 Rod olsun i amdanoion rhesnidau fineb yn awym blondio, a yna mae yn fan hogred i fel hwnerfodol, Beadwnadd i fi ei amddannu ein ffan Helwyr sydd yn cael presau gyda Nathan hon w shares a nidusanu'ch aleu'r cyfrwyngau yn ei dda yn gweithio. Wel mae son wedi gweithio i awr un要diad wedi gwld-wstbdd sef podemol i chi е�를i ar � Rydym yn hwnt i fusiblastelr, yr plan a'r proposes i gresnidau iddyn nhw bry i reflectiadau ohaid spellu yma. I'm with a major international IT firm. So John this is a case of one research institute flocking to one firm you might say. A series of three or four over the remainder this year. And this is one that I've called here the hybrid governance of cyberspace in it and actually in actual fact it's going to be something that Ryan and I will both be working on. I beg if I'm Nathan and I will both be working on. So here's some preambula words. First, very generally, I think we'd all agree that cyber space is an environment of opportunity, of cooperation and fulfilment, all these good things on the one hand, but it's also an environment for contention, predation and confrontation on the other. On the other, it's also extremely popular. According to the ITU, by the end of 2015 there could be something like 3.2 billion users of the internet around the world of a global population of something like 7 billion. That's the ITU's estimate. Forgive me for using these numbers again as someone said earlier where and where they get these figures from and who knows. The rate of internet adoption, I think we can all agree more solidly, has been very rapid. The ITU estimate that there were just of around about 400 million users of the internet in 2000. We'll listen to Eric Schmidt's prediction. He claimed that by the end of the current decade the entire global population will be online. A lot of people have thought that's a little bit over the top. Let's take a slightly more conservative estimate. Let's say just two thirds of humanity might be online by 2020. That rate of take up, if that's right, can only be described as momentous. 400 million to 5 billion or thereabouts in 20 years. A pretty good return on investment you might say. This is a momentous take up not just technologically but also I think more significantly in terms of politics. This expanding global communication infrastructure brings with it the possibility of what I've called here a worldwide conversation across all dimensions and at all levels of human life. Cultural, economic, religious, diplomatic, commercial, family, individual, non-governmental and governmental. You name it, it's in there somewhere. Communication and conversation are also, this is the flip side, made possible for criminal, terrorist and other adverse behaviours. All of that is of course very familiar to all of us. But in spite or maybe because of this rate of take up, I don't think that humanity has yet acquired sufficient familiarity. It goes back to this discussion of normal we were having last night, the sense of normality. We don't have sufficient familiarity with this relatively new form of communication and information transfer to know with enough confidence the identifiers of good and bad or acceptable and non-acceptable behaviour at any given moment. We're just not quite fully comfortable with it yet. I would argue that cyberspace is not yet susceptible to the forms of mature political organisation with which we're familiar and as a result it lies outside frameworks with which we've traditionally expected to manage and to moderate competition and conflict and even organise violence in the international system. So here's the problem as I see it. Cyberspace is neither the digital equivalent of the lawless Wild West on the one hand nor is it a closely controlled political arena in which the rules of the road are universally understood and implemented. It's neither of those two things. The supervision, organisation, regulation and even the governance of cyberspace are in transition. Very probably I suggest, this is my guess, towards a point somewhere between those two extremes, but as so often this middle ground is so much harder to describe and to occupy than either of the extremes. This is in a sense a large part of our challenge. I'd say that productive human interaction of any sort in any environment can't usually take place without a governance framework of some sort involving values rules, laws, norms, standards, codes and so on and so forth. Even non-productive or destructive human interaction often takes place within such a framework. The regulation of the conduct of war is a very useful and timely illustration. Governance frameworks in my view are not optional. They are inevitable. They're both necessary to the task at hand, whether it's regulating war or establishing protocols for commercial interaction as well as being fundamentally human. It is what we do insofar as we prefer order and predictability and fairness in our interactions. There's something intrinsically innately human about this. So I'd argue that the governance of cyberspace is best described as work in progress and it takes place on these two fronts. We get the so-called norm entrepreneurs on the one hand advocating the design and adoption of codes and standards as a means to socialise in a sense, socialise and stabilise cyberspace, either as a whole, the most ambitious or in discrete sectors, so in manufacturing industry, in