 Ladies and gentlemen, welcome back here for session nine on day two, and I hope you had a productive coffee break. Now we're back here with a very intriguing subject matter that I would assume pertains to each and every one of you in your daily life. And this is why I'm happy now to discuss the topic of artificial intelligence and the future of human labor. My name is Ali Aslan and I have the pleasure of introducing to you this splendid panel. To my left, we have with us the president of the Center for Global Developments and the former director at the Middle East and Central Asia Department at the IMF. Ladies and gentlemen, please welcome Masoud Ahmed. Also with us is the deputy secretary general of the OECD. She also served as a prime minister of Finland, Marie Kivinyemi. Ladies and gentlemen, coming to us from Munich, he's the vice president of advanced concepts at Airbus. Happy to welcome here Holga Mai. Holga Mai is with us, ladies and gentlemen. And last but not least, we have with us the group executive board member at Gemini, Patrick Nicolle. Ladies and gentlemen, now we're trying to make this as lively as possible. I know we have a brief and crisp hour to discuss a very pertinent subject matter. Masoud, I know this is a topic that throughout your time at the IMF, but of course now also is occupying your thoughts. And I would like to get your take on where you see the topic of AI and future of human labor headed. Masoud. Thank you very much, Ali. So first of all, as Ali said at the beginning, you know, artificial intelligence is playing a role in our daily lives in many, many ways. And almost every day now, if you open the newspaper, you will see some article about how self-driving cars are going to use artificial intelligence and that's going to have big consequences for the three million people who work in the transportation industry in the US, for example. But I think the point I want to make is that the development of AI will affect the future of work far more broadly than what you see at the moment. There's been work done by academics which suggest that in the US, for example, almost half of the occupations are going to be impacted in a major way by artificial intelligence. OECD has done a lot of work and then Mari is going to present it shortly, which will show that in OECD countries more generally, we're going to find many, many occupations that range across the board, which are going to be impacted through the development of AI in ways that we can anticipate and in ways that we cannot yet anticipate because AI is developing so fast. But I want to say to you that this is not simply an issue for the OECD or advanced economies. Artificial intelligence is going to affect the nature of work in developing countries and emerging markets and those countries and markets are in many ways less prepared for the consequences of artificial intelligence. Adidas makes 300 million pairs of shoes every year, employing about a million people, mostly in Asia and in Africa. Last year they opened a factory in Germany, which produces 500,000 pairs of shoes using robots. This year they're producing a factory in Atlanta, which will do the same thing. Now, question. Ten years from now, will robots be producing 3 million out of 300 million pairs of shoes for Adidas or 200 million out of the 300 million pairs of shoes that Adidas produces? And if so, the million people that are now working on producing shoes in Vietnam, in China, in Africa, will they be retrained to do something else or will they not have jobs? So will they have new jobs or no jobs? And what does that mean for countries like Ethiopia, countries like Kenya, countries like Senegal or Vietnam, which are now thinking about their development strategy? Traditionally, when you thought about development, development was basically done by taking surplus labor which came from agriculture when agriculture became more productive and you moved them into light manufacturing. But if light manufacturing becomes increasingly done by robots and through artificial intelligence, that run of the development ladder disappears. So where do these people that come off from agriculture go? Are they immediately going to jump into high value added, more sophisticated manufacturing and services? Do they have the skills to do that? Do the education systems that are struggling even to equip people to do the simple manufacturing, are they going to be able to give them the skillset that they need to be able to take on the new jobs that we're all struggling to define? Another dimension of this, and I'm going to stop after that. Another dimension of this is how are we going to cope with the fact that many people during this transition, this is a much faster industrial transition than the three industrial transitions that we have seen in the last 300 years. Because it is happening within one generation, it affects all nature of our work and many of the people are going to find that they cannot retrain themselves fast enough to take on the new jobs. So concept comes up, how are we going to deal with that? And one issue that people are thinking about is to introduce a universal basic income, which is to say everybody should have some basic income. Now it's an interesting concept, but I think it raises many questions about is it affordable? Is it feasible? Can a country like Senegal or Kenya or for that matter a larger country like Pakistan be able to introduce a universal basic income when they don't even have a universal safety net? So I think artificial intelligence is happening. It is coming at us much faster than we anticipate. I personally believe that the people who are complacent and saying, no, no, no, we have gone through this in the past and, you know, we're going to manage it are kidding themselves. I think when it comes, we're going to find that our institutions are not strong enough to deal with it. It will create a great deal of social tension in our society because young people who are coming into the labor market will not have the skills to be able to take on the jobs and the jobs are going to be migrating out. And we need to start planning and preparing for it. I'm not suggesting you try and stop it. It is on balance. It will add to our productivity and to our ability to live better lives. But to do that, we need to equip ourselves now and to make this much more of a national agenda. And I think a panel like this is a great way to get us started into that discussion. Thank you, Masoud. And