AI in Aviation
4:05PM Jul 7, 2020
Morning Good afternoon and good evening and welcome to the AI for Good always all all year, always online webinar. My name is Ayda Dabiri from the ITU the International Telecommunication Union, and I have the privilege of introducing today's webinar. He use United Nations specialized agency for ICT s, and we're also the organizers of the AI for Good Global Summit alongside the XPrize foundation in partnership with 36 UN sister agents. ACM and co convened with Switzerland. The goal of the summit is to identify practical applications of AI to advance the Sustainable Development Goals and scale those solutions for global impact. The AI Summit has gone digital with weekly online programming, allowing us to reach even more people all throughout 2020. We are pleased today to host this webinar on an aviation aviation with ICAO, the International Civil Civil Aviation Organization who is also one of the UN sister agencies participating in this year's program. We have a distinguished set of panelists today, but we're counting on you the participants to help create an engaging discussion. For this we will be using the q&a functionality which you can find just left of center at the bottom of your screen. In addition to the q&a, we will also be using the chat functionality. Please make sure to set the message recipient to all attendees and panelists, not just all panelists, you can select this option Just above the message box. Now before we turn to our panel, I would like to ask everyone to try out the chat functionality and let us know where you're coming from which country or which city. So I will type myself pressing the chat button. And I will send to all panelists and attendees, Geneva, Switzerland. So please feel free to send your message and tell them from our previous webinars. We've had people from all around the world we have Ukraine, France, Spain, Dubai, lots of different places. So that's great. We're happy to have you on board today. So without further ado, I would like to introduce today's facilitator, Mahmoud Natal jack who is the secretary of the airworthiness panel and a technical officer at the International Civil Aviation Organization. Over to you, Mona.
Hey, let me try to get this video on
There you go.
Okay, so that works very well.
So thank you so much either and the IT team and welcome everybody to our session on AI, meaning artificial intelligence in aviation. My name is maimana towel jack, technical officer and witness at the International Civil Aviation Organization, which I will abbreviate as I kayo. And I will be your moderator for the next hour or so. So this webinar is part of a series of the AI for Good Global Summit, which is virtual this year. And as I said, in the introduction from either through our this panel session, you will be able to interact, so feel free to ask questions to use the chat function ask as many questions as possible because we want this to be interactive. So we have a very distinguished panel this morning for you. Their bias can be found on the AI for Good Summit website. I will introduce them the way they will be presenting, the way they will be making their presentations, and why each and every one of them deserves a genuine introduction. In the interest of time, I'm not going to go into much detail. I will start by asking them where they are calling from, what they do, and what's their favorite dessert. I know my favorite dessert is chocolate cake. So let's start. I will start by introducing Paul. So Paul, say where you are calling from and what you do, and what is your favorite dessert.
Over to you, Paul.
Boom, thank you Mina. So my name is Phobos. I work for your control. I am the head of an operational division. We infrastructure in Europe. I my favorite dessert is for the team.
choice. So next I'll go to Marco Marco, please tell us where you are calling from what you do and what is your favorite dessert?
Yeah. Hi, my name is Michael Rickard. I'm the head of innovation for search technologies. And my favorite dessert is cheesecake.
Alright, then, Andrew, please over to you. Where are you calling from? What do you do? And what is your favorite place? Hi to
Hello, everyone. I'm Andy. I am calling from just south of London in the United Kingdom, and the chief solution officer for digital towers for NATS, which is the traffic service provider. In the UK We also provide services around the world I'm actually not a very sweet tooth kind of person. So I'd actually say pass on the desert go straight for a strong black coffee.
Good choice, I would say. As an Phillipe, my final panelists, please tell us where you're calling from what you do. And what's your favorite is it?
Yeah, so Hi, everyone. So my name is Joffrey Poli. I'm calling from Montreal, Canada. And I'm a corporate innovation manager here at santech. So we'll call it cover that a bit later. And yeah, my favorite dessert is pure frozen mangoes and just like a pure mangoes. awesome i
love mangoes too.
So we appreciate having all of you with us today and look forward to your presentations on the topic up for discussion. So during the presentation, you will be hearing people talk about the SDGs. So they are the Sustainable Development Goals from the UN. And there are 17 of them. You can check them out. online, and I will be able to also provide a link to them in the chat if you're interested. So, let's set the stage and we will and to do that we will play a short video so you know, please play the video. So, we are here at the meeting point between artificial intelligence, aviation and the Sustainable Development Goals and a lot of exciting things are happening. There are a lot of benefits we are seeing coming through AI and we have really good panelists today to excite you about the subject matter. So there will be two webinars, the one that is currently running right now another one in August of this year. information will be provided once the date is low. To get into the mode and to get you into the subject we have a short presentation people are innovating around The world there are tele medication, remote classroom educating the next generation that is happening. those in need have access to what they need. The supply chain is more resilient and responsive knowledge to make all of this possible is being spread. So internet connectivity that enables tele medication, education etc, is now being brought through the skies. Part of this is driven by innovation in aviation. AI has the potential to drive us further. And we want to imagine what we can do now that we have AI in aviation. Thank you
UN here to show us the lady.