the media, in research and innovation and so on. Very difficult questions have to be addressed in that process of norm entrepreneurship and norm transference. What precisely are the values and norms which should govern behaviours in cyberspace? Should these norms be generally applicable or should they be sectorally specific? Who or what should be responsible for developing and implementing these norms and under whose authority? As much as norms can be developed prescriptively or by design, if you like, they can also evolve spontaneously and organically through practice. This is again very human. This process of norm evolution, as I've called it here, can also see adaptive responses resulting from cross-pollination between different sectors. The research questions for this small project, the hybrid governance of cyberspace, are as follows. The governance of cyberspace could plausibly therefore be developed in at least one of two ways, which I've just described. Either this instrumental governance by design on the one hand or more evolutionary governance by practice on the other. Even so, as the technological sophistication of cyberspace continues to deepen and as cyberspace reaches further and further into ever more geographic and cultural environments, any form of governance I think is likely to face periodic and perhaps even chronic challenge as cyberspace itself continues to evolve. This thing is moving all the time, as we all know. In the context of these shifting circumstances, we're wondering in this project whether what we've called hybrid governance might have the advantage of being not only conceptually coherent, well, it's for you to judge, but also practically efficient and effective and durable. By allowing different approaches and ideas and influences and objectives in a sense to blend on demand, could a hybrid governance structure offer the necessary resilience, being able to adapt to change and to withstand stress, to adopt and to socialize new cultural and societal preferences and as a result, could it be more likely to be both durable and even global. Finally, I think we should also consider issues of implementation. While the normative basis of the governance of cyberspace is a very compelling political philosophical problem, of course it is, this is why I think we're also fascinated by this. It is also a practical matter of urgent global public policy concern. So there you have it. Don't over engineer the governance of cyberspace, but don't be too incremental and too complacent on the other hand either. Allow, what is the expression, 100 flowers to bloom. How are we going to go about this? We've got what we call around work packages. So WP1 is basically the paper that says this is where we are at the moment with the global cyber norms debate. It happens to be something I've worked on for several years with colleagues in Tallinn at the NATO CCD COE and so we'll be extending some of that work. Essentially a desk based review of where we are at the moment in the global norms debate. Then we've got a cyberspace governance workshop, which in many respects is going to be the most interesting part of this little project. We have a forecasting and backcasting workshop. Rand, if you know Rand, the Doctor Strange Love Thing, the Bland Corporation, et cetera, et cetera. Rand loves games and it loves numbers. In the forecasting phase of this workshop, we're going to draw upon contemporary circumstances, actual or likely policy choices and external trends and drivers and shocks, all of these things to describe three plausible future cyberspace governance scenarios. We've called the first intergovernmental, the second the transition, in other words it's still evolving, we don't really know where it's going, and the third the multi-stakeholder. So those are the two extremes again, the intergovernmental, the Chinese, the multi-stakeholder, the rest. In the backcasting phase, we then get our workshop participants to imagine themselves in one of the three scenarios from which they'll look back to identify these policy choices, external trends, drivers and shocks, et cetera, which in their view will have contributed most to the emergency, to the emergence of the respective governance scenario. And that's basically it in a nutshell. We have a fourth group as well, Group D, and their job is simply, they're the kind of exercise reference in a sense, and their job is simply to provide an independent forecast of the most likely state of global cyberspace governance in 2015, suggesting how norms and norm entrepreneurs might have contributed to this emergency. So once they've done that, then they're going to have a cup of tea, and they don't need to think backwards. So that's it. I'm a new college of law student. Given that the dark web is described as being many times larger than the normal internet, and I've been on the internet for many, many years, and I haven't been in touch with it, given that that's the case, isn't cyberspace rather than being one sort of space, isn't it? And this isn't about all those separate spaces that make incrementally much? Yes. I caution internet. We're talking here about the nice web. We're not talking about the dark web or the deep web, because by definition these are beyond the policy debate. I'm not trying to be glib and escape your challenge. I mean, I think it's a hugely important challenge. There's an awful lot