you got us started very well. Yes, please go ahead. Don't don't. I didn't want to keep the people from your well deserved applause because you were you really touch upon many important and relevant issues that I'm sure we will go through throughout the discussion. Thank you for your initial remarks. And of course, you mentioned the role of the OECD here. The OECD has been taken on the subject in a very prominent manner, has done many high profile policy forums on this particular subject matter. So, Marie, that's why we are, of course, very curious to hear the numbers and statistics of the OECD when it comes to AI and the future of work. So, thank you very much and thank you very much for the invitation. Happy to be here again, my fourth time participating in the World Policy Conference. And congratulations to Thierry for the 10th anniversary of this great event. And really happy to share the OECD's view on these themes on artificial intelligence. Also, this is a technology and future of work. They are very high on the OECD's agenda. And Ms. Dahonitsa said the scene in an excellent way. There's not much to be added, but I go a little bit more in details. But just to give you concrete examples of how rapidly artificial intelligence has penetrated into, not only into our homes, but also into the workplaces. When we look at the number of artificial intelligence-related inventions which were patented in the five top IP offices in the last five years, the number of them has nearly doubled in five years. And also, when we have a look at the funding of artificial intelligence startups, the number of them was in 2012, 160 deals, and 2016, there were 658 deals. So, in a very short time period, really rapid change. And we all see and can notice that artificial intelligence, it can help make better decisions, detect problems earlier, and also generally reduce costs in a number of areas fundamental to societal well-being. And let's take, once again, a concrete example of health. Deep learning algorithms combined with inputs from human pathologists have lowered the error rate for breast cancer detection to 0.5%, compared to 3.5% for just pathologists, or 7.5% for just machines. So, a huge improvement in the breast cancer detection. But then, of course, artificial intelligence also creates challenges, and we organized last week a conference at the OECD on artificial intelligence, under the title, Intelligent Machines, Smart Policies, and really the theme of the future of work was very high on the agenda in those discussions. So, people tend to be very worried about what really will happen at the workplaces and what will be the future of work. But we at the OECD think that these kind of fears could be a little bit exaggerated. And I personally also think that we should be rather optimistic when we think about the future of artificial intelligence, and the future of work. The humankind has been able to survive the earlier technological and industrial revolution, so I'm sure rather convinced that we are going to be able to survive this in an excellent manner. But to give you some reasons why we believe that kind of these fears could be exaggerated. So, first, when you see this slide, you see that there is a difference between what can technically be automated and what will actually be automated. Social attitudes, they matter in deciding where the use of robots is acceptable, and as you can see here in space technology, it's more acceptable than when it comes to health care. And another reason why we think that maybe these fears are a bit exaggerated is that what will be changed are individual tasks, not entire jobs. So, entire jobs won't be to that extent automated, but more the individual tasks. So the change, the nature and content of most jobs rather than resulting in their total automation. And we estimate, you can see, maybe you can see there that there's quite big differences between countries, what may estimate that will happen. But on average, we estimate that 14% of jobs have a high risk that most of their tasks will be automated. But another 32% of jobs are likely to see profound changes in task composition. And the third reason why we think that maybe we fear a bit too much is that technology destroys jobs, but it also creates new jobs. We all can see around many new jobs of which we couldn't even dream of or think of, one example, bloggers who thought 20 years ago that there will be that kind of a profession, what we see now. But really it is so that governments have to be ready to face these changes and we are facing risk of increasing inequality in labour markets and beyond through the changes that digital transformation is bringing to the organisation of work and the way labour markets function. And of course, I want to underline that governments have to be awake in order to help people to navigate digital transformation. And here you see the third reason. Employment rates have risen in most advanced countries in recent years. So the future of work may not be so bad as some people fear. I come to the point what should be then done by the governments. So first we need to adapt our skill policies. The skills composition of jobs is changing and here you see that in almost all advanced countries we have seen a decline in the proportion of middle skill jobs and an increase in the proportion of both low skill and high skill jobs. But really in order to face this phenomena skills composition is at most important and the governments really have to improve the education system when it comes to the basic or the elementary education people need a mix of strong cognitive and soft skills to complement their ICT skills. And also we have to have a look at the lifelong learning possibilities for adults. That system has to be improved much I would say in all the countries. And skills policies, the second policy area is active labour market policies so that workers who lose out in this transition we can provide them necessary income support but also means to find a new high quality job as quickly as possible. So skills, active labour market policies and third also social security policies which we already heard here. There are many possibilities for that but really because the forms of work are changing really also the social security systems have to change. So thank you very much for this opportunity to share the OECD's view on this. Thank you Marie for providing us with the facts and figures and already you see throughout the first