good morning after this inspiring talk about the future the near and a little bit more distant future let's come back to the present and see where Europe China and the United States stand in what has been called rightly the race of AI so about Chinese I cannot go very deeply within my short time that I have today into into the features of the Chinese AI but I am very enthusiastic about it and I'm admiring the fast progress so we know there's a strong focus on machine learning there's wonderful work envision many applications envision in China and speech robotics data analytics most research follows the most popular approaches to deep learning and there's a strong advantage through huge volume of data everybody knows that so that part is easy and there has been a national AI strategy and the national AI strategy is actually much more modest than is usually talked about very often people claim that this is a very ambitious bold strategy but actually here if you look at it 2025 China wants to
be synchronized by the world's advanced level not ahead yet here and AI industry will be a new important economic growth point and in 2030 then China wants to be leading but they don't even say we will be the only leader but they want to be leading and actually I think China will reach these goals much earlier and even the numbers that come with these goals the numbers in the effects of a eye on the GDP will come earlier AI has put into schools primary schools middle schools high schools and books in the curricula this is all happening already and now I come to the real point that I want to make today to make things charter I will start with a thesis by Khai Fuli author of the book AI superpowers and he describes the fight between United States and China and then one journalist who was very clever he realized that one continent was missing and he asked CAI fully what about Europe and CAI Fuli said in the race of AI Europe is
not even in the running for bronze medal so meaning I left them out but not by coincidence yeah it's I left them out because there's nothing reasonable to say so my question today is is it really right is Chi fully right he wrote an excellent book so he depicts the whole thing this is the book he says there are these two superpowers United States and China and his book this is my picture now yeah I felt found the picture but this is about how he thinks that the race takes place yeah there are these two superpowers you can guess which one is China and then do you see Europe the green guy up there yeah it's Europe and then there are few others you know a few others not not to talk about them Japan and some other so okay so Chi Fuli then does a very brilliant analysis he roughly a scientist would not be courageous enough to split up AI into these four groups but he does he he says internet AI there
is this is machine learning for search for recommender systems but it also has ecommerce in internet entertainment all these things the second thing he calls business AI analytics on structural not structured structural data structured theater and structured data it is also enterprise decision smart manufacturing everything that goes in business then perception AI machine learning for vision speech gestures haptics also walk for instance your vision as seeing not just faces but also how people walk and then autonomous AI robots autonomous driving so he roughly puts it in those four groups and then he this is his scoring this is actually Kylie's graphics he says now let's compare US and China and in the area of internet AI he says they are equal today but very soon China will lead a little bit and he gives very very good reasons for that and actually I have to say I didn't like it but I like I like his reasons I didn't like the outcome but I think I fully agree so in business AI China he gives
China a very bad score and he says China will catch up but United States will be much better but now question to you this United States so great and business AI okay but better yeah so perception AI you can see the results autonomous AI I think China is already more advanced I disagree that it's 1 to 9 I think China is already further advanced and but I can agree that maybe China and the US will be about the same very hight to distinguish so why does now why do I come to different conclusions not because Lee Chi fook I fully is wrong but he is concentrating on commercialization and product research and let me view it from the from the viewpoint of the real research of the basic research of new methods and I would for in in my world it looks like this up there is the and and I forgot to say the commercialization is of course where the money is made are the other things cost money basic research on product cost
money and then in commercialization is made and yes down at the basic research Europe is still very strong and then the more we come to commercialization the smaller we get because we are used to the fact that the u.s. commercializes for us and why should we do this ourselves I mean if we have somebody else who can do it so why should we do it ourselves so this is and this is there the view that I want to now I want to give you some figures and want to show you I want to convince you that this is the right view so all three continents or all three superpowers produce papers and who produces now most papers u.s.
