[00:00:06]
>> Through everybody thank you for being here in addition to the free food. So I will so I'm French I'm coming from French university and I'm a professor there on the information system for collaborative situation and I will talk about the where we find an intelligent framework mainly to walk on crisis management so we really focus on crisis management and we'll start by presenting the AI from walks it's a little obstruct Maybe but do not is the to ask if you have any question and then I will show you what we can be applied to credit management and I will assure you a demo of the system that we've built on top of this of this framework so it's everything is about work that we've done in the past in France OK it's not the world that we're doing right now Georgia Tech we've been one.

[00:01:02]
Of the physical Internet labs so let's let's let me take you with me in my world so with. Crisis management as the center of this work so 1st thing who attended the presentation the speech by Leo McGuinness last week. So OK so we started it started with. A question about what is it being smart remember and I will try to encode this question in a bit about intelligence I'm that comfortable with the world intelligence because it has several meanings.

[00:01:44]
Sometimes it's about information sometimes it's about being smart but we focus on the smart part of intelligence so there are a lot of different issues of what it is being smart what he's being what is intelligence and. What we have done we have tried to find a principle component the common parts of intelligence and basically there are 4 main things you can see are forming colors a pretty you can see the colors that there are 4 main parts for Main let's say principle component and this part or the 1st one is perspire.

[00:02:23]
To be smart you need to be able to collect signals from the outside from you see just from your environment for the situation you're facing and it's based on this. Perception that you will do something but the 1st component is about perception the 2nd one is about interpretation the way you connect the signal that you are gathered we've concepts we've I D's we've the past we've what you know what you've what you've already faced and so on so it's about contextualisation of the signal that you have received 1st Desoto one is about exploitation What would you do about deceit knows that you have contextualized with regards to your objectives it's the way you will you.

[00:03:15]
These into printed signals gotz to your objective that says that is basically about decision and the last about knowledge knowledge in terms of things that you already know abstract concept cases that you have been facing already and the way the knowledge can N.P. you of a better explanation or better interpretation and the way interpretation arcs protection can feed your knowledge so that's the 4 main 4 main concept that somehow emerge from the from the definitions of intelligence in the literature.

[00:03:57]
So we have a kind of definition that we use which is saying that intelligence is the ability to use your senses to understand of any situation you know that when act according to your objectives with regards to your knowledge and this is just it's just another definition and I don't claim it's the definition but it's another definition but the thing is that we use it to define a framework be fine to visualize it and we say OK there is data level and it's very important data level it's close to the signaled to the idea of perception then there is the.

[00:04:41]
Information level that's delivered where the data is interpreted with regards to your knowledge of concepts your experience and the context and finally there is the decision layer decision early years deliver where you will exploit your information you will do something and delve into all of these there is knowledge knowledge that you will use and basically does the finishing means that intelligence is about claiming this level.

[00:05:14]
Perception to collect data interpretation to transform the data into information exploitation to transform information into decision decision support with regards to knowledge that will help you for good perception knowledge will help you for good interpretation. You for good exploitation and knowledge will benefit from this process as well so this is this abstract from walk this graphic or from or try to use to perform artificial intelligence for more so this was about intelligence what about now artificial intelligence most of the time when we talk about artificial intelligence we talk about data science.

[00:06:00]
A lot of artificial intelligence is now connected to did a science it means that we try to go from data here to decision support that's basically the classical way of doing artificial intelligence Nobody's you collect data in a big data environment you have data management to do so and then you perform data analysis and that will help you to have classification pattern recognition visualization and things like that so it's hard to read but it's I hope you can see you still can.

[00:06:39]
So this is if I use the frame of the intelligence from look at the just presented in a map the current the current vision of artificial intelligence this is what we obtain artificial intelligence now with days it is this with regard to good intelligence from all of that just introduced.

