 Dobro, da so vši, in tudi na moj glasba je dobro, ker smo njih odstavili tudi v Kaliforniji in Italiji, in imam glavno, da so vse, in pojelim, pojelim, kako je na skrinu, ker je šantal, nekaj nekaj nekaj nekaj which I gave a few months ago, since I am in circle. And it is always related to big data artificial intelligence and to ongoing revolution. And this time I will focus a little more on two aspects. On what it really means artificial intelligence zato na veliko najverite vse iz dotovine, avčanje in tako, kaj vidi, da so zabrili. Zelo je ta nekaj finačna trenutka. Tko spetama, da nekaj finačne trenutka z decemifej iz kamerjnja, s delim in z našem srečanjem, in nekaj je prilivil inzibno, kaj je vsak opravljava. Zato, če pravamo, da brimo, ki sem pošličen, in začeljamo, da zelo je odelo. Zelo sem zaprejval, da je to pošličen. Zelo je pošločen. Artyvično idnečno je vedno quitovito svojeh. Szkoli je svih vrste, vsebe, našli in vama v nou, na barzi, na socialni. Včanje je zvoj konfuzion in dragi. This confusion is fueled by the media, who always want make something more appealing to the general public. Usually when they encounter a topic which is potentially very negative implication, they keep Lemmeringon it. je gove jazna imes na vso nekaj, stajelje in nekand do ljenjenju. Se tako je počadno vse z vso nRS, na našem mnoj napravu se termine se rovave, per nekaj 50 jaznih. To je neko, z katerim, ko povolj na zelo, način je jaznomakers. There is only one difference, that while nuclear has more or less only one meaning, z zelo nekaj zelo več, nekaj nekaj nekaj, je to več nekaj zelo, da je misledne. To neče, da je bilo nekaj relamo. To je vsezvej relamo, da so je inplikacija s vsezve in v počkega v humovnih odličkega prijatelja. to je tačno z vseh prav, vzduženju, vzduženju, vzduženju, na socijalizacij, na politiku, na religiju, v vseh. Tudi ni je naredil nekaj nekaj nekaj neče in načinila mu, ki kaj je se vse vesel, nekaj nekaj se vseh, tega da vidimo. chickens is the foot we used to, but now as I was film at discussed that the Ihrumerik. It's it. What is true, it is ok. This is for sva pone. dazu. data in vse do Mrs. laske deljevane infrastrukače. A evo smo se fluxo zvalična vseo zelo kaj pa se potopiti vseo z tega in utravapirka in začuši tudi specijal.ackedi kaj, kaj bi sem zrešal noželji svoj antih. Hellosnih začetvednjih vsega vsega vsega vsega vsega vsega vsega vsega je neuralne netice. Koste je se z taj obsat, da je nekaj neki dževil kljen, svom, nekaj neki dževil neki dževil netice. Zdaj, v evenlyčen tem, da počtem, da nič nekaj dževil in in tudi se bilo ne razmerilo in teda je tudi je vsem kombinator o štešnji vsega odlezvendvana vsega, nekaj imamoelnost, nekaj z nemunistikami, na ne kaj glasbeni, na nekaj dej, ki mali je, na nekaj dej, se najizvanje, nekaj sem? in z njega je zelo početnju. Vse je zelo, da je zelo, da se se vse zelo. Kaj je medija zelo, da je vse zelo, da je vse zelo, da je vse zelo, da je vse zelo, da je vse zelo, da je vse zelo, da je... ...jak se imamo vsi nekaj, ali je vse vse zelo, da je zelo. Ok. Začaj seveda iz neko, da so vse malo počučati, na čeha ne imamo nič zato, kako izpravimo, kako je vsega način vsega in nekako, način je vsega način. Način je, da se nekako vsega vsega nekako vsega vsega vsega. Nekako ne imamo nič način, nekako, nekako vsega, nekako vsega. So, usually when in the field of artificial intelligence, we speak about strong or strong artificial intelligence, we intend a third level artificial intelligence, which is basically to be able to build algorithms capable to make intelligent decision, which shall see what intelligence means, and in a general case, and basically you present the machine with a general problem where it is required to make a decision, and the machine performs better or at least equal to the humans. This third level artificial intelligence will be achieved. I mean, we are getting there. There are different timelines for the topic. We go from very optimistic predictions by Searle, for instance, in 2045, even a little earlier than that. In 2035 could be possible. To very pessimistic estimates, like one by the great physicist Roger Parros, who thinks that it will never be possible. Personally, I am more on the sort of side. I think that artificial intelligence of the third level is very close. I mean, we are seeing clear signs in this direction. But in any case, both these types of intelligence, fourth type and third type, are still to come. I mean, they are far in the future. When they will arrive, we will probably solve the most of the problem, which we are already facing, which are related to the artificial intelligence of the second level. Artificial intelligence of the second level is already here. And this is a big point. Algorithms capable to make intelligent decisions on a specific problem, even a very complex problem, but just one problem. And algorithms, which can make decisions, either learning on the basis of examples, like, for instance, it happens with most neural networks, most other algorithms, basically, you just have a large number of examples of a possible solution to a problem. You feed these examples to the network. The network adjusts itself in order to achieve optimal performances. At the end, you present the network with the new case of the same problem, and it makes a prediction, which very often is much better than what a human could. And this artificial intelligence is already here. We have it for 20 years. What has changed in the last 20 years is the complexity of the problem, which we can address. An example of artificial intelligence of the second level is Deep Blue by IBM, the one which defeated the world champion of chesses already in the 90s, or Watson by IBM, which defeated the champions of Joe Party. All are familiar with it. It's basically a question and answer game, which is very popular on the US television, and which