AI: The Route To A Smarter Future (Kai-Fu Lee & David Kirkpatrick) | DLD Sync

hi hi my dear friends welcome to dldsync let me start this amazing session today with a quote of the new york times of today one of the big questions it's a quote about of today about the future of the internet is whether the world's digital habits will eventually look like china's in the united states and some other countries like ours in germany technology watchers have tried to borrow from some of china's internet habits in the belief that they are a preview of the future or everywhere the nagging question still a quote though is whether borrowing from china's digital world can work is kinda a peek at the future of technology or a digital island on its own to answer these questions i'm happy to to introduce you to david kirkpatrick and kai fully dr kai fully david is a good friend of dld he spoke he is famous for having an interview 2009 with mark zuckerberg at the lda 12 years ago and kaifu kaifu spoke first time at 2011. kaifu lee is the president of innovation ventures he is a famous expert famous pioneer of artificial intelligence and he is a good friend too at dld he is he spoke the first time at 2011 very visionary what about the future of ai and david is the founder and president of taconomy a very recommendable conference in the u.s check it out teconomy it's fabulous it's a fabulous conference so my dear friends the floor is yours thank you so much steffi um it's really an honor to be here with kaifu um and welcome kaifu um so as steffy said kaifu is somebody who's got a tremendous experience in the computer industry and technology broadly he's currently an investor it's innovation ventures in based in beijing but his career prior to that was primarily working for american companies where i met him way way back when he was at apple uh then he was at silicon graphics uh then at microsoft then at google um where i believe at one point you ran google's operations in china kaifu if i don't if i recall correctly certainly you had very senior positions with all those companies uh but throughout you've been an ai expert and thinker in fact even when i met you at apple you were doing work in ai voice recognition um and then of course you wrote a very influential book called ai superpowers china silicon valley and the new world order which came out in 2018 i believe um and uh really re remade i think thinking globally about how ai was evolving because your primary point was that china was coming on in many ways stronger than other countries which i think especially americans found a little bit difficult to believe and accept so here we are a couple years later in the midst of a pandemic that china has done a much better job taming than most other countries even though it did begin there um how would you re return to the basic argument and and update us um how is china doing where's ai going uh how's the you know give us some big picture thoughts about what's happening with ai and the superpowers of the world both technologically and geopolitically um okay thank you it's great to be back at dld uh so i'll quickly recap well the key points in my book uh first i argued that the ai technologies were becoming mature and the speed at which research becomes products is getting so much faster secondly i argue that in that situation data is the new oil that it is not the company that has the most ai researchers that will build the most profitable business but it is the company that has a huge amount of proprietary data that can be used to tune some kind of a business metric and the third point i made is that china has an amazing entrepreneurial ecosystem where the entrepreneurs fight really hard tenaciously in a winner-take-all battle with much greater ambition much greater dedication hard work tenacity perhaps less original innovation but these three points created an environment in which china has all this data and the entrepreneurs are building technology applications tweaking them and and innovating um not as a breakthrough innovation like iphone but tweaking feedback a b tests from the internet and using that to build better products and then collecting data to build ai on top of that and that was really the genesis of um you know the rise to uh power for alibaba and tencent and also a whole uh host of uh companies that are now worth about 200 billion dollars from maituan uh the the food delivery dd uh the the car car sharing jd and um uh pindodo both both in e-commerce and social e-commerce um and uh of course by dance the company that built tick-tock which is uh a lot of the headlines in the news recently and when i wrote the book by dance only occupied a little bit of space because at that time it was only worth a few billion dollars now it's worth a few hundred billion dollars so i think the book is vindicated in the sense that these predictions uh became true and china truly is an equal ai power to the united states if you really go down to details u.s clearly still has the touring award recipients the world class amazing innovators people like um young lacoon and also in north america jeff hinton and and others uh but china is really rising in the next tier of researchers if you measure the top uh 50 of the papers china leads the us and leads europe uh if you measure top 10 percent of the papers china is still catching up but the trajectory is that china will lead u.s and europe in a couple of years so the junior younger resources are coming up uh in terms of applications as i predicted uh it's really moving towards um pervasive use so china now has 18 ai unicorns compared to 20 something in the u.s and these companies are creating a lot of value but what's even more interesting is that traditional companies are becoming really smart in using ai uh from pinging insurance uh to a number of taxi firms working with uh robo taxis and the list goes on so so the infusion of ai into traditional businesses is what will create the greatest value uh so i think these are the points that have been vindicated what i didn't predict was that i had hoped u.s and china would each have individual strengths and find good ways to work together and that obviously did not happen and i knew at the time it was wishful thinking but i thought there was a chance but it did not come to be but perhaps with the new us president things will come back to uh normality and that my wishful thinking might still have a chance of coming true well many of us here in the united states have a lot of we might call it wishful thinking about changes that we're hoping will happen um certainly but um so to summarize what you're saying then is that at the moment the u.s sort of still leads on the sort of pure algorithmic capability but the china because it has so much more data and a much more pragmatic approach across the board toward utilizing ai is really achieving more with it at scale i think that seems to be a summary i mean for one thing i think many people including me right now i'm a little surprised that you describe all those chinese what we call internet giants as a.i superpowers that in every case it's their manipulation and