two statements of Masut Ahmed and Mari a bit of a contrast to Masut striking if you will a bit more cautious tone not pessimistic but cautious where Mari is saying no, no we are going to prevail a much more optimistic outlook on this topic. So interesting contrast though so let's see where Holga Mai now falls into this place. Your title Vice President of Advanced Concepts. Now that's an ambitious title. I tell you that so with that title and that task that you have at Airbus of course we're very curious to hear where you think AI is headed. Yeah thank you I'm actually I'm not a techie I'm a political scientist and I study security policy defense related issues for about 35 years and my role is to look into the future a little bit and to see what the future security challenges are because the better we understand them the better we understand customer requirements in the future. So that's a little bit the idea and doing that of course you discover usually old things and new things things you have seen for centuries or millenniums even and other things that are new so I try to put artificial intelligence a little bit into context. History always is continuity and change so in a sense many of the future wars we're going to see and challenges we know pretty well from the Roman times in a sense we are Roman Empire plus cyber right things that are old things that are new. So of course sometimes new stuff replaces the old one. We have very little cavalry today although horses could be useful in some areas in Afghanistan actually but when we invented long-range weaponry like bow and arrow it was superior to the club but only as long as the archer didn't come within striking distance of the club then he had a problem. So we're going to see in the future laser weapons cyber war everything but also explosives guns and all that thing. Trend extrapolation usually mean it's about higher faster further but sometimes there's something new. When Henry Ford made this wonderful statement about the cars he's saying if I had asked my customers what they wanted they would have said faster horses. So sometimes we have to explain why we invented the combustion engine and things are going to change and the big changes that we are going to face in nanotechnology, biotechnology, robotics and artificial intelligence and the interesting thing comes up if these four megatrends merge. Well we had a chemical evolution if you wish a biological perhaps one day we have a mental evolution. Actually I can't wait this to happen if you watch TV in the afternoon and you zap through the channels and you see these soap operas and soul striptease shows and whatever I physically suffer after five minutes. I find it so imbearable but this is Homo sapiens is it or is it not? Is this really the result of evolution so far or can we move a little bit beyond that? So let's see. We are at the edge of completely going into the digital production the economy, 3D printing, additive manufacturing and all that. That's only a short step. We will move into the biomolecular production an endless amount of meat without a single animal. You lose an arm in an accident that will regrow. It's only a question of time because it doesn't violate the law of physics. So we have interesting views like Stephen Hawken who says the biggest threat to mankind is artificial intelligence. We have molecular biologists who say, no it's the mutation rate of microorganisms that it's the biggest threat because they undermine all immunization efforts. So let's look at computers. Computers are as we build them today and program them today are instruction executing machines. We write instructions and they do it. They do it wonderfully, beautifully, reliably but they do what we tell them even if they learn we told them to learn a little bit as we learn but what is it exactly that they do differently? Interesting enough, so far they do things very fast. There is a competition in supercomputers every year and three times in a row China won. We are now at or last year at 93 quadrillion operations per second. It's pretty fast. Actually it's nothing if we move into a quantum computer but what it means in other words is computing capability will not be the limiting factor but the question is what is it that you want to calculate? Actually the computers cannot do math very well that may come as a surprise but if I ask any of you what is 3 plus 5 you will immediately say it's 8. The computer can't say that. The computer has to do if this then that, if this then that, if this then that and does it incredibly fast. Is it what it does? So computer do many things better than human beings. They control complex production processes, perform dangerous tasks, all this is very useful and we move on with the automatization and autonomous systems at some stage. We start with that they repeat narrowly predefined actions but they move of course beyond that. We have seen a film where these tiny little flying taxis they of course do sense and avoid and all that by themselves. If it comes to combat robots we maintain the decision to kill and not to kill but everything else could be done by the machine. The question is if you enter a situation where all the issues at stake are predefined very well it's comparatively easy. If you're a commander of a missile defense unit and the sensors tell you there's a ballistic missile approaching one of your cities you better launch the interceptors. You are not calling for a cabinet meeting, right? You do and launch the interceptors but all the situation is pretty clear. Now if you're autonomous driving and you are not the driver, you are the passenger and you enter a dangerous accident situation and it's either you drive over the kit instead and you survive or you drive against the ball the kit is safe but you die. So how will the computer decide? At the moment we have to decide but if the computer decide we'll probably say well Mr. Mayu are now 59 years old you have three adult kits the invoice for the medical doctors goes slowly up we could use your retirement pension fund differently for other nice things. Here's a young guy, good genes, good intelligence pays the whole life into insurances. Don't we agree that the kits will survive? Well even if I don't like it I have probably to accept the overall situation why because if we move into automatic or autonomous driving the likelihood that they enter into a deadly or probably very life endangering situation is millions times lower than