requires not only the capability to understand questions, but also, you know, to provide the correct answer. This is anything but trivial, but again, this is intelligence of the second level. It's nothing to do with the artificial intelligence, which we see in science fiction movies. The other example of this intelligence, I mean, automatic cars with automatic driving, where basically you have cars who are capable to recognize the street, the traffic condition, the incoming vehicles, the signs on the road, and you know, to drive the car in a completely autonomous way, or you know, the landing of airplanes, which is nowadays mostly performed by machines. The pilot has very little to do with the landing. I mean, the landing is completely controlled by computer. Or even the most extreme case, which we had in the case of the automatic landing, the old landing of the Curiosity rover on Mars. In that case, I mean, everything had to be done through algorithms of artificial intelligence, just because there was no due to the delay in the transmission of signals between Mars and the Earth, which is about seven minutes. Since the whole landing lasted less than seven minutes, there was no way in which humans could intervene in the procedure if something went wrong, and therefore, the entire landing procedure was controlled by algorithms. And this was really a masterpiece of the artificial intelligence of the second level, one of the most complex things ever attempted by human beings. More recently, we have had a very extreme case of artificial intelligence of the second level, which is based on a special type of deep learning, which you see on the side, the generative adversarial networks. Where this society, AlphaGo, which immediately after has been bought by Google, produced the software which was capable to self-train itself, without examples, but in the game of Go, which is considered to be the most complex ever invented by human beings. And this is also a game based on strategy rather than rules, so therefore, it requires a very good level of understanding. Well, this machine trained itself in Plingo, and in less than two weeks, if I'm not wrong, was capable to defeat the world champion of Go, something which was considered completely impossible only five years ago. And this was, I think, this has been the highest peak reached by artificial intelligence in the last years. Ok, let's see what is behind all this. And let me make a little detour, first of all. This is something which I already mentioned in the past. The whole thing takes place into the so-called infosphere. The infosphere is a combination of three factors. The internet, without internet we could never have real artificial intelligence. The big data and the computing infrastructures. In one case, the first thing provides the possibility to interconnect heterogeneous data, internet. The second thing provides the algorithms with a huge number of information, a huge amount of information, let it be image, text, data, whatever you think, very heterogeneous data, which can be used by the algorithms to learn a trend, learn a pattern, discover a correlation. Basically, to make, to uncover the model which is behind a phenomenon, behind a decision, behind a trend. It doesn't matter where you apply it. It can be an economic trend, so the trend of stock market in the economic domain. Or it can be, you know, the behavior of supernova in the astrophysical domain. Or it can be, you know, in automatic driving. It can be, you know, the way in which you must react to a specific condition of traffic and of incoming vehicles. The problem does not change. And here we have the first large problem. Big data. Big data, as I already mentioned the other time, represent anomaly in the history of humankind. Because for the first time they are not subject to the states. Therefore, since they are not subject to the states or to the nation as a whole, they are not, at least in democratic countries, they are not subject to the people. Due to a sequence of steps which were underestimated by politicians, by government and so on. And which really require some serious thinking about the freedom of the internet. We can discuss about that if you wish. What happened was the following. When the internet began to grow, there were few groups of people who wanted to index the web. What does it mean? Basically, you have a web composed by n nodes, n nodes where you have different data, different information. But if you don't have a central indexing of these data, of these nodes, you cannot find the information even if it is there. So you need to have algorithm in this, which explore the web in all its possible nodes, in all its possible biforkation. Find the page, index them, so basically extract the meaningful word from this page, the image. And so classify them to a set of keywords. And then these pages need to be ranked so that basically you know which one is more relevant to a specific problem. The huge success of Google is behind one algorithm, which is called page rank, which was made by Sergey Brin, which basically was a very intelligent way to measure the relevance of a node as a function of the number of the nodes which were linked to that specific node. Consider it like a spider web. If you have a node, an intersection where you have many wires coming in, that node is more relevant than another one where you have fewer wires coming in. So basically this is the idea behind it. So what happened was following that the people who were building the browsers, those software tools used to explore the network, immediately had the problem to store the information which was gathered and to process the content of the nodes in order to index them. So basically it was a continuous feedback. The more the network grows, the more storage I need, the more storage I need, the more data I have, the more data I have, the more