All right, thank you so much Gino. So to tee up the conversation, I am delighted to invite our first speaker, Paul, who would make his presentation you know, for the benefit of the audience, and the presentation will explore the fly a report, which is a combined effort from Europe. So please go ahead, Paul for your presentation.
Thank you mimouna. Again, this time, I'm not going to be talking any further about the desserts. And I'm going to present to you the results of a report that indeed, we wrote all together. In Europe recently, maybe one or two more words about my organization. I work for your control, which is a European intergovernmental organization. And we're looking very much after supporting air traffic management. In Europe. We are also having some operational units with a central flow management function, as well as we're running an Air Traffic Control Center for three or four countries in Europe. You can go to the next slide. We over the period from September of 2019 to March of this year, we're running a what we call a high level group on European aviation artificial intelligence is perhaps a bit of a pretentious name this high level group But what we did do is that we put around the table all the best experts and key decision makers from the policy level from the European Commission, from the industry from Airbus from air traffic control organizations, from the airlines from IATA. But as well from the staff associations, beat em controllers, pilots, engineers, etc. We also included the defense people from NATO, European defence agency, etc. So we really thought we had the perfect mix of all the relevant players in European aviation and air traffic management. We wrote a report that we called fly AI and that very much concentrated on what is AI doing already today in air traffic management. We had two main objectives. One of it was to demystify artificial intelligence, because there's a lot of unknown elements are rounded fill, as well as to try to accelerate the uptake of artificial intelligence in aviation, because we believe it holds a lot of promises and a lot of improvements and performance benefits for the whole of European aviation. The report was published on the fifth of March, literally a week before in Europe, we went into the big shutdown. We had planned many rollout events, which unfortunately couldn't have happened. So if you're looking for the report, first of all, at the bottom of this slide, you will see a link where you can pick up the report, but I also still have a few hundred of these paper copies that I'm happy to share with anybody who has an interest. Next slide, please. We stated we didn't start from scratch and we very much based our work and our report on the further work that's happening in the European Union on artificial intelligence with regardless of any domain. This is called the high level expert group on Artificial Intelligence, which published its reports over the course of last year, and they very much breakdown their recommendations over policy regulation, capacity building and partnering. And you will see in our report that we have followed that same exact structure. Next slide, please. In the report, which I really ask everybody to download and have a full read of we, first of all trying to demystify artificial intelligence and that's why we call this AI is already in the sky. And we're giving you an inventory of all the applications that we found BM e Dugan exploration and r&d, or already in deployments, we are finding in air traffic control centers, for example, already applications that are doing on a day to day business traffic prediction based on AI algorithms. We are also for doing for example, the monitoring of communication navigation and surveys data that needs a lot of attention in Europe. We're also So finding applications on runway occupancy, but I colleagues from Heathrow will further detail detail that later in the presentation. And in the middle of this slide, you see all the benefits, all the areas where we think that artificial intelligence is already today or maybe already tomorrow, providing benefits to the greater good of aviation. And as you see it built into many different areas going as well into resource management workloads, automation, infrastructure, managing and even going to elements like spacing, separation, etc, although that will not be for immediate Yes. Next slide, please.
So, the core of the document itself, is what we what we try to do with the exploration of artificial intelligence, and we come up with seven steps that very urgently, we think need further attention. The first step being that we need more sharing of data and a common in artificial intelligence, infrastructure framework. There's a whole plethora of tools and methods and processes. And we think that we can achieve a lot by sharing better our best practices and lessons learned. There's a big task to be done on research and innovation. But as I said, as well, a lot of AI isn't very ready to be implemented operationally, or is already implemented, implemented. And for that, we think we need further work on the tools and methods that provide us validation of artificial intelligence run algorithms, and terms of deployment. There are some areas where we think we urgently need to look at, for example, the cybersecurity domain. cybersecurity doesn't stop at the border of aviation. And we really must make sure that we use the latest AI techniques to discover to discover the threats on our area. We also recommend that in terms of deployment, we start with the safe the non safety critical applications and we do the more delicate safety critical applications at a later stage in terms of The recommendations five and six, we also think that a lot of work has to be done in terms of communication, dissemination, dissemination training, and that training go from top to bottom. We are training our senior managers, we are training all the staff, we are training our operational people, we really involve everybody we can think of, in demystifying what artificial intelligence is, and how it can benefit aviation. And the last recommendation is probably the most important one is that we really should be working as much as possible in partnership. You don't do this on your own in a corner. It is something where we can really learn live and learn a lot from each other. And so we have a lot of practical views and recommendations on how to set up such partnerships. Again, I invite all of you to download the free report. The link is again at the bottom of this slide. And I'm willing to answer any questions you may have. Thank you very much.
Thank you so much, Paul. And I think that with representation we are beginning to see the efforts that are being made by Europe to admit to demystify and accelerate the use of AI in aviation. I think it's very interesting. I think it was very interesting to see the emphasis on the importance of good data and infrastructure as well as special as well as the cybersecurity considerations that have been made. It is all very important for proper application of AI. So with that, thank you so much, Paul. And I will now hand over to our next panelist Marco Marcos presentation is on AI for air traffic management. He will talk about how AI has changed the way we develop systems from the engineering perspective and talk to a few use cases of AI for ATM Miko over to you.