or China u.s. or China who produces most no more is easy China more than us but who produces most it's Europe so and this is in the past now this is changing by the way the latest figures show that China is increasing a number of papers but wait for a second if we go to the very top conferences say this is now nips 2017 nobody has done the same statistics for 2018 it may have changed I'm sure it has changed but this is United States Europe and then China and then Japan so suddenly things look different if we look at best paper Awards this is organizations with two or more best paper Awards and computer science now more than AI but I checked for AI it's the same yeah US universities are the majority in this case then we have European universities and Chinese universities have fewer than Canadian yeah there are fewer than Canadian universities of organizations that have more than two or two or more best paper Awards so this is simply and
and then of course US global enterprises they are it's very hard to say are the United States that's why I put them in a different bucket because is Google for instance or Facebook United States is Huawei United States yeah in a way or is Huawei China I mean in a way in the way not but now it's more difficult because these papers are written not only most Chinese papers are written by Chinese but are most American papers written by Americans not really let's look at this factor a little better because many Germans and other people are also Europeans Chinese are in America and also we have international authorship in Europe so actually if we look a little more closely at statistics also of scientist exchange we see the following pattern Americans don't really move that much but Chinese move quite a bit they go to China and some of them come back and many Europeans go to United States and some of them come back and not so much movement between Europe and China and
Americans don't really go in large masses this is a list that was put together those are very famous authors who are either forerunners pioneers or the main important thing people in deep learning and machine learning especially deep learning neural networks and this was put together for a chorus here as background literature that came from that point and all the ones in green are European the ones in blue are US and in red are China so this is an interesting thing because United States has maybe as many as Canada or maybe even less that's a that's quite interesting so that means many many things of course come from United States but we're not or didn't originate them so this is a flow of ideas let's look a little bit more closely take three Americans success stories autonomous driving many of you know that actually Germany was far ahead of that in the 70s then we gave it up but the person who picked it up and won the DARPA contests and one was hired as in
residence at Google is a German machine translation Google Translate changed all of machine translation the person who started Google Translate was toast grunts from Germany and then France off was the main architect and there were a couple of other people one even in this room who are from Germany and worked in the machine translation group take deep learning altogether the two most widespread types of neural networks convolution networks lacunae frenchman yeah and before there were basic ideas already by alex viable in germany discovered long time ago lacunae refers to that LS TMS invented in Munich by English mokuba and Zeppo Qaeda so what are the reasons for you asleep first of all it's DARPA funding that was mentioned already was very nicely mentioned by kau Inoa consequence strategic long-term funding it is not even just the amount but it is the way the funding is is the way really for a long term pushing something and then the well financed highly qualified internationally attractive education system and the continuous inflow of top talent and then
the last one is of course very important excellent commercialization climate venture capital risk tolerant attitude large avant-garde early adopter market larger market so I Burnett is a success story of DARPA funding the internet became Mao's Windows hypertext all from DARPA funding Global Positioning System conversational personal assistants call it a way mentioned that they are Siri chalo Siri statistical machine translation self-driving vehicles all by DARPA funding so in this is then then you get this prolific or virtuous circle because you get long-term funding you get success stories and because then you get great commercialization and then you get venture capital and what's the problem now in Europe somebody has to start somebody has to start that thing yeah it is a wonderful circle but somebody has to start it it doesn't start by itself either you have to inject the capital by a government or by the society or into the DARPA like research but you have to start it at some way and where you start I would say doesn't matter but it matters because
there's different speed depending on how you do it and so then what about China China is actually now copying the United States they have understood this mechanism much better than the Europeans the Chinese understand the American mechanism the Europeans understand some parts of it but never the whole thing so DARPA has Descon now China also slides the strategic funding it makes the country attractive it has the thousand talent program to attract people but of course China is suffering from the language barrier and then in the end it's the excellent commercialization climate now the evaluations and money spent is even more is even higher than in United States right now yeah so in some parts so in our European role why why do we have very good research because it's a pressure free climate of intellectual work we have this intellectual diversity in Europe people work already on the non popular topics the topics of yesterday in the pop topics of tomorrow now you are free to work on the topics of today or yesterday or