[00:07:01]
But here we're not in an artificial intelligence. School it's good of interest surely engineering and if we consider industry engineering most of us are in industry engineering approach or dedicated to. Support decision as well and it's mainly based on models we all use model all the time and dissing is that it's quite different because we use the model of the information we build the models ourselves in order to support decision and for that we have modeling experts that no mater model instance base and so on and it will do modeling and then we can use up to musician simulation formal analysis and all of this to support decision at this level so again this is not covering the wall intelligence from all that I introduce you at the beginning but you can probably see me coming what I mean is that probably now what would be the benefit of trying to combine.

[00:08:15]
Data science and model driven engineering Can we would we be able to fill in the 3 layers and climb this attraction level from the bottom to to the top in a way that is fully that would be free useful. Actually data science is here and modeled and model driven engineering is here so we can see that there is no we'll wait to change shame them that easy to connect them there is no real connection because they are not reaching the level from the same level so let's let's talk a little bit about the the weakness and the limits of model driven engineering and science so if we consider the amount of data it's obvious that there are more and more data more and more different types and so on but.

[00:09:14]
What's the issue with this amount of data and that's really where it's important to understand that that's the main justification of what I'm trying to do it's this amount of data the saying is that if you have more and more a lot then the issue in that a science is that you can probably manage a lot of them but you can not manage a lot of different types of data a lot of different meanings of the a lot of different concepts behind the data it's more like you can deal with the volume you can do with the village city but you can not deal with the value in a very easy it just like a shark shock is a bill to detect a drop of blood in millions of gallons of water I mean I have never tried but I think.

[00:10:07]
And it's great it's great performance but if you try to give a Rubik's Cube to a shock it's out of its perception it will not be able to understand it it will probably apply a default to be if you attack escape ignore I don't know what's default behavior but it's outside of the perimeter of the data that you can manage it can manage a lot very precisely but in an amount that in a perimeter it is very small if you consider On the other end if you can see them other driven engineering the issue is more about the size the volume the the validity of data it's it's more like a baby if you baby a Rubik's cube it would probably sing that it looks like a toy and start playing with it find something useful to do with it on the other end if you drop of blood of kind of birth of the baby will not feel it you can do with blood but but it's more fun.

[00:11:14]
So the issue is that to manage this issue of. Big data we've either data science or mother driven injuring both bring brings you to then OK good results a lot of thing to do but it's not the optimal way you're stuck with the things that you can actually understand in data science not the volume to see that you can actually understand and you're stuck with the volume in the case of the middle of it because you need a human to manage these.

[00:11:53]
So. What's our proposal. And our proposal is this one is to try to have data and then he's not used to take the decision but used to build the motors and then use model driven injuring tools and induce an engine in tools to exploit the motors and it's a big change in the way we consider data analytics and data analysis now we're days which is more about jumping directly to deceive but here you can have abstract concept abstract.

[00:12:29]
May time of the LET you can use to understand. World that is more open that. It is right now so. How can we use this. From walk in crisis management oaken we use this data information decision and knowledge from walk we've got to crisis management so the questions are more about what perception can we have what into petition can we do what explication should be done and on based on which knowledge so we will consider this sions one of the juggler in the context of crisis management so.

[00:13:22]
First question what's perception and what interpretation. If you want to perform crisis management there are 3 main things that you need to know the 1st thing is the where you need to know the context you to know where is the crisis situation because you need to know if there is a road if there are buildings what facilities what populations what infrastructure what electricity you know what are networks and seeing that you need to know all the stakes and old assets that may be threatened by the current situation so that's the where.

[00:13:58]
If you want to know the where what we do you know where research is that we use Open Street Map You know that we're struck from the map so that's the data sources that we use extract from the map with a little treatment all the amenities build being people that should be considered in the context of this crisis.

[00:14:21]
That's interesting because you have a lot of things in opposition up so you just need to filter and you need to give sense to some of them for example for predation you need to know depending on the time depending on the building they may be people or not.

[00:14:34]
A 2nd thing that you need to know about a crazy situation is the who who can be used to to manage the situation what reason. So it's the question is more about is there for are do we have firefighters do we have police do we have. Emergency Medical Service and so on.