use of large quantities of data with ai that has allowed them to grow so large is that an accurate restatement of what you just said uh yeah also in my book i talked about the four waves of ai the first wave of adoption is clearly uh in the space of internet because internet has the most data and the internet giants can easily hook up the data to an ai algorithm to optimize some kind of a business result whether it is maximizing user minutes revenue profit ads and so on so the the internet giant's got the lowest hanging fruit and they've also become the ai giants but there are three more waves coming so there are more there's more room for other big players second wave is the use of ai in business the third wave is the use of ai imperception vision speech and other forms that can be captured by sensors in iot and then the fourth wave would be autonomous with robotics and autonomous vehicles and drones and all four waves are happening at the same time and each wave i think can create more giants so right now yes the internet giants dominate in terms of ai prowess and the collection of data and the value extracted from the ai but there will be a chance for all other industries as well now and just to quickly get back to this issue of companies uh as a lot of things you've said i want to dig down into but would you say that all of the u.s technology giants are more or less ai superpowers as well uh the ones you worked for let's start with microsoft apple uh and google and then would how would you look at amazon and facebook and if need be could you just quickly go through all five of them and say would you call them all ai superpowers uh yes i would but let me rank them in the order of power i would put google at the top google has the largest amount of data amazing probability to process data with many huge applications and also a collection of probably uh by far the largest ai scientists practical ai scientists turned engineers so i don't think a google has yet unleashed all the power and i think they are in a class of their own followed probably by um i would say microsoft purely because microsoft has more research talent but right now they are not yet unleashed uh in the microsoft products uh i think microsoft is maybe one step away removed from the consumer i rank them number two in ai power but uh they're not making as much money from it as uh let's say um you know amazon or facebook i would put facebook as number three they have a phenomenal research team and huge opportunities but i also think and the first researcher you mentioned john le works at facebook of course too yeah exactly but i don't think facebook has fully utilized ai in the full extent in its products and we can go into details later uh fourth would be amazon if only because they are later to the game but they are in fact i think the best in terms of building products that just work i think about the echo alexa at the time they built it amazon didn't have any world famous speech or ai researchers but they realized the problem of talking to a speaker is not speech recognition algorithm but it's a noise cancellation and microphone arrays and that's the kind of practicality that i think will give them more room for the upside so i think their potential is high and yeah apple uh i have worked there and i think they have a good team siri is a good decent product but overall i i don't think apple has fully embraced ai they've certainly hired an ai team but i wouldn't put them quite in the same caliber as the other four companies yet that is interesting since when i met you something close to 30 years ago and you were at apple you were basically doing ai related work with voice at apple so too bad they didn't uh continue to advance um well there's so many things um let's you know one of the things that's been so interesting here in the states um during the pandemic is the epiphany that big companies have had almost across the board about becoming digital enterprises and um because they had no choice and because everybody was working at home everybody wanted to buy online uh in every industry a more digital interaction seemed to be appropriate and necessary so that things that had long been delayed in almost every industry have been put on steroids as we would say so i'd like to talk about this issue you raised at the very beginning of ai extending into every industry and to compare what's happened in the us this year in particular with what's happened in china and then to get a little more deep into this question of you know where will we see industries and companies aside from those five that we just discussed really become true superpowers because i think aside from the the chinese few of the chinese companies you've mentioned which are very large we haven't really seen outside of tech really anyway um this superpower phenomenon based on ai and i think you do believe that will happen right uh yes so first on the american side i do think the covit has led to a phenomenal growth in the use of data of on the consumer and on the work side so on the consumer side i would say u.s youth uh of um let's say digital payment mobile payment and uh you know food deliveries and things like that are where it's way behind china but the covet kind of forced the issue so now the you know the the products by us square and the paypal are gaining huge traction digital payments uh including connections to cyber currency uh imagine the kind of data that they have is now going to be approaching the kind of data that alibaba and tencent have and that's really good for these digital payment companies doordash is going public it's not quite the maitrend yet but i think more americans are doing taking take out and the virtuous cycle can begin the more users who order takeout the more restaurants will sign up the more restaurants we sign up the lower the cost of delivery and then that increases the data size the ability to monetize as well as the ability target and that will bring down the cost of delivery which is probably the largest inhibitor of ai of food delivery so ai is centrally related to that and there are many other examples the other huge area i think is digitizing work i think because american companies are already the most digital in the world that has enabled google and other companies to allow people to work from home and that has really given rise to a whole set of enterprise software uh from zoom to microsoft home uh to slack and to all the other products that allow this process to happen as well as digital signatures and docusign and so on so i think there is going to be a huge opportunity in the enterprise space with data being collected value being created and of course ai being applied i think the value of ai will come later when the data is aggregated so i think these are all things that bode well for american tech okay it's interesting that you mentioned square and paypal but not jpmorgan chase citibank uh wells fargo etc bank of america do you think we're going to see those companies become ai superpowers or are they going to get bypassed by these more digitally oriented from the from you know from their genesis