of everything we're discussing today that is completely and utterly beyond our control, certainly our imagining possibly. Who knows what goes on there? How do you get into it? How do you, again, it's a power up. I mean, how can you possibly bring norms of governance and so on into areas which are by definition beyond governance? So I'm evading your question. I've got a question for you around telemedicine. From what I understand in this moment of the case, but from what I understand it was put forward, so telemedicine was a budget of international legal experts within relations between cyber and international war, international war theme. From what I understand it was put out, but there's been a pretty limited uptake and response to it in part because of the composition of the people involved in the new company part and the way in which it was put out. First of all, have you found that that's the case with the telemanual that there's been limited international response to it? And if that is the case, or even if it's not, what lessons can you take or have you taken from telemanual in terms of establishing these norms, top-down models, bottom-up models, et cetera? That's a great question. I mean, where we are with the telemanual, telemanual 1.0 is, as you said, essentially what it said was international humanitarian law applies. That's it. That's the big, and it goes through it in great detail. It's a very good read and very well structured and so on. The problem is, though, it was adopted by, I don't know how many, not adopted, read by, I don't know how many governments who thought all very good, well done, and it certainly has been adopted, I think, by the ICRC. So it's got a lot of credibility, but it hasn't been adopted by China, for example, who said, well, so what? This didn't go through the UN. What's it got to do with UN and so on? So there is a kind of a bit of a stumbling block there. Where they are at the moment is, they're trying to work through Talin 2.0, which will be, you'll know this very well in the sort of classic history of Geneva and Haig conventions that you get through the basics, and then you have to think about the more difficult cases, which are non-state warfare. So how do you apply international humanitarian law to non-state cyber warfare? I'm a huge challenge. I mean, where do you go with that? So that's what they're stuck on at the moment. I think they're aiming to produce 2.0 by the end of this year. So watch this space. What do I take from it most of all? I think what seems to me to be most obviously missing from 1.0 and probably from 2.0 is, in a sense, it goes back to the discussion you and I had just a few minutes ago after your excellent talk. It doesn't really, for me, get into the implementation enough. At the moment, I want something more that's just all about saying it applies. I want to know how and why. That seems to me to be what's actually distinctively challenging about cyberspace at the moment. Don't just sort of declare that goodness ought to reign in cyberspace. Show me how it can. So that's where I am at the moment. That's my problem with the Talin process at the moment. My question is in terms of what Roger's presentation was about. Wouldn't more governance to some degree create more vulcanisation to some degree? Because if you take away governance, you're actually left with, I suppose, a free-ranging internet. No rules, no laws. That's basically the basis of the whole invention of the internet. So the more governance you have, the more normative behaviour you're imposing or saying that's acceptable in society, doesn't that tend towards driving towards vulcanisation? Was it just my imagination? No, I think that's a really good question. I think if by governance what we had in mind was something closer to government, then I think I'd agree with you. Many of us here feel slightly nervous about big government. This would be the biggest possible ever government. I'd feel kind of wary about that too. But this isn't government. This is governance, which ought to be both a matter of governance by design and what I was saying, but also that kind of human bottom-up sort of thing. We like this. We like to have predictability and trust and all those sorts of good things. So not necessary is my answer. But one of the things I'm going to be talking about tonight really is trying to suggest that at the moment what we need to be doing is allowing both things to be happening. We need to be pursuing the global governance idea while at the same time allowing that there are these separated governance structures building up. They are there and there's nothing I can do about them to stop them or any of us here. They are there. So in a sense we have to accept the obvious that these two are very valid, the very good ideas that we're all agreeing with on the one hand and this rather diminished version on the other. It exists. So somehow we've got to get ourselves into the habit of thinking or allowing both of them to coexist for the time being until the liberal triumph happens. Yes, thanks very much. I think we might close there. In closing I'll remind you that Paul is speaking tonight, a public lecture on quantum sovereignty and the Westphalian principle. He's been extending some of these ideas, but you must come to the lecture to find out what he's reading and what he's reading. So I'd like to thank him. Hand over to