[00:15:00]
And we can we can new what Tuesday of what ability what kept abilities they can perform and this is this is one of the most complicated things to collect because basically their plans. I think you know today are plans for each crisis situation at a defined before any crises actually or cures and in this plans you can see the missions of each partners and what they should do when what the so basically it also includes what should everybody is able to do by.

[00:15:33]
A locating the missions to some partners you implicitly explain with able to do work but it's quite difficult to extract we can parse the plans and try to extract the data about who is doing what but that's wasted plans and interviews that the place where we still have to consider.

[00:15:53]
The human. Input. The sort of thing is there what of course it's crazy this is a different situation in each crisis there are different. Risks different damages different. Sets to deal with even so on so for this one we need to take care of sensors. Open Data social network and all of this to try to collect the data and try to interpret the data but want to stand what are the objectives that we try to we should try to reach.

[00:16:30]
The more interesting the most interesting it's how can we understand from the road coming what are the risks they mage's and finally objectives that we that we need to take care about. So this is the main they to Florida to try to manage so that's where we try to go from the day tally year to do informational a year and that's where we perform the interpretation here we have.

[00:17:02]
Rule based systems and machine learning they system that are trying to understand the data coming from mainly to spot and not the spot is just about rules but this part is about rules and machine learning to try to instantiate the concept of Freescale of the age of objectives of the teams and so on that are the scene that we have to deal with.

[00:17:27]
So that's the way we climb from data to information for the rest of the approach what we do with that we use planning enough to musician agrees I'm more the transformational Griese I'm ended that and the knowledge base that we have that we've built years after year in order to build the club What's your response and we the crowd of investments is a business process more the lead just a bunch of tasks saying you should do that you should do that it's a functional response that's the.

[00:18:00]
The answer to do is to face attrition and that's where we go from information to decision using knowledge that's the exploitation approach that's the way we prefer exploitation between information and decision so what do we have in this knowledge base just to give you a short vision of what is going to is this knowledge base our knowledge base is based on a myth some of that OK I don't want you to read it it's not the point is just I just want to show you the way it's organized with what should and what it contains.

[00:18:39]
So we have the what we call the core me time of the liquid time of the local things the concepts and relations between the concepts that describe any of course about the situation. The kind of credit she should reason of as partners partners we've a capability they try to reach objective together they try to perform activities following procedures using imputes.

[00:19:04]
And so on and so on that's generally called a bunch of situation but the interesting thing is that based on disk or me time within that describes concept we can expend it. On domains just as we do for us to learn when you have concepts and you try to refine them when you try to learn more and more in a domain for example in crisis management he'll you have 4.

[00:19:30]
Package around it the 1st one is about sorry I forgot to mention the Corys is structured with context concepts objective concept partners concept and B if your core Sept so if I move to the surrounding layers you have for Christmas when you have. Here the context on the part where you are of more precise concept for example the work is very important crisis management the buildings are very important because management so we have this concept that in your reach from do an Vironment component here same seeing here for partners here we have partners here somewhere here but here we have partners that are more precise for example increased management it's very important to know if a partner is on site It's if it's a if it's remotely working because that's not the same to you you don't have the same issue with the spotter So we have actors on site actors not on site same thing for the objectives objectives should be more increased management objectives more about fixing the mage's and preventing risk that's it that's always it's.

[00:20:40]
You have no more no other kind of objective so by defining this in your reaching concept you increase and you improve the size of the concepts that the system actually understand and at it will be able to instantiate based on the on the received data so what do we do then with this process that we deduce and I will show you in the demo.

[00:21:07]
Because it's only about collecting data creating a more than of the situation based on these data and then exploiting the small that to create a collaborative process. It's not only that because Christ manage money crazy question is unstable completely so you can collect data create a more that did use a process and 2 minutes after that the process is not it's not good anymore because something changed or because you got new information so we need to be able to build a new jail process so what we've done is that based on them on the model that we have in the situation or that we have we duplicate it actually so that we can have 2 models one that we call the expected model and one that we call the field model and a long time along to the crisis management both of them will be updated by different sources the expected model is the situation as we expect it to be so this one will be updated with regards to the way the process progress after each activity we say OK if this activity is done then we believe that the objective of the.