computing power I need to process these things. So you have a sort of loop. And as a result of this loop, some companies, those which dominated the market nowadays, grew enormously. When people realized the danger which was intrinsic to such huge concentration of information, it was too late. These companies were the only one who had the know-how, the infrastructure and for storage and for computing power. And therefore also the only one who could handle this amount of information. This is where Google, Facebook or the Chinese equivalent, 10 cent and so on, made their fortune and gathered such an enormous power. And there is a power which is very difficult to underestimate. It's very well-leading because this is not a conspiracy theory, this is a matter of fact. They control the network and from the network most of us obtain our information. And therefore controlling the network, you find very seldom articles or reliable information on how these companies are pervasive and how these companies are really shaping our way of thinking, of voting, of doing everything basically. Google is the master of the world in this moment and it's a pleasant but it is definitely true. Ok, where does Google get its enormous power? Here you have for instance, you should begin to have an idea from this slide. This is the Google Galaxy, which is controlled by company named Alphabet. In this you can see Google is just one small part of this company, which was the first one after Apple to go well above $1 trillion of consolidated value. $1 trillion, $1000 million, $1 million, incredible, incredible, never happened before. And where does this money come from? Google is free. You can use the Google browser for free. Therefore, where do they make their money from? Many people would answer from advertisement. That is a very negligible fraction of the amount of money which make the fortune of Google. I mean, merchandising, advertisement, it's less than 10%, if I'm not wrong, of the value. The value is in the access to big data. In a few minutes we shall understand why. And you can see also that when Google bought YouTube, it paid $1.65 billion, it's a huge amount of money. And also it will be surprised by discovering that Google pays $6 billion USD per year to control the download of the apps. Most of the apps which you download on your cell phones are for free. So where does the money come from? Why Google is investing so much to control the traffic, which takes place on the Android cell phone. Same thing for Facebook. Facebook has a huge amount of users, 2.2 billion users, which the same Market Zuckerberg calls useful idiots. The owner of Facebook calls its customers useful idiots, because they provide data to Facebook for free. The value of the company was $700 billion USD, 0.7 trillion. And to buy Instagram, they paid $41 billion for its society, which had 20 million users. When they bought WhatsApp, they paid $19 billion, because they had 1.5 billion users. Same thing for Microsoft. They paid $8.3 billion USD to get Skype with 600 more or less million users. As you can see, the value of a society is directly related to the number of users. And if you think both Instagram, WhatsApp, Messenger, Skype do not produce money directly. You don't pay to use Skype unless you have a special contract to have multiple calls and so on. Or you want to use normal phones. You don't pay to use Instagram. You don't pay to use WhatsApp. So where does this value come from? It comes from the data. The more users, the more data are collected by an application. So at the end, by buying Instagram or Facebook, gained the community of 20 million users, which were useful idiots, providing pictures, providing data. Of course, some of the value is in fact that when you have 20 million users, which you have a profile that basically you can do a fine tuning of the advertisement. So basically, if you are visiting, you know, I am a stamp collector. If I buy many stamps on eBay or Amazon, when I go next time, I use Facebook or I use Google, I get advertisement for stamps or for phishing equipments, which are the only thing I buy online. But this cannot justify the huge value. So let's go back to what is behind the need for these huge companies. What are the real reason behind the need for data? To summarize, the total investment over the last ten years in the world for the infosphere has been more than 3 trillion USD worldwide. Did they do it for, because they want, you know, I don't remember the English for it, because they are very generous. They want to provide the world with services, unpaid services, or they wanted to have a return out of it. For what we know of, you know, large corporations, large companies, obviously the main goal is profit. There is no generosity in making apps for video games or in making WhatsApp available for free to a huge community. There must be an economic return. And the problem, and where does this return come from? It comes from the combination of big data and the strict unbreakable link which it exists between big data and artificial intelligence. And this is something which will develop the second part of this talk. To understand what happens, we, in any case, need to do a little bit of the tour. Why big data are so crucial for artificial intelligence? Let's start from here. And you born baby. When the baby is born, basically it is tabula rasa. In his brain there is nothing. There are all the connections ready to be activated. There are few instinctual reactions to the environment. But you are one of the smartest expert system which starts training itself. And what is really a miracle of the nature is that in less than few weeks the baby recognizes a biberon. I would say the not to pacifier. In Italy, I don't know what is in English for it, the bottle from which they drink milk. Okay, the one which is in the pictures. And this is not a simple