if people drive. So probably statistically I have to accept that no matter how I feel but actually then what is intelligence? We don't know. I think what we usually call intelligence is the result of an intelligence test. If you do a test in the result we call intelligence. And intelligence what is it actually? Does it do good things to us? Well we think usually yes but you know lots of the people in the Third Reich in Germany were pretty intelligent but they were very evil at the same time. So what is it exactly what we look at? Is it artificial intelligence or the use of intelligence by artificial organisms or machines and what makes the difference? Well artificial what is artificial actually? Human beings do something but we are very natural I guess and at the same time we apply natural laws so it is so artificial actually. The other thing is of course that we have these emotions and if I explain to my students a very rational cold-blooded so to speak world view and they ask well Professor Mayer where are the emotions? And I say oh yes good that you mentioned it emotions. Hate you know perfectly bestiality. No no no we mean love, charity and care-taking. I say well so you have this tingers with your brain these emotions and other emotions. So at the end of the day it's about us deciding things and how we value things and that is related to culture and all these things. So let's start with the automatization and autonomous systems and have liability problems. Household robots in Japan let's say they take care of the household of elderly people. If somebody gets injured or perhaps even killed by a robot what do we do? Do we put the robot into prison? Would probably make no much sense. Is it the software development? Is it the producer? Is it the guy who sold us the machine? Is it us who bought it? Is it us who switched it on? Who is actually at the end of the day liable? We have to discuss this very carefully as we enter into this game. Now last remarks. Right close by a famous theorist of artificial intelligence had a nice article about 15 or perhaps 20 years ago and the title was the computers will convince us that we are superfluous. Well if we don't want to be superfluous we have to think what our role is in nature. What is it that we can do better than computers? And it's not about that the human being at the end of the day should control the computers because who is it? You want Mr. Hitler to be in control? Stalin or Pol Pot? No, only good people, but who is good? And who defines it? So it always comes back to us our culture, to our value systems, our preferences, our priorities. And we shouldn't just educate our children at the universities to become computers in the sense of memorizing everything and to repeat it for the exams. We should encourage them to think out of the box to challenge conventional wisdom, to be innovative in all these things. And that is what is I think now at the core and at the challenge. And we will see that China is likely to program autonomous systems differently than us because it's an interesting tension between collectivism and individualism. And all these issues are so key when it comes to artificial intelligence. Thank you. Thank you, Holger, for giving us, if you will, a historical, if not philosophical approach to the subject matter of artificial intelligence and, of course, a much-needed plea not to watch television during daytime. I think that that is... Except when your show is on. Well, my show is on in the evening, so I'm safe, I'm safe. But thanks, Mosul, for clarifying that. Now, we've had policymakers, if you will, international organizations, innovators and now coming from a different perspective, Patrick Nicollé, of course, with Capgemini, the industry, to see how the subject matter, which I'm sure is taking up a lot of thought and time, immersing yourself and your colleagues. So, Patrick, take it away. Yeah, Ali, absolutely. So, Capgemini is the largest IT services company in Europe, so I will present the practitioners of you. But before that, I will express an absolute personal conviction. As Holger highlighted it, machine performed tasks. They never replace human beings. And there is un-clarity in our taxonomy today when we debate about artificial intelligence and where the age of the machine. A machine is a machine. And when you look at the component of the machines and especially in artificial intelligence, it starts with algorithm, at the heart of the artificial intelligence technology. And an algorithm is a mathematical answer to a clearly defined problem. The problem can be complex. Some are not solved. The salesman trip is one good example are not solved. And some are sold, sorting data. There are many algorithms to sort data. And then you have a lot of technologies around it, like speech recognition, natural language processes, semantic biometrics, deep learning, swarm technologies, which might be very funny, is how drones fly together. For instance, it's nice to see chatbots, et cetera. But at the heart is the algorithm, which is a strength and a limitation. So what do we see now in terms of implementation in the enterprise? So the first thing that started with the first industrial revolution is touch and move. So what we had, the robot. And we started not by robot, but by automation. Because for economic reason, it's better to invest in a technology that does something simple in a repetitive manner so you depreciate your investment much faster. And then came more sophisticated robot, like painting cars in assembly lines. This is complex. And these robots are much more expensive. And now we have developed capabilities so that they can interact and co-work with human beings. So they have tasks, everybody, so you divide the task. So this was the first sense, if you want, of human being. Because all these developments can be categorized along human senses. So this is the first that we've seen for a while. And that is gaining pace. It is gaining pace, for instance, in IT, in information technology, because the first deployment of artificial intelligence was called robot process automation, which is nothing else than doing automatically a script that another program is asking you to run to make it simple. So the machine does what another machine asks you to do. But it's very simple. It's a very repetitive. The next era that is coming is around listen and talking. This is the second sense that artificial intelligence is going after. This is the most advanced in