 many people are contributing their data to the internet. What are the implications of this to our society? Karl Umann will tell us now in more detail what these implications of this. He is a scientific journalist and he is a free journalist and he is a member of 2013. He won a prize for his blog post. The prize of Spectrum der Wissenschaft and we welcome him with a warm applause. Hi dear listeners, welcome to Calculated World. This is Tony and Klaus. Klaus here from the translator booth. We're looking forward to being here with you. If you have feedback for us translators, please tweet C3 translate, use the hashtag hashtag C3T or email us at hello at C3lingo.org. So hi, it's great that you're here, even though there's a space talk at the same time. But I also like digital things and I like to talk about the future. Who have you looked into their horoscope today? Well, there was one or two people that actually raised their hands. I wouldn't have thought that, but who knows someone who looked into their horoscope? We live in times where German healthcare systems pay for homeopathic medicines. So it's not as surprising as I thought. So who thinks that we can actually predict the future? There's 20, 30 people. This is my topic today. I want to talk about Oracle 2.0. Now first, the disclaimer. I'm a journalist. I'm not a coder. I really like reading scientific articles, but I'm especially interested in the call in what happens afterwards. So what kind of things does this mean for society? Now a second disclaimer, future, that's a really big word and prediction even more so. And to predict the future of society is not really something new. Now at a large scale, there's something like climate or technology predictions. There's also something like weather predictions that works quite well by now with data and models. There's one part of trying to predict the future that doesn't really work quite well, which is to try to predict lightning and these kind of events. Even meteorologists have to look at maps in real time in order to get an idea of that. And what I want to talk about today is kind of a big data that moves from like a couple days in the future or maybe a couple hours in the future. So we're looking at very short timeframes. Now let's look back. Oracles are probably as long as civilized society itself. And they've always played a big role. In general, humanity always wanted to look into the future, just to understand how the year proceeds when our fields flooded and stuff like that. So you kind of needed that even in the first high cultures of humanity, just to understand the seasons, natural progression. The antique Greeks then did this in a kind of different way. The Oracle of Delphi is probably the most prominent Greek oracle and it kind of made policy. So a lot of the kings of different states went to Delphi and asked Delphi. And the Oracle of Delphi really manipulated people. And so what you can really see is if you can predict the future, you can also change it. As I've said, I'm a journalist and I read in 2011 about it a little bit. I wrote something about the EU flagship initiative, which was a thing that the EU Commission gave out for a new big research project that should be very, very close to... That should be very close to predicting the future and it should be based in Europe. Six projects applied and one of them was Future ICT. Actually, but Future ICT didn't get the money and said it was graphene and the human brain project. So it's a lot of money for a research project. So I showed Future ICT and when this decision wasn't made, we thought of this behind this. It was pretty impressive to me. We have this exponential data mountain. So every year society produces the amount of data that actually has been there from the beginning of mankind to this previous year. So every year it's new again. And if we add this data to centers and ads that we developed. So humans contribute to this freely. And if we put all of this into a living Earth simulator and then we try to predict different aspects of human civilization. So we try not to simulate Earth as a whole, but we try to do different aspects that are interesting to us. And another thought of Future ICT is to do this publicly. So to have a public system that politicians can use both all of us. So it's not to suppress a minority, but it's completely transparent. So these were the implications that were suggested. And I think this sounds really good to know. So we're in a situation where nature catastrophes occur, like Fukushima and then there were technical difficulties. And then the Japanese government were coming into play. What decisions do I have to make? How do humans react to this like the Japanese society? And what can be what comes next after this? So how probable is it that certain things occur after this? So in these situations it would be very sensible to have these kind of predictions. There were other ideas like others. I found this really interesting. Like for instance, if the law is passed in parliament, is it really appropriate for what was the intended use? So it was really interesting to look into this and what implications it had on society and to simulate the implications and to have a short-term prediction for the future. Okay. Okay, so as I've said, in 2011 I researched some of that. And a year ago I restarted my endeavor. And I started reading into especially what happened in between. This basically mounted in a radio show that I did for the Deutschlandfunk. And I want to talk about what I presented back then. One example that I found, which is a little older and quite simple actually, is an Israeli coder who has a company and she developed an algorithm which is self-learning that analyzes New York Times archive of the last couple of years and is looking for correlations. So some things that happen in conjunction often and often. And then the same algorithm looks at current news and basically brings out some warnings. For example, in 2013 there were some droughts in Africa. A couple of years there were floods and then a couple of years after there was some