Thank you very much for the kind introduction, my Muna and as I said before, I'm the head of innovation for students technologies. We're leading technology provider for navigation service for lighters and airports around the world. We focus mainly on digital and remote towers, but also on automation and decision support for air traffic controllers and airport operators. Next slide, please. And the first point that I'm trying to make is that AI is a very accessible technology and it also enables other technologies to be more accessible. And what I mean by that is that the problems that we previously were not able to solve as engineers in our specific fields are now possible to be solved easier using AI and the underlying technologies such as machine learning. On the slide, I'm showing our very first video segmentation or video tracking technology circa 2010 on the left, and our latest AI technology on the right, and it took us about 10 years to get to the point on the left was a lot of effort, taking optics, PhDs and a lot of system experts to design this algorithm carefully looking at the pixel changes and coming up with this with this algorithm. On the right you can see the AI and always To do for the AI is built on technologies that other academics and other technology companies have already built, taking it and adapting into the aviation industry, showing examples of different airplanes and being able to deploy this operationally. So the point I'm trying to make is that you no longer have to have a big team of technical experts in order to solve problems, but you can benefit from from other technology providers and academics alike. And the good thing about AI also is that it is by design most resilient so that you no longer have to worry about designing the algorithm for different scenarios by showing the AI different examples of different lighting conditions, different aircraft, we're making the AI resilient by design for these noisy inputs. So overall, the the AI really has a potential to speed up the system development and allow people who have traditionally not been able to access certain technology access to the technology now. Next slide, please.
One of the most important things when we deploy these AI Solutions operationally is ensuring that AI is safe. And that is from an overall system perspective and an end to end solution. But also specifically when it comes to AI, being able to understand what is the AI work, and more importantly, where does it not work? What are the failure modes, how does it fail, and how to ensure that the AI if it fails, does fail safely. So what you can see on the slide here is we use the technique to understand what our AI here is interpreting as an aircraft. So you can see certain sections of the aircraft were blacked out. And in the top right corner of the picture, you can see the confidence score of the AI saying that there is an aircraft in the scene. And so we can do this in order to understand what the AI is actually detecting in order to determine this an aircraft here. And what you can see from this slide is that the fuselage is, for example, much more important than the wings. And the point of this is really understanding how to maybe make the AI explainable, get everybody comfortable that AI is safe, and also understanding that we don't have Pisces In our data, so that for example, we only show Lufthansa aircraft and now all the sudden that doesn't work for Turkish Airlines. And so those are the things that we can understand by making running these tests and really explaining AI and how it works. Next slide, please. So now I'm going to go through a few use cases of how we have developed this AI into operational solutions. And the first one that I want to highlight is our shining star example of our cooperation with with NASA. And it's the London Heathrow digital power laboratory. And you can see on the slide that is the overall human machine interface of our system. You can see we have panoramic cameras in the top with a map display and then in the bottom left corner, you can see our AI is monitoring the runway exits. So the tower and Heathrow is built almost 100 meters tall. And what happens sometimes is that the cloud cover comes low enough to block the controller's view off the ground. And but the ground still has visibility, so everybody on the ground still has good visibility. operating in visibility conditions. So what happens then we go into this to condition. And the air traffic controllers now has to rely on the ground surveillance system in order to clear the next aircraft to land to ensure that the previous aircraft has exited the runway. So now we have to go from a very high fidelity surveillance versus the human eye to the ground surveillance, which means that we have to put a safety buffer in there and increase the spacing on final approach. This means we're losing about 20% land capacity and as a busy airport as Heathrow is, you can understand that that is a lot of movements. It's a lot of money that is lost, but it's also a lot of fuel that's burned by sending these aircrafts to other airports or putting them into a holding pattern. Next slide please. So as you can see here to solve this problem, we have installed two cameras on two exits of the runway in order to try our technology and you can see that the blue mass begins covering the aircraft that's all AI detecting the aircraft and what we can do now is visually confirm with this AI technique. When the aircraft tail has left the runway, the white runway marker and then again a safety zone, indicating to the controller as seen in the bottom in the aircraft tag that it has now vacated the runway. And we can then indicate to the controller that is now appropriate to clear the next aircraft for the landing. So what we have done now is we've brought again, very high fidelity source, which is the visual camera technology with our AI where we can on a pixel level determine when the aircraft has to make it to the runway. And what we are hoping to do is then bring back that 20% landing capacity. And again, get the airport back to full capacity. We've tried this over for three months and now building the safety case together with NASA rolls out operationally in Heathrow. Next slide please. And the next example is from Dubai, where you can see that on the right we're tracking not only the aircraft, but all the service vehicles around the aircraft as well in order to automatically track the turnaround process of the aircraft. The aim here is to accurately We track certain times such as aircraft arriving, bags coming off catering going on in order to give information to all the relevant stakeholders to plan the turnaround process appropriately. So, that would be airport operators being able to dispense resources such as catering trucks at the correct time. But what we also do is based on historic data, give an early warning when things are not running as smooth as it should be. So that for example, if we can tell an aircraft is arriving late and what the knock on effects of that are, we can again, as an as an entire system overall between ATC and airport operator to spend the resources appropriately to use all the all the capacity is so rare, the memory capacity to the fullest that we can next likely. The last example that I'm going to show here is our ground surveillance, AI. So what we did here is we analyze the years worth of data in order to understand what the movement patterns are on the surface of this specific airport. And what we then went able to do is predict accurately what the trajectory of an aircraft is, while it is moving in real time. So you can see on the slide here we are accurately predicting the the runway exits that this aircraft is going to take. What this allows us to do is accurately predict the runway occupancy time. Again, what we can then do is use the runway capacity to its fullest. If we can accurately predict the exit and in arriving aircraft is going to take, we can then decide if we can squeeze in a departure in between two arriving aircraft or can we again, squeeze the final approach spacing.