[00:22:20]
Beach has been achieved on the other end of field more than is the real situation so really we keep on collecting data coming from the field. And this data would update the field model and they may be different so all along the crisis management we have these both to small those of in their own life updated by different sources and we will constantly measure assess the distance between Bob this model to check if there is a difference between the models and if there is if there is a description then we know that it means that what happens on the field is actually too far from what happens what shooter it's just like you students when someone to look at you have one week to write a 20 page report you said OK one week 7 days 20 page 3 page a day I'm good but after 3 days when you have just written 4 pages you have a distance between the expected situation of 12 pages and the true situation of 3 pages so that you so you need to adapt you need to change your rule that was 3 page a day because you that will not that will not walk so that's exactly the same thing it was too much the difference is too big so we should change the process so what do we do we use the field model as an input to our optimization.

[00:23:47]
Agrees I'm so that we can build a new one. And we started again. So I want to show you that in a demo So basically what what's what's the more that what's the way we instantiate the. Artificial intelligence from want to show you in crisis management I told you that we have to climb the data information and decision layers to deal with the lifecycle of crisis management which is design in that maintain and so what we've done we've been through by here collecting data in order to infer a situation model you know that to deduce a collaborative process that can be validated that can be computerized that is and then orchestrated in order to be money toward make the compression update of the problem both models and in order to detect if there is a disruption that's basically decided.

[00:24:45]
According to the out efficient engine from what I presented to you and the life cycle of crisis management with regards to these 3 layers and the use of knowledge the in the knowledge part that's the meta model at a show you that is in this knowledge part and it's used to climb from here to here and then from here to here.

[00:25:06]
So what I will show you now in the demo is the suite of tools that we have developed to to support this approach so the suite of tools so the name of the team I mean short of research team and friends the name is into a publisher of organizations so it's all and so it is a start with I also we have the tools that we have so it's the truth.

[00:25:30]
So against with the 1st 2 days you're the is the DESIGN ASSISTANT. No no no related we have you know what it's to walk through assistance so this one is addicted to build the model wanted to get it to run to run the walk through then we have you thought a trucking assistant in charge of.

[00:25:50]
Monitoring the situation we also have you got it's the governance system that man has the knowledge then we have your play you play is the. User interface it's. Called information system that you use to money tore into full of the situation and then we have you seen the simulation environment that will stimulate the world system because there is no crisis every day of course fortunately at least not yet so we need fake data to stimulate and to simulate situation crazy so that the war Suites can switch can be can be tested there are other tools than this one for example of a file here is working on another one to me is what you're zippy So we have a lot of other 2 spot of the same street but the one for Christmas want of this one.

[00:26:42]
So let me show you the mo. So this demo is. Is based on the Lou offloading you off loading. Scenario the offloading is one of the main the 3 main 3 most dangerous crisis in France it's just like a big one in the US or that kind of thing it's we have 2 big revert at the same in Paris and the well even at.

[00:27:18]
The world country and so that's one that's the ruffling is one of the most feared. Crises So here we have a lot of different use cases in our in the troop So let's start with the you off through the NG So we have different too but we don't mind yet.

[00:27:38]
So. So we are in the design assistance and we will see different models let's start with the context more than the cortex more that will be empty now because we did not start the fake crises that we just show you the environment so this isn't a very much where you can have different you can you can see it as a map or as a model and then you can have different types of of concept that we really gather from.

[00:28:08]
It's made by interviews this one is a very simple one where you can see you have partners that's concept of partners their their capabilities here they could be also resource the resume that used to perform this capability by these partners and they can be also connection between the capabilities for example if you want to do this one then you need to do this one 1st or if you want to do this one you cannot of this one of the same time that the kind of ration ship of prerequisite Plus Rick receipt of forbidden relationship that you can have to think about it is for example you can not send only name in English but these planes that that's what are on the fire if you are firefighters and at the same time so it's not compatible.