Terry Bossamara, who's going to talk about tipping points. We're going from pulsi to physics. So I'm going to talk about a range of different projects all surrounding the idea of tipping points and particularly how these might have particular influence within the cyberspace domain. So as Roger's already indicated, we're talking very much about complex systems where we have very many interacting entities, agents of one sort or another, and the interactions between these various agents can produce interesting system level behaviours which you wouldn't likely predict from the behaviour of the individuals involved. So there's a huge amount of theoretical work in this area now. The things which are going to be particularly interesting to us in this talk are the ideas of networks and system level measures for behaviour of complex systems. And as Roger's talk illustrated, simulation is a key methodology for studying these sorts of problems. So we cover a whole range of things from the use of big data, very large data sets collected about people, the way people behave from supermarket records through to mobile phone metadata. We've been using these sorts of approaches to look at behaviour of people and aspects of human brain, and the particular bits that I'm going to select out today relate to tipping points. So Roger's already given a brief description of tipping points. We've looked at them over a range of different areas. Networks play a key role as well. So the connectivity distributions of networks, whether they're scale-free as Roger was talking about or whether they've got other characteristics, can very much determine the stability of these networks, which we'll come to when we talk about resilience of norms and how this might relate to resilience and evolution of cyber norms. The characteristic of many networks is that they may themselves embody tipping points or phase transitions, so certain properties of the network may lead you from one network state to another state, which is to some extent the sort of characteristic that Roger was talking about with balkanisation. So tipping points are sort of effectively big sudden changes. Tipping points are possibly made popular as a term by Mark and Gladwell in the book of the same name. It's a fairly loose term, and we're interested perhaps in the more specific term of phase transitions used in statistical physics, although if there's any statistical physics here, my terminology is going to be a bit slack at times. But there's quite a lot of different types of tipping points, which is a phase transition, which is important in identifying them in real-world systems. But what we can do is we can study what we call canonical models. These are really very simple models, and there's one of the most important things about many of these phase transitions is the concept of universality. So there's only a relatively small number of different types of transitions, and by studying the simple, simple models we can make use of this universality principle to see how these would apply to real-world systems. So, one of the things we need there, and one of the things which drives us, is a search for indicators of tipping points. And there's a variety of these doing the rounds at the moment. The first is, as you get close to some sort of tipping point, you get increased variance. So the stock market at the moment is displaying unusually a high level of variance at the beginning of this year. Whether or not this is indicative of what was already possibly in a bear market, whether it's going to crash seriously, and it's anybody's guess. Another idea, which is somewhat more recent, is the idea of critical slowing down. And I'll try and explain exactly how that appears a bit later on in the talk. But the idea of critical slowing down is that if you've got a system which is close to a tipping point, then if you give it a prod, if you kick it, it will take longer and longer and longer to recover. So a system which is well away from the point where it will recover very quickly, but as you get closer and closer to the point where you're going to go over the edge, it takes longer and longer to recover from a small perturbation. The smaller, the smaller, the smaller. Right near the end, we'll talk about some technical measures that are something called global transfer entropy. All the technical bit comes right near the end, so if you want to have an early lunch, this will be the opportunity to take it. So the particular type of transition we're interested in is something which we call a disorder, to the order transition, where you go from a steak which is essentially disordered, where lots of agents, lots of people are all doing their own thing, to one which is very ordered. So order can be good, so you can have, for example, coherent traffic, fox and birds form coherent entities which makes their flight more efficient. Some people in the neuroscience world suggest that consciousness is a property of kind of whole brain integration, where the whole brain is functioning in a coherent way. You can also be bad, so if we get all people in a stock market buying or selling at the same time on a property market where everybody buys, everybody sells, you get a bubble or a crash, in which case the order is bad. And if you take the consciousness as whole brain integration to the point where the brain is operating in a kind of pulsating