companies and and talk about how that's looking both in the us as well as china and in europe and the rest of the world as well uh sure i think uh in the fan you know the financial space is a fascinating space because it is quite like the internet in that in the sense that everything's digital everything's virtual so it is the ideal place for ai to be applied and it is the ideal place also for a digital company to potentially disrupt a traditional company so that's why i was quite bullish on the likes of you know and financial uh tencent paypal and square and so on now i do think the traditional companies obviously have a chance they have a much bigger footprint they've got many more users they understand finance better i think the challenge that they are likely to have is that um they they they they will have some uh legacy as well as well as um a need to cannibalize their existing business because with the credit cards and other forms of monetization they're living too you know fat and happy on the business and taking the two percent or whatever transaction fees in the digital world uh all the margins become minimal so can they really cannibalize their existing business and see that the future is all digital that the plastic is going by the wayside to the extent they can really embrace and accept that inevitable reality i think they have a chance a good chance but um history will tell us that most um companies don't traditional companies don't see it until it's too late i mean we're kind of seeing for example electrical vehicles coming out as a surprise when people thought they had no chance when the giant traditional manufacturers and now i think now they will have to wake up but is it too late uh and the the i think that moment of reckoning will come for the financial institutions as well the only caveat i would add is that there are government regulations and how fintech may be regulated and that may still inhibit the fintech growth in certain countries i think the situation is almost exactly the same in china there are many large banks and they're being challenged by these digital upstarts fintech companies so it's quite equivalent to the us accepted in china we just saw last week the government sort of side with the big banks in the uh in its very strong pushback against ant financials ipo so that's a different sort of thing than we see here but it goes to and i would say what you said about the us is similar that these large companies probably still have considerably more political power to figure out ways to uh with law and other tools slow down the uh the uh progress of the digital upstarts to take away their business but whether that's good or not is debatable but look you know we i mentioned europe before let's quickly and as as long as we need let's how where where would you put europe in the discussion we've had so far at a in a big picture way and and talk about this question of ai superpowers and and let me know also do you see the prospect that there are ai superpowers emerging in europe now or likely in the future yeah greatest questions of course preface that by saying we of course are our great friends with steffi and all her colleagues at berta and dld and we recognize our audience here has a lot of europeans in it uh yes i think europe has a bit of a challenge on the positive side europe has great researchers top universities and a large number of good industrial research firms but on the other side on the four waves we talked about there are not many easy um successful industries that is embracing and lead embracing ai creating um economic value and world leading in its products so if you look at the internet companies uh there are very few internet companies at the level of google or alibaba not not even at the level of say paypal or and financial so that's going to be one big challenge then if you go to each of the other levels enterprise software there's sap you go to perception speech and vision there's not that much attraction and then in the automotive there ought to be some players but they haven't really risen up so in the uh landing and in terms of creating uh amazing entrepreneurial companies unicorns there are a couple of examples but the numbers are very small a fraction of what u.s and china have produced and and i think um there are a couple of issues i think one is that europe is not a cohesive market so if you start the company in munich your product may be a success in germany but it's almost start over again when you want to go to france or the uk whereas if you start a product in beijing or a product in new york you will go to the other cities in that country just by the way of you know internet and and and word of mouth so so the fact that markets are separate and and also these and and also governmental efforts uh seem to be more at the country level not at the eu level you know when i visit european countries and sometimes i get invited to set up a small fund in that country never is that country that doesn't care about what i do in eu they will always say come to my country we can invest in you and but you have to invest in my country uh they don't see eu as an entity so eu in the sense of entrepreneurial value uh is um not acting like a unified entity that should have an equal weight to u.s um uh and and and china so i think that that is i think a fundamental issue it's not a technology issue there are a lot of smart researchers the entrepreneurs are quite good uh i think the vcs are not at the level of u.s and china and that's the result of a balkanization that there are these french vcs uk vcs and german dc's and they're not really penn europe either in the way they think about the whole picture or the way that they are incented by the eu right in order you know china in order for china to catch up with the us you know the government gave a lot of guiding funds in other words money to top vcs so that they could spend less time fundraising more time building chinese products eu hasn't been able to do that you know countries like france has been able to do that but but but then france is too small an ecosystem to compete with u.s or china not only in the sense of building a billion dollar software company but also in the sense of the total amount of data that you could gather and of course there are other issues like um i think there there's some many languages and cultures and usage patterns and and even local regulations that make it difficult well ironically there is one way in which europe is looking sort of uniformly at these tech companies and that is at the eu regulatory uh level where on privacy and competition there is a continent-wide pushback against particularly the u.s tech companies as we've seen uh with amazon just this last week or whenever and and google all these major companies are under tremendous pressure possibly with the exception of microsoft but uh they went through it earlier um i guess i'd be curious to know given that that pushback does seem to be likely to be successful in some significant ways in terms of impairing these global american players because it hasn't really been directed at the chinese