[00:28:55]
And so that's the 2nd model the partner model. So it's a it's made through interviews and the so one is about objective it's the most interesting one so for this one so far it's simply because it would be collected when we start the crazies what their knowledge base look like I told you that we have this knowledge base which is this or graph that a base where we have concepts and instance of the situation model that are connected with existing concept or existing instance it's more about connecting existing instances with new instances.

[00:29:36]
So. Now we will show you so this is the you'll play this is the map we can focus were we want and if I want to to model the context of the place where we have an incident then I can use that what I'm doing here I use Sorry too late I'm using Open Street Map and it will gather and begin the year sort of again of all this small symbols that appear that are symbols that are relevant to my situation so that's the way I can collect information about the the situation no let's start the simulation so let's stop the fake crises.

[00:30:18]
But here you have the white curves reprise in the what are live in the water flow I think from the data that we've been collected from the it's real it's data collected from the French Ministry of ecology from the flooding provision service and they made some scenario and that's we've we've transformed the scenario into data and so we have.

[00:30:43]
And we have fake here we have one fake sense or that is in charge of sending this fake the that were system so that that there is a crisis and we will be able to see the weight evolves close in the way where we are and what is interesting and what you what I missed sorry is that.

[00:31:07]
I assume it will go over the threshold yes no assume as it's going over some threshold of some acceleration of some ruse of been activated in that case it's rude base so that's why we have this security here that is the danger area of the of the of the sense of the sense all say that we are over some through or of a somewhat a liver so there is a danger in this area and if there is danger in the Syria we need to we don't fly what are the stakes from the previous model that are actually in this area so that we can say are they impacted on not so is their risk so that's exactly what appeared here we have the Syria one week point that I want to say that it shouldn't be a 2nd it should be depending on the topology of the area but we are not that clever yet so we just put the 2nd where there is a danger and everything that is in the area of the security is impacted we have all the instances of consent that has been committed has been reached right from the Open Street Map are know that our nice area creates risks so we have all this risk that appeared and in our modern know now we have all these risks that are the results of this data flows at this point.

[00:32:39]
That's it that's the kind of data that that are sent by the system so now what do we have we have. The model of partners and their capabilities with their relationship we have the model of of context with buildings rolled and sing another has been extracted from Open Street Map and we have all the object of the risks and fact that has been.

[00:33:08]
Associated with objectives each risk is associated with one objective of preventing it if there were the major already there is no damage because just the beginning then we would have also objective associated with this the images so now we make the link with the. Knowledge base so we have each partner each Actually each instance of concept may be connected with existing concepts ensued into existing instance of concept into the knowledge base what is what is complete if we are several P.S.G. just working on this part because you need to have semantic and syntactical Grissom that say this new instance based on the name looks like this one and based on the on on through G.'s not based on the name but also based on the.

[00:33:58]
Bit concerned we try to suggest connections and then we have to validate it or not but the interesting thing is that if you say OK I have this risk connected to this one actually that never crossed want to close to a lot but to this one and then we have the subjective clues to this one and we have here beyond that we have a cup ability that is crucial to this one managed by the sector that is close to this one then we may suggest that to suggest to succeeding that objective we could use this capability of course it's not a $11.00 connection it's more complicated than that it's and connection that means that to reach several objectives you may have to reach several capabilities to use several capabilities.

[00:34:41]
But it's the basis of the recall solution agrees and that we have so now we'll use the. The did you shouldn't cross the dictionary or ism we have as you can see we are for the diction agrees and what does the diction agrees I'm dictionaries I'm as the model of context the model of partners the model of objectives all of this more those are connected to the knowledge base and what you try to do try to find which capabilities should be performed in which order depending on priorities depending on actors one actor cannot do several thing at the same time depending on the input output and so on.