rhythmical fashion, you have an epileptic fit. So order may be good or bad, but we're interested in the transition from disorder to order. And the key idea, the key idea of an intuitive level is that if you want to go from disorder to order, you need information flow between the elements of the system because the information flow is necessary to get the ordering to work. If you want any order to develop in any system, then essential agents need to synchronise with one another, so information has to flow. And so what we want to do is to be able to measure information flow in the most general possible way. So this is an example of work being done with colleagues in the University of Tasmania. They've done a lot of work studying tipic points in ecosystems, so this thing on the left is a giant lobster. I'm first of all this, I haven't realised that lobsters who get so big, but Tasmania lobsters are the only creature large and fierce enough to devide a thing on the right, which is a sea urchin. And if a sea urchin is left unchecked, it will eat away, completely render a coral reef, completely barren, it will just basically eat all the seaweed and you end up essentially rocks. And these seaweed forests are very important for marine ecosystems, lots of little fish and all kinds of other things live in that. And so the Tasmanian people have done a huge amount of work on this, showing that you can get effectively a tipping point in the density of the sea wind. So this is essentially sea wind, and this is a son of lobster, a sea urchin density, and you get this very sharp transition from lots of seaweed to no seaweed. There won't bother with very many more of the details of that, and that's actually what we call first order transition. Another problem we're interested in is the resilience, the stability of things like shopping malls. So this is a picture from somewhere in the United States where a shopping mall has just kind of eventually crashed and burnt, it's just kind of a shell now inhabited by rats on the birds. And what happens, or can happen, and in many ways, in many sorts of situations, particularly in regional areas we don't want this to happen, is where you get a positive feedback effect where some shops leave. It makes it less desirable for people to go to the mall, so fewer people go and therefore more shops close down, eventually no more. And since these provide entertainment and other jobs and everything else, this can actually be quite important. So this aspect of things like competition from online sales can potentially bring about this sort of transition to a loss of these major shopping centres. The work that we've been doing on information flyers done by the University of Sussex, and this is a photograph taken of Europeans' starlings over the old Brighton pier. And these birds collecting flocks are about 40,000. A static picture doesn't do justice to them because when you watch them at night you will see that these huge flocks produce the most amazing kind of flow patterns. They produce this amazingly diverse and synchronised behaviour without any central controls. Each starling is only interested in what its immediate neighbours are doing. And yet somehow the flock behaviour is large scale, as I said, in the case of these birds, about 40,000 to 50,000 birds. Now coming to sort of human agents, this shows a stock market crash. So on the top you've got the Dow Jones index, which has a dip here, and associated with that dip is a peak. Again we won't go into any of the details in the time available, but essentially what we have is a peak in a measure of the extent to which people all do the same thing. So you've got a stock market crash which basically means everybody is selling. So instead of some people wanting to buy this and sell that, everybody is doing the same thing, everybody is selling and the stock market goes pressure. So we want to measure that degree of cooperative behaviour amongst all the trading agents. And so this looks at this in a slightly different way in that now we've got 80, outward 130 equities, 128 actually, on the left, on the y axis and date going on, time axis going on, the x axis. And basically what this colour shows is the degree of cooperation between the equities themselves, so the actual shares, share prices show a very strong collaboration as you can see by this kind of bright red band in the middle. Now staying now within the neural domain, there's quite a lot of work going on now, showing the existence of tipping points in the human brain, psychology and human cognition. So we showed, for example, that there's a transition in the development expertise in the game of Go. Go is very fashionable at the moment. It was actually on the cover of Nature in the last couple of weeks because Google have just succeeded in developing a Go programme which has beaten the European Go champion. For a long time it goes just way, way, way too hard for computers, but it's made headline news in the last month. But there have been a series of articles which show, for example, the measurement of tipping points for depression. So when somebody falls into depression, in some cases this is not a gradual slide, but actually a tipping point where at one point they're depressed, or not depressed on the next point they are. Migraine has appeared quite recently, trying to look for neural signatures for the development of