companies yet will that benefit europe in the terms you're describing or will it just hurt the american companies well certainly extracting some taxes is a somewhat positive thing for for europe i guess you call it fines or taxes whatever but in terms of the the vision i know there is a vision by many people and i greatly respect them for this in the sense of giving the data back to the individual user and creating a whole new ecosystem where the user is in control of his or her data and let services compete on top of that more protected data ownership and privacy and i think that is a very respectable thing to do and i think gdpr makes a excellent step towards that goal but i think when it becomes unrealistic is to hope that there will be a european google or facebook that competes on that new uh playground and force google and facebook to play on that on that premise i think the companies uh have too much control in the situation and i think it's not you know pure coercion isn't necessarily going to get them to play because i think they hold the power the users are basically addicted to these services and can't extract themselves i think what would work better is to fund vcs to invest in new european brand new applications that don't compete against facebook and google uh that's kind of the you know the competition in the last century right so it's time to move on to the next big thing if you if you look at all the exciting things happening with uh you know uh all the types of ai autonomous vehicles um speech recognition if you look at new types of you know even facebook once thought invincible now tick tock is giving them such a hard time and you know um so big players uh can be challenged with not with a direct equivalent product built on a new foundation but i think the giants are always defeated when something totally different comes out and catches them by surprise so i if people really believe in giving data back to the user using gdpr then i think eu should fund european companies to build brand new applications built on that premise it can still do other things like find the american companies but that isn't isn't going to change the level of dependency that people have on these companies yeah good um so more money invested on a pan-european basis that sounds like a good formula which i'm not sure we can uh hold our breath for but uh i want to step back and ask you a bigger picture question that perhaps i should have started with um the world is in a pretty dire place right now i think we would agree in many respects uh with climate uh with geopolitical tensions uh with uh the rise of autocracy in so many countries and democracy on the on the defense in many parts of the world um with the with the pandemic being so severe and likely to remain with us indefinitely and in some form and and likely to future pandemics not being able to be ruled out at all so i guess what i wanted to ask you to do is to step back and and look at everything you know about and tell me and us are you fundamentally optimistic about the world's ability to keep progressing forward and to basically create more value and and health and wealth for everyone uh yes i remain an optimist in the longer term i think you know the world has faced much greater calamities than the ones we're facing and eventually we find a way to get ourselves out of a tough situation i think globalization has maybe made us too optimistic and now we're in a moment of awakening and seeing that there's really multiple sides of the equation that globalization has brought huge benefits to the world but it also kind of led to a polarization of the haves and have-nots and which is the fundamental issue that is uh causing trouble in a lot of countries so i think um i think we're in a in a you know every is we're an up and down um uh wave right now and i think if we look at the history of technology similarly i would say i think people are quite negative on technology right now uh just as people were very negative back when i was at microsoft and it was accused of monopoly so so i think these are waves and and every time i think uh negative wave comes the technology companies uh seem to generally have a a way of learning their lessons and building better products right i mean take a look at microsoft it was um you know 20 years ago it was probably uh generally regarded as the the most evil empire i think it was called but after the antitrust situation i think microsoft now behaves like one of the best citizens in uh in many countries so i think you know google facebook and others are facing some challenges but i think these super smart people will figure out a way to to to get out of the the situation so and technology in the long term has always done a lot more good than than bad well let me ask it in a slightly different way um you know we have 5g beginning to be deployed globally the quantity of data which is something that you talk about as being central to all of this ai capability is is clearly growing how much is ai itself likely to be a tool for us to address and possibly solve or come close to solving some of these massive global problems talk about it from that perspective sure well ai the most amazing thing about ai is that it's an omni use technology that could be applied to any situation where there's data so large amounts of data comes in and you can optimize a particular objective function so the same way that amazon uses it to maximize sales facebook uses it to maximize user minutes uh we can use it to uh you know minimize the suffering of the people who have covid minimize the uh maximize the likelihood that the drug that you invent will be the fastest and most um uh efficacious in in in the way it's applied and and mac and also finding the maximal ways to help people out of poverty you can use it to look over climate data uh to discover solutions uh you can use it for sustainability and and i so i think uh the fact that it's omni-used tool uh the first wave is obviously people who can make a lot of money using this tool i i think we're seeing more and more so-called ai for good or tech for good efforts and i would urge people to search for that term i think you will be uh your heart will be warmed by seeing a lot of the great uses out there interesting well one thing i would respond to is that governments yet still don't really understand the potentiality of any of these tools uh in general and and i think that's a problem uh especially if we're looking at something like climate and i totally agree that there's a real opportunity and with 5g and sensors there's going to be a greater and greater opportunity to use data to to address those problems more more exactly and and deliberately but uh aside from in your in your country i do think china is an exception the government there certainly does understand the potentiality of these tools it's a country with far more engineers in its leadership and that matters i want to go to a a question that steffy asked uh which i could lead to a bunch of other things that i wanted to ask and we will get to audience questions more broadly in the not distant future steffi wanted to