And I think that that's it and I will be happy to answer any questions later.
Thank you so much, Michael, for that brilliant presentation. And it's interesting to see how the use of AI could improve operations, especially in low visibility situations and to improve efficiency and predictability. So that's very interesting. So thank you very much for that. The next speaker is Andrew. Andrew will is with Andrews presentation will provide information on the optimal solutions for airport air traffic control. So over to you, Andrew
is my man.
So there will be an element of some of what I cover which my colleague Marco from Sirius technologies has spoken about. But I think that's that's quite relevant, because it's a key part of where we focused. Well, what I want to do really is look at this from the point of view of how AI can really focus within traffic drone operations, specifically, those that are ports to optimize what what air traffic control is there to do. I'll try and make this as jargon free as possible. Should anything be needing a little bit more of a clarification by all means we can come up during key questions. But if we can move on to the first slide I've included this one because it's 100 years since air traffic control was born. This is London's first airport back in 1920. In South London, in Croydon, and you can see front and center there. Air Traffic Control is a key part of the operation 100 years ago in 1920 so it doesn't perhaps look quite as technically advanced as maybe you you would expect when arriving at an airport. But certainly in the 1920s this this was the site to to meet you as you arrived at Croydon now return. If we move on to the next slide. This is basically where we are today. Certainly for London and this is our main airport, Heathrow 2020. So 100 years on, but the reason I've included this slide In particular, is because while things are significantly more modern, and you can see that our shed has been somewhat improved, in that it's now a five storey glass building at the top of an eight seven meter high stalk, very technological in the architecture and also in some of the systems that are inside of it. But effectively the concept of over of operation from 100 years ago, an operating Croydon airfield from the shed that I showed you previously, to move into the control towers that we know and love very much these days. The actual concept of operation has not changed significantly. So you can see that it is dominated by the ability to look out across the airfield, which is obviously significantly larger than back in the 1920s to accommodate significantly larger aircraft and to larger numbers, and also large numbers of passengers, but the underlying concept is basically a visual one to look out and scan 360 degrees. So for an airfield, which may expand four kilometers or more in in many more all directions, and also the skies immediately around there. And you can see that controllers still rely today on certain levels of technology that would have been recognizable by our forefathers in the 1920s. And you can see a controller there today holding a pair of binoculars to look out of the window because of the size of the airfield that was today. So there's a significant similarity in operations for the last hundred years. With with technology supporting and AI now is really the cusp that's going to significantly change our operation. So if you can move to the next slide, this is what I want to just start talking people through use cases. So I've included a few graphics here. And this is the use case that Marco was discussing before. But basically, you can see on the left hand side of the picture, I've numbered a couple of graphics there. One is of a system which, which wasn't available in the 1920s. But it's quite common in major airport towers these days, which is the ground surveillance system, a radar based system using multiple iteration and also a primary radar, which provides us with a blip. So if you're looking at image number one, you can certainly see very close to that image. There is a horizontal black line which represents the runway on the airfield. And you can see a yellow blip which is sort of aircrafts shaped and that comes from our primary radar, and then that's overlaid with a label telling us that it's a British Airways flight and where it's headed to which is standard 561
The Key of that that particular system is to enable the controller to visually monitor the outside operation of the airfields when they don't have the benefit of the view that you can see an image number two, which is of this system sitting in the foreground, and then beyond that, you see the window out of the view out of the window view. And in fact, you can see if you look very carefully there, there is an aircraft just underneath the number two, and that is actually on the runway, which is depicted in the larger image one. So, primarily the controllers as per the 1920s will look out the window and will observe the aircraft to clear the runway, and then ensure that the next aircraft can safely use that runway. So, visual observations are key. But you can imagine that when those visual observations become difficult because of reduced visibility, low cloud in particular, which is common with today's modern tall control towers, We then have to rely on the T system above. Now, the one thing that I will say when you look at the system above and you look at the aircraft, in image one, you can see that it has a tail shape, you can see which direction is roughly pointing in. And you can see that it has wings, although you can definitely tell from that radar image that in fact, the whole of the aircraft is not entirely certain in its position is provided by the system. So it's tail crossing the little white box that you see an exiting the runway safely, is why we apply an additional margin when we use a controller observing a system rather than observing directly out of the window, because the level of fidelity of what they're observing particularly the aircraft is is lower. So we increase our spacing between successive arriving aircraft and that reduces our landing rates. This is common at airports all around the world. It's not specifically a A UK or London Heathrow problem, but it is particularly an issue at airports that are busy and capacity constrained as as Heathrow certainly was prior to the current