migraine, headaches, consciousness, we've talked about epilepsy we've mentioned already. And one of the issues that Roger and I have been talking about is whether or not these neurocognitive transitions underline the change from somebody who is maybe kind of disillusioned or unhappy to somebody who will commit acts of terrorism. Because there's a big change from being unhappy, maybe a petty criminal or whatever, to walking into a crowd of people with a suicide bomb. So in the case of Go we see a big change in the difference between amateur and professional players in terms of the way they look at the board. Social norms, which as Paul was saying are critically important in the cyberspace world, tend to be actually very resilient. And since IT is moving so fast, this resilience is particularly important. So a lot of work has been done by Peyton Young and Oxford on this, and there are examples of where medical treatment is determined not so much by the actual symptoms and the best treatment option, but by the actual local culture, medical culture corresponding to whether the age group, depending upon the age demographic. So to now get to the technical end, the sorts of systems we look at, simple systems are things like cellular automata where you have very simple systems with simple rules of behaviour which can be very complicated behaviour. And they show the same characteristics of tipping points of the real systems we've looked at with a close link to random graphs, which is in some ways similar to what Roger was talking about, but this is effectively the inverse of organisation, where you get what's called a tipping point, you get a connectivity avalanche in the graph, where in this collection of small sub-graphs, sub-networks, when you add the red connections, you add just a couple of connections and the system goes to be essentially fully connected. So the way we deal with all of this is through looking at statistics, so for example you can have statistics of cat colour and you can look at how one thing predicts another, so if you know that the cat's blue then you have almost no information about whether it's male or female, but if the cat's a tricolour like the one down on the left, you know it's female. So that's what we call mutual information, which is the, I was never intending to try and explain this formula, but the idea of mutual information is how much one thing tells you about something else, so how much the colour of a cat tells you about its gender. The more recent idea is chance to interpret, which measures information flow, and information flow is closely rated to mutual information, but it's looking at how one thing influences another. So we've then looked at one of the simple models called the IC model, which is a model of spins, which can either point up or down as shown here, and these spins show a remarkable phase transition where they will essentially be completely disordered or they will all line up. Again, skipping over lots and lots of details here, and then what we were able to show was that all the indicators that I talked about, a critical slide down the increased variance, they come to a maximum peak at the point of the phase transition. So I'm here. The red line shows the position of the transition, and in mutual information and the transfer entropy between each pair of spins peaks at the transition itself, but a new measure which will introduce the global transfer entropy peaks in a different place, it peaks on the disordered site where the green line is. So effectively, it's a predictor of the disorder to order transition. So the red line here is the transition, and the green line is here, which is the peak of the global transfer entropy. So we've got some new theoretical ideas. Testing these things for real data makes huge computational demands. You need large amounts of data to do this, but some of the new data sources which are becoming available now for studying social phenomena, for studying academics and so forth, give us a possibility of getting some predictive tools for predicting changes in individual behaviour or changes in cultural norms. Thank you, Terry. Thank you very much. I should mention that Terry is being a little too modest. His transfer entropy work has been published in one of the most prestigious physics journals, and is getting a lot of a buzz, because, as he modestly said, it's predictive of these sorts of changes. So the work that we can now do, looking at the way norms might change, looking at the way the internet may balkanise, particularly in networks, is for the first time we've got the possibility of getting to grips on prediction of those catastrophic and very hard to understand events. Predicting that there might be a way to occur with very high probability, and it's just never been possible before. So this is a real advance in our ability to handle this level of complexity for this level of real-world problems, to deal with the management and governance and control of the internet. So I'll take a few questions for Terry. Terry, I'm wondering if you've done models with the collapse of that shopping centre, to see what that looks like as a parallel to that. I'm not sure if you know about the two-degree housing development in the United States that was demolished in 1960s. It was basically a progressive thing. So it sounds like pretty much the same, so it cannot go into the supermarket. We've