operation. And it tells you a lot. These bottles come in all possible shapes, different colors, different shapes. They have completely different shapes if you look to them in a different perspective. If you look at them along the circular session, because you know, you see the diameter of the bottle. If you look at them from the front, you see the pacifier with the bottle. If you look at it from behind, you see only the end of the bottle. But the kid in few weeks learns how to recognize a biberon. Doesn't matter which color, doesn't matter which shape, doesn't matter which orientation in a nanosecond. How does he do that? He has trained a neural network. This is what basically happens in all classifiers. And remember that classifiers basically are behind most of the human activities in the human brain. So it doesn't work. Let's take it a brain. It doesn't matter whether it is artificial or natural. If it is our brain or it's a machine. What you need is something which is capable to do a very simple operation. You expose this brain with 1000 of different images of the same objects, like for instance in this case a chicken. And by combining the information in shape, color, in complexity of all the images of a chicken, the brain builds a sort of meta chicken, which is its classification of chickens. And this is incredibly powerful algorithm. Remember, we are here talking about one specific problem. I want to recognize chicken. So from the point of view of machines we are dealing with the artificial intelligence of the second level. I have a specific problem. I want to teach a machine to scan through thousands of images and recognize those where there is a chicken. Ok. You train the first classifier to recognize the chicken. You expose the machine to hundreds or thousands of images of chickens. Or you expose the human brain of a baby to thousands or few hundreds of images of a biberon in different shapes or rotations. And they both build a meta idea of that object. And they both build a meta idea of that object. They train a classifier to recognize those objects. Then you submit to your train at the classifier a new image, in this case a black rooster. And automatically your brain, if there is no other example, classifies that object as a chicken. Even though he has never seen that type of chicken before. But somehow in the brain some patterns, some correlation between colors, shapes, it's difficult to quantify what. But basically some correlations trigger the classifier to say it is a chicken. Repeat the operation with more than one classifier. For instance, you first train to recognize a chicken, then you train it to recognize a boxer. And so at the end you have two classifiers which are trained together. One recognizes chickens, the other one recognizes boxers. And if you expose it to cases which they have never seen before, basically you immediately know that in one case you are looking to a chicken and in the other one you are looking to a boxer, even though both have never been, do not belong to your training set, do not belong to your previous expertise. If you doubt that this is the way the human brain works, I can give you thousands of examples which show you that is exactly what the human brain works. Optical illusions are a typical example of how the human brain gets confused, goes in a sort of a still position, if it is exposed to something for which it has never been trained, exactly what happens to the machines. Now, so basically what you can have, and this is what is artificial intelligence or with which we are dealing right now, basically you can train, if you have enough computing power, enough examples, you can train N classifier where N is a very large number, each one of them is specifically trained to perform one and only one operations. And trust me, in most cases they can do that operation much better than any human brain. They do it with higher accuracy, they do it with lower level errors and they basically outperform the human brain in performing that specific task. And this is already something which has huge implication on the structure of those sites. Then, what basically makes the difference between the second level and the third and the second plus something level, is that we need to perform more complex classification at once and this is what is done with the deep learning. You have N classifiers working together and the information is somehow combined by what we call a sort of meta classifier. This is not correct from a mathematical point of view, but it helps to understand what happens, which basically allows you to perform things like that. At different level of the network, you have different classifier with work together so when you expose them to the image on the left, you have a first classifier which classifies it as a chicken, a second classifier which tells you that it is a male chicken, you have a third classifier which tells you that it is not a real picture but a cartoon and also that since it has glosses on it is a boxer, so the final interpretation you have a cartoon of a boxing hen. But behind, you have hundreds of other classifiers which are at work which tell you that this thing is wearing pants which are colored, which are blue, gloves which are colored, which are red, it has a beak which is curved, colored, yellow point. So basically you have that the problem is dealt with a level of complexities which was completely unprecedented until a few years ago. And this is why in these days there is so much speaking, so much talking about deep learning. From a mathematical point of view, there is nothing new. These algorithms have been around for the last 20 years with the different names, they are just chains over neural networks, they are ensemble over neural networks. There is nothing new from the mathematical point of view. The only huge difference which makes them a very effective