terms of technology development. And when you discuss with R&D from vendors, such as Microsoft, Google, in that space, they believe that in five years from now, speech recognition, language capabilities from machines will be better than the human beings. We all do mistakes when we speak, when we interpret, when we understand, and they believe in five years from now. It would be a huge leap because today, when you ask and you can try with your phone, you all have my phone with me. But if you all ask Siri or whoever on Android, the rate of response is 30%. Of course, if you ask what is the weather in Marrakesh, you will have an answer. If you ask a more complex question, you won't have an answer. But you will have a polite answer, but not the answer you want. That's the one thing, by the way, socially, is that virtual assistants are very polite. They create problems in human interactions afterwards. But that's where we go. So that's the first area of development. So it touches in terms of activities, of course, everything related to call centers, help desk, which is an important part of the activity. So the third one in development, but is more ten years away from now, we see where, again, the timeline is defined by when the technology will be better than human being, is watching and monitoring. So you've seen a lot about face recognition, about what you can do. It will completely change elements related to cybersecurity. But today, the application we see in terms of what the eye, the vision can do, it's about self-healing. So you can detect default, be it on hardware, be it in systems, through this technology, and then you can anticipate and automatically launch, for instance, the self-healing of a system through this. There are big developments, as you know, in cybersecurity, because this is a much more advanced capability than what we have today, notably when it comes to human being identification. And there is a startup in India called Fluid AI where you can open a bank account just on your, watching your screen and through the recognition, you don't need to touch your PC anymore. You just talk, you move, the machine recognizes you, and it's good enough to go. So this is the next development, but again, it's ten years from now, but it will accelerate. The impact on the employment will be more limited than the first two. The first two, everything that is about moving, touching, and everything that is about listening, talking will have a bigger impact. Here it opens a new field. I'll come back on this. The next area is about knowledge. And here, this is, my view, a big revolution because we used to look as human being, and Google started like this, is we build knowledge repository. Libraries are knowledge repository. We build knowledge repository. And in fact, with AI, this is meaningless. You don't need to build a repository. You need to build the ability to ask the question and access the data wherever they are. We had the question yesterday about the fake news to many data. There are many data that are structured and unstructured, and the structured one are not structured the same everywhere. So you can consider that overall it's all unstructured. And so you scroll through this, and there is a huge increase in data, and 80% of this is totally irrelevant. But it is produced by the machines. So how do you build your knowledge? And I think I will come back on this. It has a profound implication on education. The next one is about analytics. So that's the next area of development. It is another of our human capabilities. Here we started with so-called business intelligence. Business intelligence is making, trying to understand some patterns from structured data, not to be your accounting system, is a set of structured data. It's historical. An analytics is forward-looking. So you must understand trend and adapt to it. Here we are making quite a lot of progress, and a lot of the human-to-machine interaction is driven by analytics. The type of service you propose, the customization, etc. And this is primarily what is changing the business model of almost all industries, this part. And then you have in analytics, but probably we won't discuss it today. It's machine learning. It's where you can program a machine to learn by itself, to execute the best task, and the latest breakthrough you probably all read through is AlphaGo Zero from DeepMind, a subsidiary of Google. It's a machine that could learn the game of Go without human interaction, because you train a robot. When you launch a robot, you program, you always have human beings that accompany the robot. That's a 30, 70% question. So what do you do when you cannot answer the 70%? Someone is doing it. And then the machine is learning from the answers they get, and they progressively improve. AlphaGo Zero is really a breakthrough because this machine started to learn the game of Go without human interaction, without human training. So that's a new frontier. That's what we will see. So these are the areas where we see the application. In recent, thanks to Thierry, I could join the World Press conference on such session two times. So I talk about the impact for employment. I talk about the social impact. But I think the biggest one is in education. And Marie alluded to, she concluded the presentation. I'm convinced, and it starts from the very early age. The way we will have to look at the world compared to the way we were looking at it is fundamentally different in the way you interact. The type of working organization will be completely distributed as well. So the hierarchical, the social model, all the institutions we have built are not geared to address these elements. I'll stop here because I see your look. And I think it's about my time. It's a polite look. Very nice look. Thank you, Patrick. And I think we have really listened to four very intricate and very diverse presentations as far as where the subject matter, where this field of AI is going to take us. We've had risk versus opportunities. We've had opportunism, if you will, and optimism versus some slide of remarks of pessimism here and there, all of which entails this discussion. Now I'm looking at the time. We have approximately 20 minutes left. And with your permission, I would like to take the opportunity of calling upon some individuals here in the audience to address this very panel. I already see two, three individuals. You've had your hand up first. Get the mic. Please introduce yourself quickly. Yeah. We've