hear how is a.i already being let's put it this way how is ai already affecting the daily life of the average chinese citizen and how do you see that evolving over time uh right i think the first is the the use of consumer internet uh it's incredibly convenient uh the example i always give is that um in in china you can now uh for example order takeout and from any uh about a thousand restaurants that can deliver to you for about 70 cents within 30 minutes so this kind of capability would not be possible without ai and this optimization and the impact is not just that we can get more takeout the impact is that it changes the way that the chinese people eat right and and also um take and financial as an example there is a tool that you can go use to pay for anything without ever having to carry cash or credit cards and also the amount the money that you put into and financial will pay you higher interest and you can borrow money against it and you'll soon be able to buy insurance by investments through that so it's um amazing convenience moving up to other areas take robotics as an example covet actually really um pushed up the capabilities of robotics in china because the need for social distancing so in many restaurants the robots are serving the food you know they don't have hands or legs or anything they're basically carts that move to your table and then you take the dish off so that's kind of become a big change and you can place an order on your phone and then the robot brings over the food and you pay and suddenly the waiters waitresses are you know no longer needed obviously not at a high end in a michelin restaurant but these are average restaurants and i'm afraid you may have been disturbed by my phone that was the robot bringing my grocery to my to my apartment uh and uh but the robot was uh another thing it was there is that what you're saying yes the rope the robot it was outside your door yes i uh i actually asked them to turn off the robot from ringing my bell but i think they still called me so the robot's doing that and then did they leave the food at the door of your apartment then the robot yeah when they left the food there yes what does that robot look like really quick uh the robot looks like you could easily do it looks like r2d2 r2d2 and that's our coming that's a routine thing now in beijing it's not routine but it's in any apartment hotel that chooses to spend the money can have it built in because you know autonomous vehicles are hard but robot to your apartment is easy and also there the robotics i think is really taking off we now are seeing a really smart um basically we call them amrs these are robotic devices that can move things in factories so when you go in the factory packages and items are moved automatically you just currently still have humans putting them on and they move by themselves so it's like a fully science fiction but within the warehouse they they don't yet operate so we're taking advantage of the fact that these technologies are not yet good enough for the streets but it's great for controlled environments like warehouses and factories and robotic arms are getting really smart we've seen this silicon hand that can pick up an egg yolk uh from from from an and break an egg and pick up the egg yolk from the from the egg itself and it can pick up irregular shaped objects and we have now robotics systems that can pretty much do the work of technicians in laboratories basically you don't have people moving things around pouring fluids it's all done mechanically and controlled by ai so the list goes on but i think robotics has been pushed forward by covid and i think the consumer convenience is greatly provided oh and my favorite is education the the chinese kids are getting such amazing online education nowadays the power of ai can deliver the following things first for the kids it can the teacher can be become a cartoon character it can be a real teacher but you morph it into a cartoon or it can be a fully animated 3d object that's a lot of fun it's very personalized the teaching can be different for each kid if you're very advanced it moves you faster if you're slower it gives you tests to make sure you capture all your foundations and uh and it's not a talking head next to a white board all of the best chinese educational online tools they are fully interactive so the kids are interacting with an app that is fun while the teacher is talking so it's fully interactive and now there are some virtual teachers and virtual students in the classroom so if you're in the classroom of yeah are a lot of those things available across china are those just things available in shanghai and beijing and a few major cities or is that kind of thing beginning to pervade the whole country oh well the education is pervading for pervasive in the country because it's just an app and and the app has an ai teaching it's very inexpensive it actually makes um education affordable to all the people if it has a human in the other end then it can be expensive depending on the size of the classroom so the last point i was making is there are now virtual teachers who can teach entry-level subjects and interact with students in simple ways not for advanced classes but for entry-level classes and they're also virtual students that can be placed in the room with you you don't actually know who's real and who's not and the virtual students are programmed to make the class interesting and interactive so there are actually measures that show the students are scoring higher with online learning and also more engaged and more satisfied and the parents are more than happy to pay and this is creating a whole new wave of education unicorns in china so that's probably the biggest one that i'm so excited about well that's a diametrical contrast to the united states where not only do we not have that kind of thing very widely deployed at all but 30 percent of students aren't even connected to the internet during a time when their schools are typically closed or highly impaired but let's talk about health care because there's tons of great questions coming in i had a whole bunch more but health care is something we've got to touch on because i think you believe that ai in particular has allowed china to make leaps and leaps forward especially relative to other countries around the world so talk about that sure um i think from a high level uh the current practice of medicine is trained based on the inaccurate inadequacies of our brains so that when your food is coming again kaifu sorry about that uh the people the people um who go through medical schools are that what they're learning is limited by their ability to grasp and analyze and understand because we have a limited amount of memory now when ai comes along it can learn from a huge amount of data from billions of cases of all kinds of cancer and other kinds of illnesses and and it can also personalize and give different treatment for people based on their family history and genomics so this kind of ability to target and to to remember all the cases and optimize for the most the the treatment that leads to the greatest