tomorrow it's fine so great for the ones who work on tomorrow highest degree of academic freedom less Purpose Driven deeply rooted in intellectual culture and a slightly deteriorating but still ok school system we have yeah so that's our pluses and what's the reason for the fast growth in China the size of the market and thews he has an energy of the actors actually the main AI people are known to the public yeah in the morning there was this question asked does the people in this do the people in the street know about a aia when you ask them here in China they do in China they do and they can eat everyone in China knows the four or five names of the most famous Chinese AI people they know them yeah whether they are in China or in Stanford doesn't matter they know them dedication and strong action of government on all levels creating the ecosystem the I echo system awareness and popularity among potential customers end users and then the last thing is also
important more pragmatism than the European yeah so a minimal Viable Product is something that would not be a prototype here now so that is course already out in the streets and if it doesn't sell so well then it gets improved so summing up that fight Europe has excellent basic research bro both in bright and in depth Europe has still great chances in a few areas for instance in enterprise AI business AI I mentioned earlier was the part where China is not doing very well but also us is not doing that well and true it was mentioned several times Germany doesn't have the Facebook's yeah at least not as companies coming from Germany may be present in Germany you know the googles and Facebook but not coming from Germany but we have s AP and software again and we have quite a number of companies that are rather successful in bringing IT into enterprises and not just big enterprises also small and medium enterprises so here we have a chance strangely enough I even believe with
it we have the chance in robotics not in building the robots no I think China can do that better and faster and in greater masses but the Chinese robots are not as intelligent as the best robots we have here in in in Germany but the German robots would be too expensive and too brittle and so but if you join forces yeah there's much space for cooperation health is a big problem and I believe that Europe has a chance there and of course smart energy because we are doing and we have an energy strategy a very strict one and we need AI Europe is weakened commercialization USA a great and research excellent and Brain Gain and successful in commercialization Chinese research is growing fast in size but it's growing not so fast in breadth but why because most of the people are impatient they want to be famous or rich or both very fast so they would only publish and work in the area that gives very fast return they will not work on topics for
tomorrow or the day after tomorrow because they don't think until that time by that time they already want to be millionaires they don't want to wait that long so China will take the lead in commercialization and Touma markets I think there's no way around the only reason why they don't do that in the internet area yet is the language problem and is the missing international spirit in the Chinese companies actually companies like 10-cent and Baidu of course they have the technology to go already and conquer the world at any given time if they really try very hard but right now they have a very big market themselves and China will need more basic research more enterprise AI and brought AI applications so there's plenty of space of fruitful cooperation and there should be cooperation because even if we got use that we don't have to commercialize ourselves and we are quite comfortable with it but maybe there will be some pressure in the future on Europe to do it after all yeah even if we
are used to United States doing it but if we want to do it we need a strong partner with very big market and somebody who needs our basic research and why not for a change work with China we had the illusion that we were working with the US but that only happens on a corporate level I have been in 30 year of 30 years in AI it was very difficult I had maybe had two projects together with the US in those 30 years and it was very difficult to get and those were only very few projects that were funded by both sides jointly it simply didn't happen we were under the illusion that United States and Germany Europe we're close friends working closely together no we didn't I mean United States hired our best people and then they commercialized but we didn't really work together that's wrong so maybe we can find modes to work with China and China cannot yet compete with a brain game of United States I would look Europe does not
have a chance anymore to compete with China in areas where China is already good but in these areas we profit from cooperation but the age of a is just beginning we have great chances in future complex technology paradigms because the next paradigms for instance the paradigm of mixing explicit knowledge like a knowledge graphs Google Baidu knowledge graphs and deep learning those things Europe is not so bad in these areas so we have great chances in the next revolutions of basic methods and in complex industrial applications for instance I work in China also in a working group on AI in industry 4.0 that's a joint project by ministries in Germany and China and there is a true need so the Chinese companies want German experts and companies to come and help them and there's even space for small and medium sized European companies the problem is we need to find the modes of cooperation and the models to allow our companies to get on to these markets that's one of the things I'm interested in now
and working on in my time in China thank you very much for your attention