introduced with the OCP Policy Center. So I have three very quick questions. One is, is it jobs or is it inequality that this is going to be the big issue? The second is education, education, education. But do we run the risk that we're educating people who will not have any jobs at the end of the day? What kind of education? And the third is, does this mean the end of globalization? We just use the machines instead of putting the jobs out. Thank you so much. Now I already see quite a few hands up. So with your permission, I would collect three for questions and then throw it back to you. Who has the microphone? You have the microphone. We'll go to this gentleman and then the lady in the first row. I'm not forgetting those in the back, of course. You will all get your say. Can we get a microphone here in the first row, please? We're going to collect the questions first and then you get an opportunity. Thank you. I'm Eishitry from Israel. Coming from Israel, I've seen a few developments in Israel, especially in the artificial intelligence, which are really amazing. And I'd like to ask the panel members, what do you think are the risks of the artificial intelligence? I'm saying it because it was published that Facebook tried to teach computers to develop by themselves, a software by talking to each other. And after a few months, they found out that those computers developed a totally new language and circumvent all their own guides, human guides. They started talking between themselves without the interference of the people who were working with them. And they closed this operation totally and put off the computers because they were very afraid of what's going to happen if computers will take over. I'm asking to you if you are aware of those developing dangers and what are the dangers. Thank you. Revoking terminated Blade Runner theme. The computer is taken over. The robot is taken over. Go ahead, please. Okay. Daniel Khateeb. My question is for Mr. Nikoli. Many believe that technology had made our world less secure. With the free flow of information, everything is accessible on the Internet. Radicalization is happening on the Internet, especially now with social media. Who can contain this monster? Do you think artificial intelligence is the antidote, especially when you spoke about monitoring, about the voice recognition? Can artificial intelligence be the antidote for this monster created by technology? Can artificial intelligence make our world more secure five years from now? Thank you. All right. Thank you. Could you pass on the mic, too, roast behind you? The gentleman has been very patient. Go ahead, please. Thank you very much. That's almost from Japan. I have a question to Mary. When I was working at the International Energy Agency in charge of oil market, I was constantly told by some producing countries, you are institutionally overestimating supply and the restimating demand to suppress prices. So if there are any institutional risks in OECD to say we shouldn't fear, we are exaggerating fears, but it is really the real feeling of you that there is no institutional tendency to underestimate fears and overestimate the positive side. Thank you. Quite very heavy, important and productive questions. We take two more with your permission and then round up. Go ahead, please, gentlemen. Thanks very much. Don Johnston. Really, my questions are to Massoud and Mitty on the labor market issues. Massoud said, people who say we've seen this before are really not correct. I think that's true to the extent that it's the speed of change that I worry about. But we have seen it before. When I was at the OECD, the concern was Donald Trump's jobs moving to the developing world, from the developed world. The result of that is we have the kind of problem that has created his base in the United States. We did not adapt. We did not solve that problem. We didn't solve it in the United States. We didn't solve it in Canada. What makes you think we're going to be able to solve it with displacement by robots? And also the other question is, the jobs that he wants to bring back, it sounds from what I hear, they're the very jobs that robots will be able to do. So where does that take us? So I just want to know whether you think we're going to do better in the future than we have in the past. Thank you so much. Thank you so much. Would you be so kind to pass on the mic to the lady there? Go ahead. You get the last question. Hi, my name is Natalie Cartwright. I run an AI startup out of Canada. My question is for you, Patrick. And we were the winners of the surge camp forward. So we are grateful for Capgem and I support. You mentioned data in your presentation. From my perspective, the way that we manage data in this AI world is one of the most important and pressing policy questions. Curious if you've got a perspective on how we should start to approach that. Thank you. Thank you so much. Great questions, Masoud. Take the ones that pertain to your field. I want to answer the questions that Uri raised and then that Don raised as well because I think they're connected. See, I think Mari's presentation basically lays out nicely what you could achieve if everything was well and we were a well-organized society and we did the things we needed to do, trained the people, retrained them, and things would work. The fact of the matter is we're not. The fact of the matter is that the pace of technology for the next 10 years is going to be much faster, much deeper than the last 20 years. And as Don said, we've made a mess of it. I mean, so our explanation today is that it wasn't globalization, it was technology that accounts for 70% of the problems that we're experiencing amongst the unemployed. The next 10 years, the pace of technology will be faster. Why do you believe that we will be somehow so much more effective at tackling a bigger problem than we were at tackling a smaller one? If we can, I agree with you. You know, we'll be able to get there. I am not so confident. Now, Uri's point, is it going to be jobs or is it going to be inequality? I think it will not be so much that jobs will disappear, Uri, in my view. I think what will happen is that the nature of jobs will change in a way that many of the people that are currently doing them will not be the right people to do the new jobs and other people may be able to do them, but the ones who are displaced are not going to find other jobs for themselves. And this will exacerbate the inequality that