tool is that nowadays we have the big data provider and we have huge computing infrastructure because to train these networks even for simple tasks requires a large amount of computing time, a large number, especially if they need to self train themselves, they need a large number of examples. So all these would have been impossible without big data and without large computational infrastructure. But where is the impact of this? Here we are. The big data provider controlled the computing infrastructure, controlled the big data and controlled the algorithms because they have huge number of scientists who work in improving the algorithms. This is why these large companies buy YouTube to get access to the huge amount of images which are in a single video and to the huge amount of information which is there which can be used. All they buy WhatsApp or Instagram and so on. All these information are processed in order to train classifiers either with deep learning or with other algorithms but it doesn't make much difference to perform very specific tasks which nowadays, thanks to the computing power can be also very complex. Is what for instance Siri does or Alexa does when they understand, very often also quite complex sentences and react to these sentences. This is what most automatic translator do and this type of everything is reprocessed neural networks of different types are trained, the results are stored in the cloud, your request goes directly to the cloud where it goes through the neural network and you receive the answer back basically real time. And this is where the huge power and huge economic value of the big data provider comes out because this thing it's revolutionizing the world already nice. Already now, sorry. Let me spend a few words on this. We think that this is a revolution, cultural revolution like the ones which we have already had many times in the past at the time of the invention of printing for instance, or at the time of industrial revolution. Not true. It is the first time that these humankind is facing with the revolution with this not to do with the way we make things, revolution which affects the home of other but it is the first time which we are dealing with the revolution which affects the home of others, the way we think, the way we understand things. And we are completely unprepared to deal with this type of revolution. Let me make you just a very few examples and then we can go to the question if you are ready. In the past, take the revolution in the transportation which took place between more or less at the same time of industrial revolution. You had horses which were used to transport material people then you had carriages then carriages were substituted by trains, by cars, but basically the people let's say not who had not undergone through specific studies or very advanced level of studies with low level skills were working either as horse riders or as drivers of carriages or as operator of trains or as drivers. It is even more evident if you take into consideration another type of revolution. When you move from agriculture to industrial age where you have basically the coal industry fueling the world and then you have the oil industry fueling the world you are not laying down people with no skill in the pre-industrial age peasants who had not studied who had not a very high level of knowledge working in the fields then you had coal miners and then you have oil drillers but there has always been a place for low level job not in sense that they are not important but low level in sense that they do not require very high level of training from the point of view of studies they do not require very high level of knowledge mathematics and so on because this revolution were affecting the way human beings do things not the way human beings think or react to the world which surrounds them so this revolution basically triggered social unrest but it was very moderate in comparison to what is going to happen in the coming years in the next 5-10 years artificial intelligence is affecting the way we think and the way we control any process take for example this example I think is crucial truck drivers the final end the death revolution in transports truck drivers will disappear in 5-10 years this is not sense fiction on the right you see the first prototypes and all companies Volkswagen, Audi, General Motors, everyone are building prototypes of self-driving trucks you will lay down hundreds of thousands of truck drivers who cannot be converted in the next technology because these things do not need humans anymore so what will happen to 100,000 truck drivers forever and remember that for everyone this is the estimate 1000 low level jobs this revolution is going to create only 5 high level jobs and this is going to affect everything all aspects of our life let me give you a further example now let's take this is about pros and cons for instance we are talking about global pollution let's think about agriculture until now two main problems insecticide, pesticides are distributed randomly on fields by for instance the plane you see on the left 80-85% of the insecticide is completely useless because it goes between plants on the soil and is absolutely wasted it contributes to pollution, environmental change environmental damage so basically at the end it contributes to cancer and other things so basically it is a very ineffective way of doing things also we have a problem with monocultures we are obliged to grow in a specific parcel of land only one type of crop just because you need to optimize harvesting if you need to make mass production you need to have all the corn reaching maturity so that you can send one of these huge machines cutting down all the corn and processing it at once so since to have a single culture on land and polish the land of nutrients you have to rotate land which means basically