restrictions in in traffic due to the global pandemic. But, but that aside, you can see how the impact of the system limitation then brings in a procedural mitigation, which reduces flows and potentially has impacts including those on fuel burn that that marker mentioned previously, and delays. So the system on the right hand side shows how cameras surveillance can basically provide us with an image and there's an early image in fact, he's in one of his previous AI models, tracking an aircraft but you can see how that aircraft can be tracked in industry, right down to the last pixel of the little tailfin. And in fact, the elevator that's on the right hand side nearest to us as the aircraft clears the runway. So that's providing us with the kind of tracking that a controller would be able to provide on the best of days. But actually, the one thing that it can provide us with beyond what what even nice they can give us from a control tower is the ability to see close up because those cameras are distributed around the airfield. And the AI is able to track that aircraft while it's tracking another aircraft, which may well be behind the controllers field of view. So a controller normally sitting within the control tower has to move their head and look at various parts of the field as well as looking down on operational systems. So that that involves them for a human element and a scan that has to be built into the way that we operate. So effectively, if you can now bring artificial intelligence in such that it can monitor all areas in high fidelity in all conditions, that then becomes potentially a significant use case. If we move on to the next slide that shows how we've deployed that use case at Heathrow, so Marco referred to a laboratory at Heathrow. This is in the base of the control tower, you can see the image there with, in fact one of our operational controllers from Heathrow coming down from their operational position in the cab way above us, and coming down to operate on the system that we were trialing future operation within the control tower using all digital, all video, and integrated systems, which we brought together so far in a two phase environment. So the first one was actually establishing the laboratory phase one, putting out the cameras for the artificial intelligence to be able to use along with the existing systems, and that's also involved an integration with those existing systems. Including the radar based one that I mentioned already, because this very much is a case of taking all available technology, not necessarily simply replacing what we currently have, or replacing, specifically humans, this is about improving the whole operation. So the second phase is all about a proof of concept for those enhanced operations, and then involve training the model. And then also, the key to us from an implementation perspective is then doing validation of that with live data. So the controllers, such as the one you see in this picture, would have operated life and then come straight down into the laboratory and seen exactly the same situation, the same weather conditions unfolding in front of them, and work with us and the software engineers on the model training, but the model itself was allowed to run independently. And we have assessed more than 40,000 flights with regards to How well it's performing and if we just advance on to the next part of this slide. So if we can just go to the next slide said, our results basically are showing us the elite the artificial intelligence engine from cirrage. And the software EA vd, which, which Amy runs with,
has been confirmed from a validation that the cloud cover the cloud height, aircraft type, the size of the aircraft, we call it its weight category, but basically the size and weight of the aircraft and the airline operator something that that Marco referred to earlier, has no effect on the impact on the impact of the system. It is effective in all those conditions. So regardless of the patterns on the aircraft, whether it's painted in the standard operator's color scheme, or whether in fact it helps as an advertising logo on it temporarily. The system can deal with all of these artificial intelligence can so that's a key part of proving that the system is is resilient for our operation. We've also looked at it from the point of view of fusing it with other information, the radar data. And we're now underway on the second phase of testing, where we're looking at because of its its capabilities already, we're now looking at using it in much more difficult scenarios where because of fog, the visibility as reduced below 500 meters. And that's the point at which pilots themselves have to rely on systems such as the category two, three auto land facilities of of the aircraft. So we certainly now looking at extending our business case and using it in a lot of further impacted scenarios. Another key thing for us is to prove that in heavy rain, strong winds blowing snow. Well, that goes directly onto the camera lenses, that by use of fuse data, we're able to show that the system mitigates those and again, can be really Upon entirely by the controller, if we move on to the next slide, this effectively is just a graphic which is explaining overall the concepts, you can see our tower in cloud. And you can see that basically we're saying that the cameras are able to process the information faster than the radar update radar update is one update per second, whereas the cameras are updating at, in our case around 25 times per second. So the system is more faster with this update rate, but it's also higher fidelity, as we've spoken about with regards to direct pixels of the aircraft that are being observed by the system. And and the visual elements of that system, again, just being shown on on this slide. So this is to aim to get back the 20% of lost capacity when the tower does go into Cloud. But potentially there are further benefits if we can prove further that we can operate with AI doing this Same job but basically in much more reduced weather conditions. So my final slide if we we move on to that there are further benefits of adopting artificial intelligence within ATC operations. And specifically, I see AI as being very usable in areas where human performance is limited. So the overload of controllers, so where workload becomes increased, then AI support is a significant use for it, but also unconscious at the moment that a lot of our airports are suffering from a downturn, which we certainly hoping is, is temporary and it's its nature as far as the significant impact has had on the flights at our airports and through the international network. But the underlaid of air traffic controllers where they are not working at their normal high rates, workload rates can actually help The negative effects with regards to human performance. So, again, an AI system will simply be monitoring and processing the parameters that it has learned to process regardless of the workload. So it can simultaneously scan in 360 degree view, and it can scan two aircrafts simultaneously in opposite parts of the airfield, or it can scan multiple numbers of aircraft up to the normal sort of hundred or so per hour that we would anticipate through the major London airport. So that's one of the key areas also improving efficiency of the infrastructure particularly runway capacity is is a key area for for us and again, enhancing that efficiency of infrastructure means that we can do more with what we have already.