cholera. And then they had a warning for Cuba where there hadn't been cholera since a hundred years. But in 2013 cholera actually reappeared. So you can see that this warning can actually help something. That was one small example. How something like this can start. In 2011 there was also the Arab Spring. And if you look for literature about Arab Spring social media, that was actually researched quite a lot. So that was a pretty good learning data set because social media was actually really important in contributing to people being able to meet up. And then in this data you can actually see how these protests developed spontaneously. A little more developed is another system developed by Virginia Czechs called Embers. Virginia Czech and some other universities. They're sponsored by the YARPA, which is a national security service funding organization. The results that you see here. For protests in Venezuela. You can see that where they predicted the protests to happen, they actually did exist. There are warnings, but those warnings are not published for obvious reasons. But they're actually given to U.S. secret services. Now, Embers looks at news and blogs, but also at geolocation data. That is geolocated Twitter and Facebook posts. And then they give out these warnings that like workers or business people would be protesting. Even before there were some warnings about protests. So the protests that they predict are statistically significant. And this tool is quite successful in using semantic parsing and then predicting the near future. Now we can think about this quite logically. There isn't any guru. We can kind of parse someone writing, well, I'll go I'll go protest tomorrow. Then I can parse protest tomorrow, which is quite easy and primitive. The nice thing about Embers is that U.S. universities and U.S. secret services, they actually publish their results, but they aren't very good at critically critically looking at that as well. Because they say, well, it's a good way for governments to prioritize citizen grievances. But of course, it could also be abused. So I think there isn't enough self-criticism in this. So this is kind of where we are right now until two or three years ago. First, you need unfiltered access to social networks. Of course, you also need to be able to check it with reality, to check your prognosis with reality. So after having made a prediction, you should always check whether what you predicted actually happened and then change your predictions for the future. And of course, if you do semantic parsing, you do need to change a different code, for example. So is this so? Do we need this? No, that's not exactly like this. This is a result from this year's summer, which was in science from the group of University of Miami. And they tried to look into extremism. So they wanted to look if Islamists from the Islamic state, how do they organize in social networks, because they are important and it's important if they organize globally. So this is difficult to do in Twitter and Facebook, because they are automatic filters and they're jurisdiction prohibitions. So this is recommended. So you don't have data from there to be analyzed. But they looked at contacts, they had few contacts. This is a big network in Russia and Islamistic posts are centered and they are not happy to have them, but the censorship is not up to date. And if you do this analysis of data pretty quickly, then you can look at it. So what they did, you have the y-axis and escalation parameters and they did count the actual posts within a certain time period. And how long did it take until it was censored and deleted? So they just checked activity. So what they found was if the activity is high, then you have real events like in Syria and you have the strong escalation and then you have the first offensive military action by the Islamic state after you visited this in the networks. So this is like within thermodynamics if you heat up a liquid and it gets hotter and hotter and then at a certain point like the vapor point, then you have a phase shift. So the liquid is getting hotter and hotter and then it's boiling and that's a certain point. And something like this you view also in social networks. So in theory this is nothing new, but now you can measure it. If you look at this curve and the event, then you can actually predict the actual time or the date of the event. So they also found this in this data. The other example is from Brazil. So this is basic research on social available data. It has become an own branch of physics, sociophysics. They look into self-organizing behavior in large crowds and the models are from thermodynamics, but they do not always work. You have to have some self-organization within the group like this happens often in social networks. So it depends if you have the head that says we have to become active now or if this happens on its own. And they can only make predictions if this happens on its own. But what the physicists say also, it's not like it's a pot of boiling water. It's bit different. So the states that you can have are actually different from solid liquid and vaporized and they are still within the beginning of this field of research. But we do have the data and we can say social research can be applied on the same level as normal physics. So much for the first part. We see collective behavior and we can analyze this data even if this is really big data sets and their collective actions and they follow rules that we can actually predict. Now, obviously, if I'm talking about the future, I should also do some extrapolation. I should also talk about tomorrow. I don't know who of you read 1984 this year again. I also did that. And also in this topic, I found a lot of parallels. Now what we found out is that collective behavior can be predicted in the short run. Of course, what does prediction mean? This is a graph from the German weather service. You see this point on the left, the point where we measured the data. And then in different kinds of conditions, we look at the near future and we see that these random effects are