper website to take down the article which they allegedly had a right not to have existed anymore it just made the Americans search engine not point to that our anymore mm-hmm that feels like distance okay I think we have time for a couple more questions there's no huge breakthrough in the implementation tools have become more convenient it's there we can't just stuff it up in half an hour so so what say please all about data and I think our money so the big companies have they have money so they have data and they have a didn't their results with this certainly didn't deserve and they would have more money so the beverage in hiring hiring Israel for startups to fit into this field so you actually touch on that you're looking a little bit you know and you mentioned the fact that you were a little concerned if I remember about academics working alone without access to a lot of data and how they were gonna be able to get into this game yeah so for academics there are a number of efforts including we are funding and then for called AI Challenger which is providing free data for all academics which i think is important otherwise they'll get left behind by the Giants now as far as the startup well I think the art of building a great startup is a combination of course you know a scientist like yourself but also someone who can really raise money right and as well as finding out find the an application area where the Giants aren't going after people right about three years ago the Giants didn't care about face recognition so all these Chinese companies jumped on it in a sense time and face plus plus and then they actually got more data than the Giants and they got more contract and they got their valuation is up now the Giants can't quite stamp the mountain anymore so it is your job as a CEO to see what's the next application that the AI the Giants don't yet have the data and then come up with a credible plan and story to get a lot of money from VCS right I mean look at how much money these computer vision companies have priced and once you have that money and then you secretly go build it don't let their time they know hopefully you have not just that they a lot of data because they can always spend money to get more hopefully you have a pipeline that will get more data coming right one of the first things first things we did to help the face plus plus when there was no vase day that was to connect them with me to all the southeast high so we said well hey let's have face plus plus help you you know make people more beautiful and to do that you have to know where the eyes are the nose are and in order to the value and the who it is and how old the person is and so on so that was the loop that bootstraps up then they got more business and they kept going so the CEOs job is actually the operate especially in China it's about operational excellence and it's that's for rate to the VC and it's about building that product and it's about keeping secret when you need to and finally we want to break that hundred million dollar round that's when you got to do that your PR so it's a CEOs job business is probably a lot tougher than someone from academia with imagine you're asking a question about economies of scale right and your aspiration but David really Jesus yeah yeah economies of scale where that's wrong there would be no disruption there would be no startups there was no intervention when startups do is they turn that strength that economy of scale into a weakness into the ocean away the big cameras but effectively and my career is long enough that I remember when we thought idea who never been disrupted when Microsoft had a stranglehold over innovation in technology when no one was going to hedge no key I think about food disrupted IBM it wasn't another personal delayed main foreign companies who disrupted Microsoft it wasn't an operating system right so Google's not going to get upset unseated by better search engine that's not what's going to happen but are they a permanent a novelist all bet against it since the case for a I am impact investing in my book I talk about it six books with a very different sense because I personally believe the service industry means to have an injection about remorse and capital but they don't provide the same returns and these high tech companies and I would think for impact investing it's kind of making a case that you can return decent amount of return on investment what five 10 percent a year whereas VC's shoes for you know 30 to 40 percent a year me so few percent a year and not lose the money and also the find out fine define metrics of social impact that is degrading to be very specific if we're relief jobs are an issue and that we believe jobs of empathy compassion is what's needed or show that you can create that many jobs and keep people gainfully employed if that goal is met while giving a 5% return I think you can resupply so it's not a question [Music] right actually one that a very strong China regulation on data came out about a year ago very few people here follow that but that was regarding a sale of data that doesn't belong to you and actually the punishment is imprisonment so what happened between Facebook and Cambridge analytically would be actually more severely punished in China so that is one law that has I think made a big difference prior to that there has been various types of trades of data a sale of data and then once that came out I think people stopped that practice and I think I think people don't look to buy data or started a company when you start a company you think about I have technology I go implemented for you as a business user they I never take your data outside of your company but possibly I can take models trained in your company so that's kind of one model the other model would be while I'm gonna collect my own data I'm going to build a consumer company collect my own data use my data to have an iterative loop I think this dependency on buying data and using it to business but my question share your hesitation about the infinite ability of governments the monthly exhaust what I see in the public policy landscape in America is almost a blank slate there's no attention being paid there's no discussion wandering to see that play now because I contrast to China whatever it is they're doing their pain well I think our federal government is not taking an active role because I think they don't know what any I stands for secretary minuchin was asked by a I sympathize and wages and he said we don't see that happening for fifty to a hundred years it's just not on our screen and also at the federal government unfortunately I agree with you we were doing a lot better than the previous administration I don't know if you were at to come the conference that Obama put together in Pittsburgh try and visit administration they would have blue print out late there was a good momentum like I was right on board with it yeah we just got a very different kind of a philosophy at the table but there's any among policy circles in different think tanks and universities and things like that I'm participated this is super active discussion I think are you just talking with the vacuum at the federal level in which case I agree with you sadly I think that the I mean I spent a lot of time talking to these folks the level of knowledge is it's like when I other ways we shoot ourselves in the foot it's by having a government adjust just deeply clueless about this fundamental phenomenon all right we are basically at time but I want to end by asking both of you to put out there a red herring positive or negative a thought that we haven't covered yet in terms of what you're expecting the AI landscape or what you're hoping for in the next five years or something and we do want to start with the last word economy we spend a lot of time of these you know fancy discussions being negative and we need to keep in mind the world's worst problems have never been getting better than they are now the note there the number of people living in dire poverty is lower than it was in 1820 not a lower percent of humanity a lower number of people we have never had the mothers dying at such low rates shown infants these records the gap in infant mortality through the development role in the rich world is closing quicker than we've ever seen before so over the past generation it's appropriate to talk about the problems in the middle class individual it's absolutely appropriate it's appropriate to talk about these problems of a security bias and ethics absolutely while that's going on technology is making the world a better place faster than the world has ever moved before all right we want the end optimistically song for well I think we're looking at about a lot of chaos in the next 20 years I think we actually are looking at other two stages the next 20 years there's a lot of uncertainty on the one hand due to all the trade dispute issues nationalism issues policy issues on the other hand due to the uncertainty about a safety job this job issues and and other things privacy related to AI but I finally believe if we get over this 20 years which I think we will 50 years out I think we'll look at this very differently I think we'll start to ask a lot of will be very thankful we'll be thankful that maybe we don't have to work anymore maybe we just work part-time but we thankful that AI has come to take away the routine jobs which we don't have to do your children and some of the older of us our grandchildren may not have to do any more routine jobs there may be no more household chores there the jobs they have will be fulfilling interesting they're passionate about and and I think you know maybe it is because we have been so engrossed with the workaholics are in this generation or the past by two to ten generations I think it is maybe our maker got sick and tired of it and said I'm going to throw a ing take away all your routine jobs stop running like rats on your wheels now think about what it is that you really want to be and what does it mean to be okay so you can go home and tell those hard-working Chinese entrepreneurs [Applause]

Channel: China Institute

Views: 3,387

Rating: 4.9322033 out of 5

Keywords:

Id: OfVKTyvabDk

Length: 73min 6sec (4386 seconds)

Published: Tue Oct 09 2018