We also have a benefit here where there is a limited number of things that are controlled or any human can basically process simultaneously. And by being able to process higher numbers of parameters That affects outcomes, we can balance that to a greater extent reducing taxi time and impacting particular outcomes with regards to some of the sustainability targets that we're aiming for, at the moment, so improved resilience all around. And actually, one thing which I'm sure you would have expected me to have mentioned and clearly as a key part of this if we just go to the last part of this slide, but it is an important part on the the next part, please, which is obviously all of this underpins an enhancement in safety. Now, our systems are very safe as they are so it's it's always, I think, good to have safety where it's provided an additional benefit. But it is not the necessary the first part we start to use case on, but we certainly ensure that the outcome of the implementation is as safe if not safer than our current operations. And that pretty much concludes my initial presentation. I'll be happy to take some more questions.
Thank you so much, Andrew, for that brilliant presentation. And I quite enjoyed the operational use cases for here in areas of ATM, quite interesting to see the software and how it's working that the progression of the testing is also interesting that you mentioned that, you know, the intent of AI is certainly not to replace humans, but to enhance human performance efficiencies, and to, at the end of the day, improve safety. So it is interesting indeed. So thank you so much for that. And please, you put in the questions and put them in the q&a section of zoom as well, and we look forward to addressing all these questions. Thank you so much. So our last panelist monopolies is a brilliant gem Phillip, so generally we'll be talking about empowering aerospace company companies for using AI solutions. Over to the Shan. Philip,
thank you very much my noona so yeah, so we've seen a lot of the grave, awesome technologies, how it's applied in, like really AI in the sky, I cheated a little bit, I'm trying to do the ground to the sky. So actually, it's more from our point a bit what we're doing. So again, I'm roughly playing corporate innovation managers. So going to the next slide a bit very briefly so so understand a bit the context so I work for collision, nada tech. So really what we're doing and my role is to help corporations invent more like to consume any innovation ecosystem that's really my role. So like connecting with like young startups and all that and vulnerabilities and even how to treat all that. So going to the next slide. So a bit what I've been seeing so really, I work a lot with just to understand like with, like the big the bigger airplanes manufacturers, and as well as like older suppliers. So that's a bit like just to put you in context and it's a bit outside of the ideation. But they have a big impact on that. So basically, the reason is we say behind, but it's not a bad thing at all. So it's because every day, every solution, everything else is extremely high impact on everything we're doing at the end of the day, like the life or are at risk. So that's why it's one of the point also, everything is built for resilience. At the end of the day, we're very happy about it. So when things works, well, we don't want to touch it too much, because, well, it was very well, it's been working for decades, then if we go back to Andrews is like, even 100 years, the same thing that we're losing and optimizing every, every single year. So so that's why it's built for that, but it creates a big a big chain a bit late to, to, to the mentality, I would say model mentality, and also the General Dynamics at talking with some of our clients. Basically, it's okay, like they want to innovate and stuff, but at the end of the day, just everything is much slower in terms To confirm the sales pipeline all the way to applying contracts, we talked about, like we're talking about years and years to do everything because let's say, a contract and takes three years to sign if it's not more, but after that you're good for 2030 years. So that's why also it goes a bit with the, with the flow of it, and also the regulation but again, like we're very happy that they're there, they're there for a reason. And it's to work with them. We've seen a lot of it's great that the consensus like we've seen at the beginning with Paul is great that the players are there I was there all the entities are there really to to really to sit down under question and really to look forward. And for I would say also the language that's in every single domain like the lack of cross domain expertise, what I mean by that is a either right especially say especially so data scientists, specialists with a knowledge of aerospace aviation, aviation education, domain specific I would say expertise is very hard because I wanted to New technology. So basically, it's very rare that people would have been doing their death for decades. Usually they actually professors, and after that like to be going to the market, and for sure, most of them, most of them, and it's it again, it's a general it's like very highly optimal old processes. I put it that way. It's again, it's more they're really use a completely new way of building things of designing. That's why a bit what Michael was saying that kind of the new ways of designing new way. So it's, it's kind of like to disrupt a bit a bit that part but again, it's a very, it's an older, it's a very old and because it's built for resilience. So just again to put in context. So going to the next slide, there's a lot of there's a lot of hope. No, there's a lot of strength, and we've seen all of them, what they what they were saying. So whoops, they all skipped but so there's a lot of great that a dangerous restraints, so because it's a very controlled environment. Okay, strange seems to have a an admission that I don't know why is there
I can so that it will settle.