we are now seeing, which we discussed yesterday in your panel. So I think inequality is going to become a bigger problem. And similarly, education, I think we don't really understand what is the education that we need for the jobs of tomorrow. What we all say with great confidence is that the education we provide today is not the right education for tomorrow, but then you, okay, fine. So what should we teach our kids to do? And we say we should teach them to become better at problem-solving creativity and learning as they go, but really we don't have our education systems, and to shift them to do more of that, particularly in countries where education delivery is localized and has a lot of diversity across, is going to be a hard slog ahead. So is this going to be the end of globalization? I think there are a variety of things that are impacting on globalization, which will make it happen in different ways and slower. But I think this will certainly exacerbate the internal social and political tensions and will fuel the kind of populist response that we have seen which has conflated technology with globalization. So I think we do need to bear that in mind. Thank you so much, Masoud. Marie, which questions do you want to address? So there was one directed to me concerning that do we at the OECD underestimate the risks? No, I don't want to say that we don't think about the risks, but I'm a bit afraid of that. If we are too afraid of the technological change, we don't use all the opportunities there are for every single country to perform better. And also when it comes to societal well-being of people. So there's a lot of opportunities. But of course we have all the time had to look at the risks and the threats, and that is something which we have to face together. And also when it comes to artificial intelligence and the regulation, that is something that the countries really have to do together when it comes to risks and also to privacy and security issues. But then to this question of jobs or inequality, what really is needed in all the countries a situation can be improved is the equality of opportunities so that every single person has the possibility to educate themselves that it doesn't depend on your background as is the case in many countries. Now that the countries really can provide quality education from the beginning so every country can improve in that sense. And then when people are qualified which is not the case yet in most of the countries and when we think about the ICT skills we have found out that 50% of adults in OECD countries have almost no ICT skills or at least they are not adequate to really use the opportunities and take up the job opportunities that you have in all the countries. So not only the basic education also the lifelong learning possibilities. But then the question of unemployment and like in Canada and USA, these countries have not been able to solve the challenge of technology or technological development and globalization. So the jobs lost are mostly due to technological development not because of globalization. But we have countries which have been able to solve the problem like Germany where the unemployment rate is close to zero so you can see concrete examples of how to face the challenges and how to solve them and how to reduce the unemployment rate also in a globalized world. Thank you Marie for your perspective and also once again clarifying the OECD position on underestimating the risk which obviously is not the case. There were a couple of questions that were relating to your field I heard. I think that first of all many of the jobs that we know today won't exist in 20 years from now but many of the jobs that will exist in 20 years from now we don't know even today. So it will be adapting. Now in the past if you wanted to become rich you had to invest into a company and invest a lot of money capital to build up a steel mill or whatever. Today you need a computer internet access some good ideas and in a few years you might have a stock market capitalization of your company which outweighs anything we know from the industrial age. So I think the opportunities grow actually. Are there any risks with artificial intelligence? Yes of course but I think that the most terrible things we have seen in history that have been done to human beings have been done by human beings. So perhaps if we think about artificial intelligence and how to program computers at least at the beginning we might actually improve humanity in a sense. We don't know. It's very open. But I think our legs did not invent the earth and the ability to walk. Our eyes did not invent the light and the ability to see. Our brain didn't invent intelligence and the ability to think. It was an evolutionary biology term the other way around. It was an answer by nature to a challenge if you wish. And it was developed of course in a context. So our brain was developed so to speak in a time when we never thought about quantum mechanics and we do it because we can. Now if the brain is not constructed so to speak on hydrocarbonate but built on silicone or gallium arsenide what will be the difference? The artificial intelligence doesn't carry the package of the old days in the jungle and the savannah. Maybe an advantage because as I said I mean there's so terrible things that people do to people and perhaps this whole development helps us to progress in an interesting direction which contains risk but also lots of opportunities. Thank you Olga. Also again for putting it in the historical philosophical context for us. Patrick, I heard a couple of questions that related to your field as well. So more practical. So on the risk side you had a question on the risk side. So the good news is that Facebook could stop the experiment so it meant they were still in control. So I think it's important again the machines do what they are asked to do. Now you have to put in place the mechanism the monitoring and management mechanism that allow you to stay in control. I have other risks that you have not mentioned something I think very troubling like transhumanism development in transhumanism and I would refer to the message of Suzanne Liotto yesterday on Ethique. It's a topic we've been discussing with her for years. How come that in this industry maybe because there are too much money the big companies are too rich it's very difficult to have an ethical launch an ethical debate and Suzanne is working on it. The same that we have had with