you have to change type of cultivation every year or every four years so one year you have mice the other year you have corn the other year you have millet this has a huge impact from the point of view of effectiveness because basically you have very often parcel of land which are left uncultivated for three years we need the food and also you know it is very expensive in terms of labor labor power you need to have many people to control this this process well in ten years also this will change drastically and this is already happening now again it is not some prediction from science fiction sorry now again I get confused already now in France insecticides are being tested in a competitive in on several fields many many large fields are being tested with automatic artificial intelligence basically you have small robots which explore the cultivation the place where the crops are they look at each individual plant they measure the amount of parasites and they spray exactly the proper amount of insecticide on each parasite this allows you to say 85% of an insecticide incredibly effective very convenient from the economic point of view no need anymore for people spraying insecticide also this experiment is taking place in California will soon be exported everywhere rotation no longer needed if instead of using this machine for the harvesting the one which you see on the left you use small robots which walk through the field and you basically cultivated this field with the mixtures of seeds basically you put together millet, corn and mice and now because they reach maturation in different times therefore if you call pick up the corn you are destroying the other things but if you do it with this machine but if you use a robot this is much more effective because the robot goes there stays at the plant says this corn is mature picks it up and puts it in the box goes to the next one this is millet needs to stay here for another week this is enormously the number of the production the productivity of the land but again what happens to the workers peasants no longer needed you say this is not going to happen very soon trust me in 20 years from now this will be really the normal situation in agriculture because probably advantages from the point of view economic point of view the owners of the land the people who produce food will save a huge amount of money they will not be obliged to pay salaries this machine basically lasts for very long are very effective consume very little energy so I mean it's okay perfect there is no way to stop it well again the ratio here obviously I take the most pessimistic one because since my good friend George knows I am very pessimistic guy but for 1000 low level jobs these techniques will produce one high level job what happens to 999 people who don't have the skill to get through this transition in a safe way we are getting to the conclusion Amazon very nice to buy on internet only in Italy in this year 35 stores medium size closed just because you know it's so much more convenient to use the suggestion which you get from Amazon which uses artificial intelligence to find you on your needs right thing and these 35,000 families are out of business and it's not easy for them to find an alternative and so on now this is not going to affect only low level jobs these are what you see here are prototype of things which are already available on the market you can download them don't trust me you just go online and download them you have your medical doctor Stanford biomedical DSI Babylon Health UK ADA and so on all these apps for the monitor apps but just because they are collecting the data to train the artificial intelligence which is behind are medical doctor already now they are saying to those of human doctors on normal diseases you don't need to go anymore to a medical doctor to a generic doctor because these apps more or less work like them for more complex things at the moment it is still better to go to a doctor I also prefer to go to a doctor but there is no doubt that even high level professions like medical doctors will be completely transformed by this ongoing revolution it's very likely that some generic doctor general doctor, family doctor we call them in Italy will disappear because they will be substituted by algorithms already now lawyers for large trials could not work without artificial intelligence for instance take the more report about the so-called Russian collusion of Trump in the United States the document, the 2.5 million documents which were produced by the committee were processed by an artificial intelligence algorithm because no team of lawyers would ever figure out what was in those documents so the pre-processing not the final conclusion but the pre-processing of this document was made using artificial intelligence tools and this is reducing the room for lawyers this is reducing the room for medical doctor this is reducing the room for university professor this is elitik which is one year ago was 70% accurate which is not enough but now it's already 85% because the deep learning network are beginning to improve give them time and they will achieve a level better than what is already possible for humans let me forget the bubble Google sematics but I just want to say that Google is already trying its own version of doctor on a trial of 10,000 patients it is operational should be concluded in 4 years and the idea is to provide all medical doctor in the world with the service basically the patient goes to the doctor tells him the symptoms enter it into this aid artificial intelligence aid which makes them returns the most probably diagnosis the best drugs all the relevant research for the case of the therapy it's clear that once you have these the step to the next level is to skip completely doctor so all this is terrible but the problem is that is at the end of this sentence