Sorry about the noise. But it's kind of the the industry standard I want to I want to say is, there are a lot of different great, great stuff. So it's a very controlled environment, I can share my screen if, if you don't mind. or otherwise, I'll just I get there last night, that's fine. So control environment. So they have a lot of data. So the everything is control, because at the end of the day wants to control the output. So and they actually have a lot of ways to continue to track to track everything we thought we go all the way down to metrics, you're processing all the way to all the fleet management, the the flight when they went all across the so they have a lot of data, you have a lot of them. And it goes really far in the entire supply chain of that. So that's very One of the big strain of that is control of governments. So, AI systems, they love controlled environment because they know they can rely on that control environment. We've seen though in the aviation, like the fleet is like, temperature visualization is never in control, I would admit that. So it's not as easy. But after, after that, we have a very high level of documentation as well. So in every processes is documented because the industry asked them to do that. So again, we see like the regulation, that's the strength of that. So there's a lot of the commission documentation that is like faulty, good, bad bills, and measurements, maintenance, whatever. And everything that is going through different formats, all sorts of data can be like a report images, video captures, and so many more hours. So the data exists, and because of the documentation, which is caused by the regulation, but that's a great restaurant of that industry. And after there's a wide array of repetitive processes, so Another great thing for an AI system. So there's a lot of different events that that occurs multiple times. So it's it's good for AI system to learn about those, like the multiple occurrences of different events or good events or bad events or whatever events we're looking at. So that's another big, big thing that we we can leverage in that industry. And also they have a wide range of variables we've seen even like with with Marco and Andrew is even they're implementing new new ones. So they're trying to come claimant, enhancing like the what is already there to announce new new sensor data capability to create different types of data. We call it like real world data. This is simulated data, a bit of mix of both historical data. So they there's a great mix of different types of variables that could really help building your feature efficient feature system And reprioritizing towards that. So now we're going to the great opportunity that is already there. So we can have to change into switches, right? So all that to come so what we've seen again means more a bit at a ground level. So a lot of potential is all about like the machine vision for quality control at every senses of the words that we can yes talk about body manufacturing, but at every at every step, step of the way, all the way to the to the to the airplane is what we we've seen a lot is because if we look at the aerospace industry, there's a hundreds of suppliers for example, for one part, so and they have to have the same processes or quality system or quality assessment. So it's very good. So it's very scalable solution and they want always to automate so this is that we work with the with those solutions. For people it can be fun, really especially difficult. defector detection, that we see that through computer vision and even other sensors because they want to go deeper and deeper we need to categorize the type of defects really to help because there's a lack of expertise itself just in terms of quality control is an expertise in itself and is going lower and lower. After four we call it like predictive maintenance but I would save and predictive analysis we've seen great solution like application a bit before with Andrew a Miko about like a bit the predictive analysis way to to assess what is there to two different things. So that's a bit in depth part, really to understand where we're going from what we know. So it can be at the defeat management level it can be at just for purely like the if we look at the air boss of this world, so really to
understand all the supply chains about it like predictive when we do maintenance For different specific parts, we look at engines we look at every single part. So those are the solutions actually we're currently working with. So it's really to look at different ways we can get there. So they depend on understand what's coming, what's coming for them. And we need to again do the right prediction towards that after that for sure all the supply chain management so I put it from the ground up. So yes, it can be just a supply from everything around the world, but also mainly to for how to assess and supply all the airplanes in the air. For all the latest trailer, they own traffic as a company in itself. And in collaboration with with others, that's a hard part, but I want to get back to it. And also, another good option that we're seeing is all is all about, like the chatbot for consumer experience, how that's really something they're looking for. If you look at the statement inside, like inside The airplanes aircrafts, themselves if you look at just belief on the customer experience all the way through the journey of acquiring tickets all the way to their final destination. And so those are great opportunities to actually working with with buyer with different partners. And I also want to, from the beginning were also mentioning it's everything to collaboration because again, there's a lot of stuff that can be done young companies and your talent is really what we're looking for. This is why for me my my role is to know you all and actually helps you guys to work with the big the big guys because evolution is is highly regulated. So you need a partner to get to get into that space. So if you have social flagellation, well You better work with Marco Andrew or different entities like them because and approach them because they're looking for that they want to get better every day. And you as a younger company or like a great new New technologies you need that platform to get in there because you need the resilience you need the security, you need a whole bunch of infrastructure that just that is very hardly to doable. So for me on the last note, like you can switch to the last slide, basically, but we're always looking for for new technologies, but at the end of the day is it's very, it's a very stiff industry. However, they're very hungry to look now it's like I've seen a lot of movement that everybody wants to do to get new solutions. And elsewhere, we're there to help that. I know the other ones were very, extremely concrete uh, Mark, Wendy, but that's why I went a bit up because it's like, I want to leverage what you guys have to offer. And even if you want to propose what you guys have some really like ideas and stuff like that or things you work on in your, your research or anything. Like, feel free to contact me and I'm pretty sure we'll find the like, solutions for our great partner. So thank you very much.
Thank you so much john Felipe and I do apologize their association for you for the problems with the slides. You know what you did brilliantly anyway, it out the slides. So we come to the next segment of the webinar and that this will not be the question and answer segment of the webinar. And that time for this I will be doing it in coordination with my colleague, Yuri fetter, also from IQ. So you read we have the first question.