kind of deviating more and more. And then we have a couple of different futures possible. So of course, first in a positive way, the vision of future ICT for example. This was an advertisement from future ICT. You see a lot of researchers and politicians asking questions. What could this mean for the society which is very great for society? Society is complex. We know that we have kind of many crises that are overlapping. We have a lot of technology that stops working at one point and then we have these kind of cascades where a domino effect basically happens. So this is a good idea. But if we say, well, I don't want to predict the future of society because I don't want to manipulate it, instead I can use this kind of instrument to improve resiliency. What does resiliency mean? Well, resiliency of a complex system is that the system should be stable even if there are some sort of changes or interruptions. Now with Arab Spring, with the Arab Spring, we also said, well, political predictors, they really weren't able to predict the future. So maybe this is the end of the political experts. Do we even need political experts anymore? Maybe we need political experts to actually analyze the predictions. But a tool for the near future might be Voodoo for a politician because he won't be able to understand the algorithm behind it. There's an article by Sasha Lobo, which says that blindly believing in technology is difficult. And we should maybe make clear that predicting the future like this isn't as easy as some people might think. Maybe you remember a Google flu trends, which is a tool by Google that basically looked at how the flu spread or distributed. So they used Google search data that worked for a while, but at one point didn't work anymore. So even if these correlations work for a while, like with the Google algorithm, it might not work in the future and you might not see this difference. And secondly, what's even more important is that you might be able to manipulate. I can say, well, it's still early, but I really don't want a certain development. So instead of that, I will be doing something else. This is a quote from 1984, who controls the past, controls the future. But today, what Orwell couldn't see is that who controls that we can control the present already. Now, if we look at China, there's a paper from 2013, where a couple of social scientists from the US analyzed the Great Firewall, the Chinese Great Firewall, and how the Chinese government censors information from the outside, but also from the inside. And they looked at what exactly is censored. So we see there's like zero, there's almost no censorship, but on the right, 0.8, there's big censorship. And what they found was that being critical to the regime did not actually invite much censorship, but instead kind of any sort of assembly, any sort of collective action. That was the point where a lot of censorship was happening. And in the case of China in 2013, that was still a lot of manual work. But right now, that's probably more automated. What about with us? Like China is far away, we're here in Germany or in the Western world. Well, most of the data is actually in private hands, for example, in the social networks. And we see that there's a big public interest in making these kind of societal predictions. And there's a couple of articles. For example, I've shown that the bomb exists. It was cited a lot. And so I don't really have to say much about this. But you really have to look for the hints and look whether there's proof for whether there's social predictions. But if we look at Germany, we do see that there's a political pressure to social networks. What we call hate speech or sometimes fake news. And the question is, where does this lead to? Is election manipulation that might or might not have happened, have been from Russia in the US recently? Is that a new quality of life? Like choosing certain electorates in order to go somewhere? Obviously, that's not new. We've also seen that happen before. So I would be hesitant to point fingers. Now towards the end, it's kind of becomes a little bit vague. I have a couple of questions or thoughts towards the end. So I think social predictions, they're actually an instrument of power. If this is an algorithm for the security services, for US security services, that's a way to manipulate elections in certain countries. So it would be on the time to get the magic out of data predictions. It's a tool that we can use, but we have to be careful with it. And propaganda versus fake news. There where you start. So where do you start, actually? Many scientists researchers told me propaganda of the AFG or the CSU political parties in Germany. Is that propaganda bad? So my thoughts on the subject are these, what we see and what actually the nudging of these parties are actually getting, you get polarized on the topics. And maybe we should actually get out of these filter bubbles. So this for the political part of it in general, if we look into the future, this is a citation by George Manois and the Greek oracles. It doesn't count if the event actually occurs, but actually the action that the prediction predicts. So there's your prognosis, there's the future TM and there is what really will happen. Thank you for listening. Thank you for the interesting talk. Please for the people that are leaving us, please go to the right side. If there are questions, please go to the microphones. So I wanted to ask, do you know Project Cybersyn? You had the picture from Future City and this reminded me of Project Cybersyn. This was done in Chile and I wanted to create a socialism supported by the computer technology of the 70s. So this is just for you as a hint because this could be a historic reference to the subject. So not with digital means at the time. So at the time, no, they used telexter transfer data. Did you read the foundation cyclists by Asimov and how close did you feel in that? I was asked that right before the lecture, but I haven't read this cyclist before. Thanks. Thanks again. I don't think we have other time for more questions. Thanks again.