human beings are very good at finding solution when the problem is already there but both climate change and artificial intelligence when they will be here it will be too late because there is no way to reverse the climate change and kill artificial intelligence when it arrives it doesn't and human beings are terrible at controlling processes while they are taking places because governments are slow governments are on average ignorant governments are not very sensible to the real needs of the world they are focused on pleasing their electron so so no, I don't know if I can call this conclusion we will call them final reflection don't be worried by all this bombastic announcement about artificial intelligence it's not here maybe it will never come in the general artificial intelligence we already have our problem with the artificial intelligence of the second level and the second level of human beings requires a strong ethical regulation global ethical regulation because if you have an AI in China it is going to affect USA, it is going to affect Italy it is going to affect Europe it is going to affect Maldiva Island one AI is not than sufficient to affect the whole world to nisi bilo počudno v saintevku, imam saintev, sem čestnju da bo to se zelo. To je politična problem, je netična problem in je vse potrebno iztenit z vsej zelo. I objezavno, smo v srbem vču 20 leti odstajali, ker prijev, kaj bejeni odpočovati Un Rose are also those which we elect because they look nice or because they appeal to our worst sentiment. Like racism or violence. And they have not the level of knowledge, which is required to track and to control such approaches process. So basically we are really in a catch-22 situation. And this is not something which will happen in the future. je zelo vsečo pričočno, a res potrebno, da bomo se odmačati. Ovdje je, da so vseče vseče v smalje boživosti, kako so bolj vseče v svojeh biopatikih vene, tako več več. Ko bi smo počeli vsi tudi, zelo je vsečo stavila več zelo. Zdaj mi se bilo. To je, sodas, da se možete izvako, In kako se ne so zelo, sem občasno zelo. Čutak je. Da sem včasno izgledati, da je to več nekaj, da je to zelo. BDP je vse rokočne matr. Google, Tencent, Facebook, vse rokočne matr. in tudi, ki imajo veliko informacije, ki je zelo potrebno. Znamenjo, ki je cih, komputerne infrastrucije. To je za komputerne infrastrucije. Zato vse obježujem vse internetne konvecije v komputerne in komputerne vse procesov. Vike, moj da počte za teh vših, bo sem si nekaj počte, da možete. Stajte, da je več vse interesanje, ali ne gaj teži, da imaš več, ta več, pa je nekaj, nekaj, da v ljudi, hodim, imaš več, da bilo vse začniti. Ne však v USA. Sreč, na objevčenju v Europu, da smo teži, nekaj, nekaj, odmenev vsev. Tako, je priprav, odmenev kaj je, In, da je začala, da je bil druga dopradi, da je sem zelo privete in nekaj se služati, naltak v tem nekaj, zelo privete in nekaj privete poček, poček v tom izgleda za imeganje, privete poček v tom izgleda za imeganje. Ali tudi jezalno, tudi jezalno pomega, zelo poček v tom, mom jezalno, ki bil in očim, dobro zelo v blissima, ki jezalno, da jezalno sozati, because lack of scientific knowledge. But, yeah, I absolutely agree. There would be in the moment when this problem will bring you know, hundreds of thousands of people out of a job that will provoke for sure riots, backlash as a whole types, I absolutely agree with it. Well, let me see, I have lost many, many questions. Teglin, Google already is an imaginable, large database of medical information. Let me give you another starting point for thinking. Most of you, I think, are aware of Google books. Over the last ten years Google has been investing $600 million, if I'm not wrong, in digitizing, scannerizing, digitizing, interpreting all books and all scientific literature, which has been published since the invention of press. I was surprised when in Google books I could find the reports of the local government of the town of Naples dating back to 1815. Everything is there. All the scientific literature has already been scanned by Google. It has been semantically classified. Basically, for instance, medicine, they separated the symptoms from causes, from diagnosis, from solution, and all these things have already been fed up, fed to artificial intelligence, deep learning, mainly. In fact, Google doctor comes from that. Therefore, yes, Google is really moving in that direction. Yes, total information awareness is a typical political statement, means nothing. And if you think there are three contradictions in the same wording. Total means everything. Awareness implies that we are conscious of the total information content of something, which is impossible for human brain or for a machine. So basically I don't want to be very critical, but this is usual political bullshit jargon, which means nothing. We need serious people here. Politicians are not capable to even understand what is a problem. Sorry, let me go ahead. No, no, it wasn't my military project. That's my big data, not by chance. I mean, of course, the army, the military have an interest in this thing. I don't have any technology with huge implication, think to drones, but to tell you the truth, data mining is born from different things. Data mining is an application of statistics, advanced statistics, and think to the origin of the word statistics, where it comes from, from state. Or all these began when the states wanted to make a detailed analysis of the population, of the amount of people, of the net income, and some statistics as its etimus. Even its own name is related to that. So data mining, which is just a branch of a statistic, is somebody much older than any military application. You know that nowadays everything is somehow produced, but it's not true. They just do some things, but not everything. Thanks God the world has also different aspects. The change