built simple models, similar to the ones that Roger was describing for balkanisation, to look at this collapse phenomenon. We've just collected data from 100 models around Australia to look at statistics of different retail categories. So we can start using that data to actually parameterise the simple models and then to take it towards, to look for the sorts of stresses or changes which would bring about conval collapse. So that's essentially work on going. Another question. Not that we've done. I don't know of that, but it's potentially one of the triggers that we're talking about for changes in internet structure, but that's not something I'm familiar with. Oh, yes, it would be extremely interesting. We're also having a look, as we speak, at the idea of changes of radicalisation of individuals through their texts on the internet that we might be able to use these techniques, these very advanced techniques, to predict the likelihood that this teenager in a basement will become radicalised. But before they become radicalised, and this would give police and social systems the ability to intervene for the radicalisation of individuals. So that's still a very early stage research project. Well, I'm thinking in terms of how things go viral, whether it's a tweet or a meme, and how then you get relational viral texts from, have you modelled some of those? Because I think that's a really interesting phenomenon in terms of shaping out all the sets of relationships. It's a very interesting problem. We haven't done very much work on that, but one of the things that we'd be doing in the study of shopping malls is precisely the issue of viral marketing and the extent to which the decisions people make about whether to use online services or bricks and mortar retailing are dependent upon essentially social media word of mouth. Well, we've got a PhD student who's just finishing, just writing up now, who's been studying ecotourism in one of the wetland areas on the New South Wales Victorian border. And one of the models that she developed looked at essentially viral marketing by essentially a word of mouth, social network marketing for developing ecotourism. So it's extremely interesting, and we've started to look at a couple of cases of it. And it can be really very effective. It can be very effective in terms of costs and very effective in terms of speed. In terms of the indicators and influences that you use, is there a way that you sort of distinguish between causal versus correlation? So you've got to be very, very careful about what's really causing that. This is just a correlation and you could end up going pretty long, Frank. This is much of deep and dangerous. Yes, well, one of the things of course you have to do is to condition out other influences. And one of the, this is kind of a very technical issue, but one of the things that essentially did to the idea of transparency of mutual information, is that mutual information measures effectively correlation plus all nonlinear effects. So it's effectively more powerful, more general measure for looking about whether one thing, whether the two things have got any shared information at once or at another. If you then want to look at whether one thing causes something else, you've got to make sure that you don't have a common cause. So, because a common cause could make it look as if two things were connected when it affects the common cause that's doing that. So, a model example is if you see a kind of a, your garden suddenly seems to have kind of more snakes and cats in it. You might think that there was something going on between snakes and cats. Whereas it might actually be that there's just been an increase in number of moths and both snakes and cats like moths. So, you then need to adopt a statistical procedure of conditioning out the effect of the snakes, at which point you see there's no behaviour, there's no link between cats and snakes. So, yeah, this is a very important issue for them. OK, I think we might close off there. And firstly, thank you all for coming today. We've had a ball. You've enjoyed it. It's been a potpourri of different sorts of talks. I might remind you that Paul, as I said, is talking, is doing a public lecture tonight. Fred Kate, the Vice Chancellor for Research at Indiana University, who hasn't arrived yet because of it. He's giving a talk tomorrow evening on cyber and the law, in a broader sense. And that's a public lecture. And John is giving a public lecture on Thursday, on the cross-domain deterrent. China and cyber security. China and cyber security. At the China and cyber security. At the China and the World Centre. So it's not here. It's co-hosted by China and the World Centre. So that'll be a good easy. And we're having some more research workshops that you might wish to attend. And one tomorrow, which is looking, we've heard Min from Stanton University, who was looking at, well, we're looking at how do you shape up a game-changing environment? How do you shape up a game-changing research agenda for cyber in general for Australia? And what contributions can Australia make to the global effort in cyber space research? And on Thursday, we're having a next steps workshop on where we might go with our strategy and statecraft in cyber space programme. So if you've got any interest in those, please contact us and we'll give you an invitation to attend. Thank you very much.