number of people cured that has never been trained and there's no doubt in my mind that given 10 years we will be able to train that individually for some diseases but over time for everything so i think the opportunity is huge but but that's obviously a huge goal and there are a lot of issues with you know malpractice liability and so on that need to be solved but there's no doubt that you know while i admire our doctors uh they're only able to remember a limited number of things for example when i was diagnosed for lymphoma the first doctor misdiagnosed me because the type of lymphoma that i had did not fit the general categorization but in medical school they couldn't make you remember the 300 types or however many hundred types of lymphoma and exactly what factors contribute to them but ai can can do that perfectly so um so i think that's the future opportunity immediately ai can read mri ct radiology better than most radiologists and i think ai can move into pathology ai can be used for a doctor's assistance assistant to provide possible based on the symptoms possible types of problems and treatments but letting the doctor make the final decision ai can be used to optimize insurance ai can be used to help an individual prevent a disease such as diabetes can be connected with you know wearable devices and alert possible cases of anomaly like you know the heart may be not not having some problems so all of these are huge opportunities and also ai can be used for a drug discovery so one could one could apply ai to help a scientist and propose possible types of treatments possible drugs to to look into that have the highest chance of passing clinical trials so i think the opportunities are really really huge yeah well i think you told me on the phone too that given that you know medicine has not been as widely deployed and how health services in in china as maybe in some other countries historically the ability of since everybody's now connected in china unlike most countries and because ai is available at least at some levels in many parts of medicine diagnostics etc the the ability to to allow another in effect to leap forward a major progress in chinese medicine is is far greater than in some other countries so there's a leveling that's happening with overall health care would you am i summarizing you correctly but also let's quickly talk about with covet and the discovery of vaccines how are you seeing ai play a role there uh to be fair i think ai has made some contributions but it hasn't been earth shattering ai has been used in certain cases of drug discovery uh it's used in various types of contact tracing ai is being used to you know um suggest to me for example on my phone there is a special code that indicates that i may have been in contact with someone who has coded and and that i should go get a test so i think it's also i've mentioned the robotics uh in terms of reducing uh contact so this made some contributions uh but i think um you know pandemic is not something we have a lot of data for or experience for after this round if there's ever a pandemic again i think we would be very well prepared as well as using ai to predict and alert before as we all know now if if one could catch this spread early one could potentially extinguish it much more readily so i think you know ai can be used for for anomaly detection and that may be very well applied in the case of potential pandemic okay we've got some really good audience questions i want to go to some of those now in fact what i'm going to do is i'm going to read a couple of three or four of these that are all sort of related um and and and i what i'm interested in in particular in when you address these is they all have to do in one way or another with regulation and ai um and and who really gets power in a world of ai so one was uh where do you see ai developments in the future as countries tend to establish stricter data protection regulations another person asked um uh let's see uh it took it's another person says it took many years to establish existing human-centric laws in the real world now with digital and ai are these rules going to be largely ignored as we design ai driven solutions um and then another person asks you know uh that the world is being divided so much more into believers and deniers winners and losers you know it very inequalities is becoming greater and greater is ai contributing to the division what can ai do um and and finally this one i think is also related to these and that is often the creators of ai algorithms don't even know what they're leading to and the questioner points out that you said that at dld um and so if the people who make ai aren't sure where it's going to lead and if ai is getting more and more pervasive in society and if as i said before governments don't really understand it very well what kind of regulation do we need and what kind of regulation are we going to get uh yeah i think that's a good summary of all the issues that people and concerns people have uh i think we can address them one by one uh i i don't think i know or anyone knows all the answers but i can point to some ways of thinking about it uh at the uh a new anytime a new technology uh is invented there are potential dangers when pc first got connected to the internet there were viruses later antivirus software were invented when electricity was first connected to the homes people were electrocuted then circuit breakers were invented so i think we it's understandable that we see these issues but i actually think technology can be a in a large part a solution to many of the problems that are caused for example when people talk about fairness that when trained on insufficient data on women ai system could be biased in gender but why can't we build ai systems that when you're training it it acts like a smart compiler and says hey wait a minute don't launch this because it doesn't have enough coverage of data so that's something technology can solve we're concerned about privacy you know ai works better when you have a lot of data put together but we don't want to give our data to all these providers well there are new technologies for example federated learning that allows you to keep data on devices that you trust let's say your phone or sites that you trust let's say your hospital and then never share your data outside the trusted environment but only share the models that are trained from yours and other people's data this is not yet a solved problem but one could see that this could potentially let us on the privacy problem let us have our cake and either too that is have some degree of privacy a better protection but but still having the benefit of aggregating data in terms of uh what if we don't know why ai is making decisions well ai doesn't know why it's making decisions it's a very complicated you know thousand layer neural network and the numbers are the best way to explain but unfortunately we don't have the intelligence to comprehend a thousand layers of of numbers but what it can do is take those numbers and tell us that a.i made this decision because of the following 10 factors primarily