[00:35:27]
And here we are for 143 that is more for road freight transport to force a play chain and one for crises they are basically different because they some are more on efficiency Some are more effectiveness some are more on using or not resource and so on so in that case we will use the crisis one but it can be it can be interesting and we do that sometimes to stew Cheik what different result using different for to so we use this one and then we'll have to manage a priority because of course the system cannot know if it's more important to take care of the injured people or if it's more important to fix the road so we have to classify drunk the the objectives at least the 1st ones to take your 4 1st one in wrong them that's what we will do now wrong them we slide them in position and then we thought we were indeed in action agrees them and we get that kind of process model where you have to ask and we run it we look at straight.

[00:36:34]
So now we are in the. Darkest ration and joints we deploy the process that's a military step. And we can we can read it and supervise it so let's run it so you see that we are in are we your what are you what now when you go because we will govern it so we have a different part here we have the To Do list of actors that will be informed about the tasks so OK so once this task is activity than I do partner will receive a message on it's on his own interface even it's presented here now but you can see it's your God for the partner hits your wife or the this in it's tougher talking we put 3 different.

[00:37:22]
Tools on the same screen so here you will have the To Do list if you accept the task and if you do the task then the progress the process will progress and at the same time will keep on more need to ring the expected model in the film order remember to models just like you report 20 days and so on 20 page so let's start it so I have a tasks I will accept it OK acceptance will do it and when it's done you will see change the Cohens we move to the next one so just a regular orchestration and join.

[00:37:57]
At the same time both of these models are updated one from the progress of the process and one from data coming from the field they taken from my fix it and they can accept and so on and so on and there is a task there is does that if a creation of the city of all evil.

[00:38:15]
And the scene is that this task has been lunch but we thought another fix and saw this fix and so is a road sensor in this routine so we're not sent us the amount of traffic and cars that was expected results one of a creation so there will be a difference so we will see that we were expecting an amount of 2 of cars and it's there is a difference between both there is no noticeable with a civilization of all and so there is a difference here you can see this it was not on this one so that's we can measure the distance exactly and so we get a value disavow you for is over the threshold that that's a question of the weight and the values of its concept but that's the result of this and so because this value is over the threshold then we will just use our model and try to deduce a new.

[00:39:16]
Business process that use Process same to force or to be kept the one increase management we prioritize and wrong the the risks in a new process will be deduced that we can start again I will go to my conclusion now. Before doing my conclusion I just want to say a few words about what we are doing now with.

[00:39:54]
This during the last months we're trying to use the exact same. Framework with date information knowledge decision and so on but in a different way in a different world we're trying to and some of you explained that already in a lecture but we're trying to collect the data about organizations in order to understand.

[00:40:18]
To what characteristic of the deception bill so that we can in fair risks and report. And the idea when the idea is fun is that we try to infer these risks and opportunity US forces. As mechanic forces that would push the system in one or another direction into from work of its K.P.A. some forces may increase some of you keep your eyes decrease some other and that's finally considering.

[00:40:50]
The risks and opportunities as victors and some are actually because that's a natural forces that is a new system so more potential It's a 4 that may or you know it a cure depending on your choice depending on what happens you can choose some you cannot But the interesting thing is that if you are if we are able to collect this data to create this model of information living where we have the forces then a decision lever we can say OK if we want to reach that area of our key P.-I framework maybe we should use this force is that this one of the good one to reach that point it's was it's not maybe abused but instead of trying absolutely to go straight forward to the objective maybe using forces that will make us having another trajectory is more is clever even more we can say OK we absolutely want to go there but we did not realize that it would be so easy to go there it's not the same objective but it's really objective that are far more achievable we regarding the forces that we have right now and so that's again the same framework of official intelligent with data information decision but in some sadness in another to me so I just wanted to tell you that because that's what we are doing and I think it's it's repricing.

[00:42:15]
As a conclusion I will I will take a bit. Usually I think bed just when I'm sure to win but this one I'm not but. When you know when I was a student so long time ago my major was artificial intelligence which was in the ninety's so you almost were not born but during these days it was it was interesting to do artificial intelligence but on the other hand it's been very very soon.