biotechnology. We will need one so that we understand the consequences and that it creates a framework for the progress. So that would be one answer to this on cybersecurity. Yes, you are right. We are only on the defender side to be clear. We never attack not only because I'm Swiss but it's a group policy. The fact is that we have attacked all sorts of attack now. As I speak we are attacked. So it's permanent which means these attacks are automated. These attacks are AI operated. So you must defend yourself. So it's machine against machine. It's a norms race where you escalate your means. It's cost of fortune for the companies. We don't like it. We have no choice. So it's a norms race. You must compete. And yes, having said that in the companies, the one defending 90% of the compromises, let's call it this way, are created by human errors. Human error is you let a subcontractor plug a PC directly in your network without going through the right procedure. This PC is infected and then you have a 50,000 machine in a matter of three seconds that are compromised. And you have a lot of these every day. So there is a lot about the discipline. And here again, as you cannot change all the behaviors, we would know it, you have to automate. So reduce the number of human interaction with your systems that will reduce the number of mistakes. So yes, the answer is the arms race in the cyberspace. When I come to the next one, which is about managing data, I would like to put two things. The number one, and it was mentioned yesterday, is identity. Because there are data where identity matters. We discussed about fake news. You can think of your HR systems, et cetera. And for the ones not involved in it, you will be surprised that one individual and one company has multiple identity. So you can imagine when you look across all the systems where the individual are connected. So identity is a big challenge when it comes to data. And the second is ownership. And I think here, it's a topic we are addressing with the Ifri and Thierry Nombrial is ownership of data. The American companies, the GAFAM, et cetera, have very rapidly understood the value in data. And they are fighting. I heard yesterday about the money, but Microsoft, I thought one day I should sue them until I realized they had 800 senior lawyers in house. Then I thought, I better find something else. And they have built these capabilities to protect and keep and leverage the value. And I think there is a deficit in Europe. And that's a question we are addressing with Thierry. So identity and ownership. And last point, practical on education to give you one number. Practically what it means, we made business by programming. And now we are rescaling 100,000 colleagues in India from code to order to assemble to order. This is a complete different work. And we have less than, we have maximum two years to do it. It's the scale. And so for us, we need to equip them with soft skills that they didn't have. And then the hard skills to change the way they work. So this is a real practical aspect we are confronted to. Olga Marie and Massoud's time is clicking mercilessly, but very quick last remarks. So what you say is, it's always an offense-defense competition. This is quite natural. They're microorganisms. We have a vaccination. And there's a mutation. It undermines the immune system. So that's very natural. Technology development is natural. But we have to think also in evolution biology to learn about cyber resilience or the resilience against cyber attacks. If the cookie falls on the ground and the kid wants to eat it, let it eat it. It might get ill, but without a provocation and a challenge of the immune system, it will be not strengthened. So to be overprotective as parents is not a good idea. But there are certain diseases which you don't want the kid to get, like smallpox or so, because it's so deadly. So you do have to decouple. Evolution has the same. It's called isolation. It's niche developments. And you need both. So invite red teams. Let them attack you. Learn to survive. But certain things don't put in the internet. Don't put a nuclear power plant on the internet. At least not the critical stuff. It's firewalls don't help. Fire walls are as relevant as walls around medieval cities after invention of artillery. Nice but useless against competent opponents. So at the end of the day, it's resilience because the fighting dog is not just dangerous because it can bite. It is so dangerous because you can beat this thing almost to death and it still bites. It can absorb strikes. Our societies, our economies, our companies need to learn how to absorb strikes. So resilience will be key for room for maneuver, actually. Thank you, Marie. So what is really needed, we need more research and surveys and evidence in order to really face the challenges of artificial intelligence and of the future. And we discussed already a lot of education policies and skills. But what was also mentioned by a colleague here was the universal basic income. So in that area, I do encourage the countries which have introduced these trials on these systems to go further so that we can, based on that evidence, really have a look how we could develop our social security systems in order to face the future of work better. Thank you so much. Masou, we started with you and so we're going to end with you. Take it away. Well, I think the only thing I want to say is that while I'm a bit cautious about the way in which we as society will manage the consequences, I'd say at a personal level, I'm quite looking forward to it because at the moment when I talk to my car and ask it to dial a number, it either lowers the windows or it decides to switch to music for my daughter. So I'm very hopeful that the improvements in listening and learning that Patrick talked about will come soon enough that we will all benefit from many of the opportunities that artificial intelligence will bring to us. And at the risk of offending anyone, we won't mention what car you have. So with that, ladies and gentlemen, Thierry, I think this is a topic that will be with us for a very long time to come. Probably we'll be talking about AI at the 20th anniversary of the WPC as well. For now, ladies and gentlemen, I think I speak for all. When I say this has been a very intricate, substantial, and of course intelligent debate without any artificial ingredients. So with that, I want to thank this panel and thank you for your participation. Thank you.