which may be correlated or uncorrelated and would tell you that this is the kind of work people are doing in explainability so in each of these problems i think people are looking at solutions i do think there needs to be regulations uh and the regulations should be used in conjunction with technologies but i don't think either technologies alone or regulations alone will be will be enough in particular i'm a bit concerned about certain um uh you know government entities that think that regulation is the only way to go um and that um and and not and maybe not even fully understand what how what technologies do that i think is quite dangerous we really need to think of it could i ask you are you thinking of europe when you say that uh sometimes yes yeah okay continue sorry yeah that's that's it okay well let me just say what i just recently i was sort of you seem distracted that's because i had charging near me two devices my apple watch which of course is everyone knows and my amazon halo band which many might criticize me for even being willing to try and i had charged them and now i have both of them on and you're reminding me i mean these are two devices i'm currently wearing that are sending extensive data about my behavior to a giant company amazo apple or amazon apple or amazon uh what whichever like the video is reversed so i'm confused but um and this halo thing which people may not know much about it's it's fairly new and i i think i got a relatively early one you have to apply for it you know it measures my tone of voice it measures my sleep in very great detail every aspect of my activity uh and a number of other factors in terms about my weight etc um and it list the thing that's really amazing is it listens because it listens to my voice my tone of voice it can tell me later you know was i angry or was i nice or whatever i mean clearly we are subject to the data predation of big companies theoretically at least right and and and i want to throw into this well i'll get to a china-related aspect of this in a second but one of the key questions that i have you know and you talked about regulation i you implied this in your answer just now but aside from the possibility that regulation could be more exact and knowledgeable and and and effective do you think that the other way things will evolve is that as in so many parts of technology there will be an arms race where ai on the one hand will be you know invading our privacy but that other companies and institutions will develop ai that will then in in response protect our privacy so is that the world we're going to that one of the big business opportunities may turn out to be to protect privacy with ai and to offer maybe a new unicorn ai superpower maybe one that's and you could say in some ways apple and microsoft both are talking in ways that move in this direction is that is that a reasonable way to think uh yeah i think you can see it that way the example i gave in federated learning is such an example so i think the device companies could be racing to the market with the phone where a very small amount of your data ever leaves the phone but it would then need to have applications that learn to train on the phone and send appropriate data that's not your data it's either you know trained model parameters or encrypted content or something like that so i think it is possible to have more trusted devices that that would protect the privacy the details really have to have still to be worked out uh i don't think the federated learning is a panacea just yet but i do think that opens a interesting opportunity for products made maybe particularly suitable for europe that offers a high degree of protection of your data what are some of the most exciting new companies you're seeing kaifu you're sitting in beijing investing in a lot of ai related companies what either specific companies or categories that you're seeing that are bubbling up that are especially exciting sure um i mentioned a couple of the robotic companies um the the the soft hand that can pick up a egg yolk i think is very exciting uh the education companies i mentioned are very exciting especially when they use you know ai technologies that improve the children's on not only the score but also their engagement and i think their overall interest in learning rather than viewing learning as boring they find it interesting and game like that's exciting another new area i haven't mentioned that we just invested in this week along with softbank is a company called gsx and this is a company building agricultural drones and agricultural autonomous vehicles and they're demonstrating phenomenal return on investment so when you deploy it on your farm you can use it instead of people to do uh harvesting seeding fertilization and the drone will make an aerial map of your farm and figure out what to do that it can do and what's left for you to do and your equipment is paid paid for by the cost in labors just in one year and then it's net savings so it turns out agriculture is an easier space to apply robotics than manufacturing because actually farms are more like each other than factories are like each other right you can imagine the factory for t-shirts versus shoes versus phones are completely different but the farm is a farm relatively similar so that i think is a huge opportunity to allow the whole world to produce much more produce at a much lower cost and that can help eradicate hunger over time what's the scariest new ai development you've encountered company or idea or technology yeah it's not a company but more a video i think made by a professor called i think it's called killer drones or something like that that's when a drone has a little bit of a dynamite and a face recognition it flies in crowds to search for someone's face and assassinates the person this is actually technology that can be built by a terrorist group without with maybe a few thousand dollars or even less so the the fact that um ai uh combined with other technologies uh is um really lowering the bar for access for not just the good guys but the bad guys that's pretty scary um okay here's a question and are you you have any concerns regarding governments using uh or abusing ai to monitor citizens and it goes back to this issue of you know uh the arms race question and one of the things i wanted to mention before and i forgot to was that even in china with all the surveillance technology that's been deployed there has been a citizen pushback and people are concerned i think that's that has surprised a lot of westerners that there was even something of a movement of people saying hey you know some of this stuff might be going a little far with watching us everywhere we go um how do you feel about this issue of governments using ai to monitor people well i live in china i don't feel watched all the time i i certainly know there are places in which is used uh i think we have to remember there are different cultural norms that the idea of individual right to privacy is something very fundamental to europe and u.s and i fully understand and respect that and for asian countries and china as an