[00:42:50]
Not long time after that it's it was useless artificial intelligence was a dead. Domain in the beginning of that of 2000 I know you can it's hard to believe but it was the case no let's work all of this approach we just simply had no data to feed that kind of system and then 200-5010 Big Data everything was the everything was big data was everywhere and so people are bringing back brought back artificial intelligence and I have some colleagues who worked on not official intelligence in the in the eighty's they suddenly saw some of their old papers claiming very high in the situation index while there they were they were just saying that there were these people with it were dead.

[00:43:37]
And the thing is that my bet is that just like Big Data bring out efficient engines back on the stage just because it was useful to manage all this data I truly believe that now artificial intelligence to perform real intelligence and not only artificial reflects the artificial instinct that I believe nowadays e's artificial intelligence artificial intelligence is going straight forward from data to decision just just like if I put something shop on your home you will move but you will not sing about it it's straightforward data to decision.

[00:44:18]
So because we need we need artificial intelligence will actually be smart I believe in my bet is that model driven engineering and model based same in any case model we've abstract me Tom of those with abstract knowledge base all of these would be back under stage as well to move the artificial instinct and artificially reflects that is artificial didn't know where they were real artificial intelligence I don't know if it will or in any case not see each other in 10 days so we'll not be able to talk about that again but that's that's what I believe so I bring you back to your real life.

[00:45:01]
Just thank you very much for a listening and if you have any questions I just finished by mentioning that we've been what we are starting to walk on a common lab and that's the immersive response and it will club that's something that I hope will be. Successful in the next in the next years thank you very much.

[00:45:39]
Yes. It's made for any kind of collaborative situation for which we have enough instances in the knowledge base. The condition is the knowledge base. And partially some rules because we may need some of the diction rules that may not be generic enough depending on the on the domain so if you are in a domain that is completely unknown from the system that means the NOR Dimitar model in the layers nor the knowledge base are.

[00:46:25]
Regards to this domain then it will not work if you have some instance in the knowledge base even to me to model the general minimal there will be enough to do some things the key is the knowledge base and we work it on health care on supply chain on on a lot of other domains and.

[00:46:52]
Now it's all it's all made by a so we're in our team we have are research team research and junior team of 4 people and they are of wanted percent of the K. to develop the IP we have research or peer destroyed and researcher that are doing draft software some proof of concept tools with the language that you want with the technology that that they want them and then if something is interesting then we give it to this research team and they have to implement it again and maintain it in these big heels 3 tools that is.

[00:47:29]
That is stable. It's open source it's it's one of the present open source our goal is to find some. Industry partner that would say I want to sell this and I will fund your research for the next 10 years on this subject but I will make a product on this so that the French at least the French crazy says we move from Cypriot priest to recall age to middle age technology speaking.

[00:48:17]
Using it using these defer for so for the for the crisis management Paul The idea is the rent the people in charge of the root of the crisis and so it's more like. Local authority or government and then it's dedicated to manage the collaboration between partners so the part of where you have the tasks and the To Do list should be.

[00:48:44]
Managed by each partner of the collaboration the some troll thing is managed by the process and manager. Of. Index of tense and that kind of. That's that's. That's that's. We've been starting to work on this more than 15 years ago and very soon we've been working with firefighters and local authority and I said yeah it's great it's great screen.

[00:49:50]
And after that when you write when you when you see when you reach. The set at the really. The will responders then it's you're stuck because of 3 main reason they don't that they always say the same thing because 1st 1st of all you need to understand that in France most of the people that are doing crisis management are here because they they a technology they want to feel use food they are.

[00:50:25]
Very proud of they still stare on than things like that and there is a crisis time here it's my time now I can do sings leave me alone we manage it I mean I'm a firefighter and so but most of the time you have 3 they give you 3 main argument the 1st one is I don't want computer to take decision instead of me.

[00:50:48]
Which is from my point of view stupid because it doesn't take decision it just like ask. A big plane fight pilot to drive to pilot with no computer that would be fun not a long time but that would be 2nd argument is what if it's what if it breaks or what if there is any bug same it's stupid just like you for the firefighters in the 20th century said I don't want to take drugs I will keep my horses it's a they don't need they don't need an ever break and the soda one is.