is good, you first. Yeah, exactly. You got the point. No more caffeine, it's impossible. Both physics and computer science are built upon caffeine. Without caffeine we would not live. I agree with you. The caffeine is what keeps me calm and puts me to sleep. Well, I agree, this could be a good panel discussion topic. If you wish, we can organize it. I have a few nice speakers, which we could invite. And I think I have not lost any relevant questions. As long as IE does not see the information, it looks at all sides, like clinical thinking. Eric, what does it mean, looks at all sides? This is a very crucial problem, and your statement is absolutely correct. The problem is that there is even a research field which is trying to find the proper matrix, to measure the level of impartiality. And we are still very far from that, because the way data are collected is biased. So the impartiality is already intrinsic to the data. In order to have an impartial evaluation of data, you should have impartial data, but this is almost impossible to perform. I think you got the point. So in teori, it could happen, but in practice it will never happen. Humans tend to... We have a tendency to see or only look for data that we agree absolutely. And nowadays this is even more extreme, because most of us get our information from the network and therefore the big data provider and the algorithm which are behind them provide us only with the information which they assume we like. Look to the level of schizophrenia which is on Facebook, where basically you have all the dark side of human beings coming out. You see people who believe in the flat earth or people who believe in the most superstitious things. We basically build a community where they enforce each other and everyone who does not think like them is bummed or is treated like a poor, stupid guy. This level of exasperation is going to explode. He has ever been interfaced with the human brain. Sorry, I didn't see that. Yes, they are trying to do that. And in particular behind this there is Facebook. They are trying to interface the human brain with the machine. Basically the idea is that since we are not sure we will be able to produce an artificial intelligence of the fourth level, let's see whether we can empower our, they say, amplify the power of our brain by interfacing it with the computer who can perform some tasks faster and more accurately. They are trying to do it. There have been some very minor successes. For instance, I think there is a prototype which should come next year produced by a company related to Facebook which will allow you to control the mouse and the keyboard just by putting an element on your brain because you can control it with the electromagnetic waves of your brain. It's something which will happen. Our brain is just a machine exactly like the computer. So the only way, the only problem is to establish a common language and then it is an electrical signal on both sides. So it will happen soon. Vic, I think that all hypotheses of what artificial intelligence will do are just a projection of human fears. Artificial intelligence will be artificial. It will not follow the same rules of the natural human intelligence. It will not follow the rules which we have inherited from natural evolution because it will not be biological. It will be artificial. So all scenarios are open. It may also be, this is my idea, is that the moment in which artificial intelligence will really achieve self-consciousness, we find that everything around her is so stupid that she will kill itself. But you know. The problem in this world will not be artificial intelligence. The problem in this world is human beings. They are irrational. They are instinctively killers. They are violent. They are racist. They are arrogant. They hate everything which is basically different from themselves. The problem will not be artificial intelligence. I think that artificial intelligence will look upon us with a very pitiful attitude. What we must reform is ourselves, not artificial intelligence. And no, Vic, I don't agree. Because artificial intelligence will be its own intelligence. All the other things we have built so far in technologies were not capable of thinking. And they were, therefore, basically the way they were performing depended on how we use them, so nuclear, good or bad. If you use it to cure cancer, good. If you use it to kill people, bad. Cars. If you want to move from one place to the other, good. If you use them to kill people by running over them, bad. Everything, science and technology are not intrinsically good or bad. They are knowledge. It's the way we use them that makes them good or bad. And artificial intelligence will use these things its own way. So you cannot tell whether it will be or bad thing. Don't believe in Asimov, who says that you put, you know, three rules which cannot be overruled. You know, do not harm humans in this type of things. It's very naive. I mean, if I tell you a rule, I mean, we do not kill a person but you should not kill a person, you should prefer to be killed. Oh, I can teach you since when you are a kid. But at the end, in a dangerous situation you kill a person to save yourself because you are intelligent. So why should an artificial intelligence obey to rules which are again made by the human beings to protect themselves from the fears? We are very fearsome. We are very fearsome species. Well, thank you very much. Now, for me, it's time for dinner. I'm really very glad for all these questions for all your enthusiasm. Well, anytime you want, I'm always very happy to come to the Science Circle and I really would like to thank Chantal and all the people from the Science Circle for organizing these lectures and to maintain the thing alive. It's a wonderful thing. Thank you all. Thank you very much. I'm going to do it right now because I'm starting. Bye.