example things like safety is ranked higher more greater importance and that's why for example during kovit you know the chinese users are okay to have the mobile operators basically uh capture their whereabouts and use it to inform a contract contact tracing algorithms it's not just in china but i think in korea singapore and several other places that have tried it and and i think the point i'm getting at is there are different cultural norms for how much privacy you're willing to trade off in the social contract that you have with your government and obviously in china there are some who disagree with the current social contract just like we see people in the u.s who would also say hey take all my data i just want the convenience that may be in the minority um but but i think you know in each place there are social norms and people who agree and disagree with that somebody here asks an interesting question something that is very good european question it says regarding europe and ai what do you think about what's really being discussed more and more in europe the desire to build a quote-unquote european cloud to counteract the ones of google amazon facebook etc i think it's incredibly important to not fall into the airbus mentality that while airbus was a successful european coalition where the countries together built a a good company and a good product that competed effectively against boeing the internet is a different story and i think great companies have to be based on a market economy investment tremendous technology solving user problems not by coalitions of governments to fight against another country's companies so i would appeal to people to focus more on building a great product that solve users problems and should there be some governmental regulations that says hey you have to satisfy these requirements to run the cloud service in europe and let amazon microsoft google and the european companies compete i would think that would serve better the kind of values that europe is built upon yeah one of the things that we haven't touched on and is related i think to this whole question of data the control of data the use of ai on top of data is this weird reality that a very very small number of companies are the ones creating the clouds that increasingly literally everyone is relying upon and that would be amazon google microsoft alibaba who else in china would you put in that category anybody well tencent huawei uh baidu um all have clouds they each have their clouds of that same kind of scale yeah uh well china cloud is behind the us so they're of a smaller scale but they're growing very fast okay the reason i'm asking is uh do you think i mean that to me is a whole another category of risk that is not understood or discussed when it comes to pushing back against these companies that that if you're especially if you're in a business every business is now having to choose which one are you relying upon and it's a reason why snowflake has been one of the most successful companies to recently emerge because it essentially allows you to have a little bit more of a layer in between that gives you some capacity to switch but uh does that concern you do you think how does it dovetail with this and we may have to make this the last question or we can summarize after this uh well i think you know the business when it's b2b you know companies are really building their reputation on the service level agreements they make with their clients and if you mess up a business data your reputation is ruined so that is kind of the check that we have against the cloud companies that's not to say they couldn't do something bad but that seems to have generally worked there are also encryption and other things that they provide which can add some level of safety but it's interesting that you bring up snowflake because it is a layer that allows interop allows people to move from service to service without creating that strong a dependency so i think that should be an inspiration for people in europe who are thinking about uh gdpr and the possibility of switching um uh away from some american software so maybe that's the kind of uh maybe a european snowflake social layer is what europe can build that's a very interesting way to think of it because everybody should pay attention to what snowflake has accomplished in such a short time it's it's almost shocking but i would just comment you know if one of the things europe is so angry about with amazon is that it's using data that it has about marketplace sellers to create its own products and and then out compete them and that's an issue in the united states too certainly if if that turns out to be true that amazon's doing that there's no reason to be confident that it wouldn't be doing the same thing in the cloud and after all it does host netflix and it did build amazon video so in a prime video so uh you know you could argue this has already happened i think it's a real serious issue which is for another day but just let's wrap up kaifu and maybe stephanie wants to come on with a question of her own but what haven't we talked about that you would just add that we should have touched on in this very extensive and quite exciting conversation we've just had with you any big points we missed you think uh well i i remain an optimist and i believe that technologies over time will contribute to positive aspects of humanity that ai may have all these issues and problems right now but i believe collectively we should have enough wisdom and experience to overcome the problems and make sure that ai ultimately contributes much more than the problems that it creates thank you so much that's a wonderful way to end i think this was a very much enlightening wake-up call for all of us to get more into the topic of ai of their in of its impact of its danger and i really recommend to my dear dld community this wonderful book well here please read it can you see what amazing it's important for all of you to read it because it brings you insights you all should in inhale another promotion here for what is this what is this david oh that's our thank you stephanie that's the upcoming conference that teconomy is having on december 1st through 3rd we have something called the health and wealth of america which is a discussion very much looking at while the pandemic economic downturn uh social division uh is is changing our country and and some of the things that that may need to be done to to combat that we have some great speakers uh coming up and i welcome everybody to go to teconomy.com and check that out so thank you for inviting me to do that session thank you my good dld friend really i really was blown away about the insights you just shared thank you so much see you again at dld whenever wherever stay healthy and stay being an optimist bye thank you stuffy thank you thank you thank you bye bye you

Channel: DLD Conference

Views: 3,304

Keywords: DLD, Burda, 2020, Adding, Sync, Kai-Fu Lee, David Kirkpatrick, Websession, AI, Techonomy, Sinovation, AI in China, China, Smart Future

Id: cPBlCP4V5hg

Length: 67min 31sec (4051 seconds)

Published: Tue Nov 17 2020