[00:51:29]
We need to learn and we don't have time because crisis management I want to be efficient immediately and I don't want to search so or does it work or it's not like you fame stuck on Excel or something like that and stuck on something that that includes life something that there is no there is no way to do to understand it so what it shows is that this 3 main argument and it's the only argument that I always of these 3 main argument of just the proof that these guys.

[00:52:00]
Are afraid of the change because they have no real argument so to answer your question most of the time the decision maker they like that they said I want this well I want this in my crazy cell because they understand that the crazy cell are so poor technologically it's a mess there's nothing the just of a beamer they sure Google map and the 3rd series that seat and sometimes an exit sheets are they can notice what everybody has done and that's it it's in the US so they know that it's bad so the decision maker they want to change but the person they don't want to they don't want to change and it's the things that the industry is just waiting for that and say OK once they will be ready to buy something we will jump on it but some days there will be some day there will be the war footing or the same for the inference there will be a big big big mess with a lot of of the mages and after that we say OK now we need to give several millions of euro to buy a new saying and somebody will jump on it with no research no real product that's my and as a bit the answer the question thank you that's a that's of interesting question that's why you're so we are working on a lot of our research project and not only technology we always have social scientists working with us and makes most of the time they walk on the acceptance and the way we can run exist size of exhaust size and improve and climb the curving lot.

[00:53:37]
Of people so that they can. Agree on and like this kind of product. Yeah that's a very good point because in France we have a lot of exercise too just like in the United States as well so you have to do exist size if there is a if you have a sensitive any sort of site or if you have a big university there are exhausted that should be performed through to so that you can train and practice the seeing is that it's it's difficult to have practice every day every months with everybody so a virtually zation just just like you mention using the authority can be a way to improve the not only the quantity of science but also the quick and with connection to the use to the new use age because you can circle let's make decisions as with these tools resume or let's do it again with another tool and let's get you the feedback from the data of the tradition about the tool you will see that where people are stuck in this place and you can very good we hope that we can have you can have a very good improvement of the acceptance if you practice it virtually just like pilots have virtual training you can you can do that as well.

[00:55:28]
I think. Yeah I did not mention that at all but Crisis Management is about prevention preparation response and recovery that's the 4 usual step and. Prevention is really not about risk management 3 a lot about risk interview cation and in and walking on the risk before they actually OK or so if you want to walk on the on the risk you have several options either you remove the stakes or you protect takes the one of these impacted by the risk so you try to.

[00:56:13]
It's to prevent the triggering event to actually occur I say you try to remove the danger because that's a that's the $300.00 danger event stakes and if you have these $31.00 then you can you can have a and so our troops are not considering prevention at all that's another part of our team that is working on prevention but what we want ground risking to figure out and trying to.

[00:56:38]
Said force is it's really about prevention it's really about these 3 these 3 component that deal with the danger the stakes in the event and trying to see on which you can act because that that stream is very interesting because once you don't if I did go what is the danger what is the take what is the event then it's used it's often very interesting to see on which one you can act the faster the more efficient the more a sustainable e and so on and so on so I can just give you that answer but don't search the rig but your question is no distal is not connected with prevention at all.

[00:57:22]
Thank you. Or we do that we have we have these days we have 2 piers G.'s working on anticipations of suspicion and the objective is to what you said I mentioned to to observe the 2 models the expected model in the field model and the way we monitor them in real time but what we're trying to do now is we're going to project them in the future so that we can say OK there is no disruption for now but if we just keep the exact same process the exact same condition then we don't know yet but there will be a disruption in the future just by projecting the model in the future and also we can include some specific if what if it rains on what a what if this.

[00:58:21]
Reinforcement never come and then we can anticipate on the distance between them the model and so we can say OK you don't know yet but if everything goes that way there will be an issue in 6 hours if everything goes on perfect in 6 hours there will be an issue due to this or that the so that's close to what you suggest Thank you.

[00:58:47]
Thank you.