Good afternoon. Good evening. Good morning. Thank you. Muna so yes, we have quite a few questions. And I think safety is always first in aviation. So I we can start with that. We're getting quite a few questions broadly on you know, how do we know it's safe? What kind of proof of concepts are we using? So what I suggest we do is a question can first go to all also he has a great Or overview from your control. And then maybe our links from that system and ether can provide a few more specifics about how are they? So, Paul, if you could generally cover, you know, how how do you know how do you build that trust and confidence that the public needs to know that this artificial intelligence applications are safe?
That's a very good question. I think first of all, as I mentioned in my presentation, we highly recommend that all layers in the organization are are well trained, are aware about what artificial intelligence actually is and what it isn't. There's a lot of scares and fantasies and fears around it. And then when you sit down with people, you explain it, that it is just I would say standard analysis with just many more parameters. Then you would have a normal algorithms and people want to quicker understand that. And another thing that we that we definitely found out as well in the applications that we put operational in our own organization is to involve the operational people very early in the design process. It is not that the engineers went loose and came up with a solution. And then we're then looking for a problem to solve. No, right from the beginning. We sat down with the operational problem people and we asked them actually, now what's the problem that you think that needs solving? And together step by step, we build the applications in a very iterative and dynamic way. And I think those two things have proven to be really key success criteria for putting things into operation with confidence.
Thank you, Paul. So Andy, if you could provide maybe from your experience and your very specific use case, how you've applied perhaps what Paul has been Explaining.
Thanks. Yeah, so first of all, I would agree with Paul entirely that involving the end users in the design process is a very important step. Particularly right now because AI is not something in common use in aviation and particularly in air traffic control. And you'll have noticed in my slides that I mentioned straight off that the team of current air traffic controllers are invited into the laboratory facility which we have put on site there at Heathrow, for exactly that reason. So they can be involved in in the design requirements, and even down to, you know, how the system then interacts with them, and what level of interaction they want. So, you know, we offered from the outset that if they wanted to see the video, that the AI was being monitoring that they could do that. That was an early decision that came from the end users that No, they didn't understand what the AI was doing with that, because of the demonstrations of it running live. And no, they didn't want that, because that was just more information to survey and process themselves. So that was a, that was a good start. But I think it is key that you have that involvement. The other part is predominantly really in ensuring the safety of it like any other system, and it's just a key part of that system. So that's why we've done operational data collection validation to show how the system operates in different conditions. And we've done quite a lot of analytics. In fact, we had something like a 90 page report of the different parameters that were analyzed as to how the system was performing. So from that perspective, it's it gives us a good grounding to be able to build a safety case as per our normal safety management system, which any navigate service provider. You know, that's that's their day to day. I think the other key to it, though, is actually demystifying at AI from the point of view of it isn't some kind of strange being It is simply a way of being able to program a system with algorithms capable of processing a huge number of parameters, as Paul mentioned, as well, rather than manually inputting those parameters. And actually, I think Marco would probably step in and tell you how different the amount of time it takes to manually input parameters versus machine learned parameters. That finally the only other thing I would say around safety assurance is simply ensuring that you've got the system focused on a key element of your operation because that way, your machine learning is more focused and Likewise, the ability to test that machine learning is focused. And the last part of it is simply, we have used closed, machine learned models. So once we've trained the model is is no longer continuing to learn within the operation. At this stage, I don't think we're at a position where we would be able to certify that type of system within our own operation, the levels of an unknown would be too great. So I think simply by by following those different steps, you can ensure that the system is as safe, I can in fact able to do more than a manually programmed system.
Thank you very much, Andy. Marco, since any had mentioned, would you like to supplement that with a little more information? Thank you.
Well, sure. I mean, Endian, Paul already made most of the points that I was going to bring up and the only other thing that I would say is that when we design systems is important to know that the AI is in running by itself. In the operation, there is some logic traditionally program systems in place as well for data integrity monitoring, for example, so that we can ensure, and I saw a few questions in the QA How do we ensure for example, that the data from the that we receive from the cameras from the radios good, so we have some integrity checking before it goes into the AI. But then more importantly, we also what do we do with all our AI systems not all our systems period says we have a dedicated safety module, the only responsibility of that 50 modules ensure that everything is running as smooth operation and make sure that before we present anything to the user, or in the case, for example of a traffic light or the mission before we send a signal to the traffic lights to change color, we have a certain safety check to say everything is running good. And it's the last final gate as a human would do to assess some information say everything is running smoothly and then we can present that output to the user. So for example, we get the confidence output from the AI and if something is not recognized to high enough level of confidence, then we say, you know what, this is not safe to make a decision based on 75% 80%, then we can say this is in a situation is running in a situation that it doesn't recognize, Let's fall back to standard operating procedures and analyze the situation, do a root cause analysis and come up with a corrective action to handle the situation the next time.
Or much. So
I want to in this particular question on regulation, one of the comments that we got, was that the question on what the role of AI tail might be in terms of safety regulation, so you can either take that question now or when you wrap up, but the wrap up.
Okay, thank you.
I want to shift a little bit to another area of questions that I think is very interesting. And that is, so in aviation, traditionally you had to kind of build on top of The previous level of maturity you had is infrastructure. You know, you went from nothing to slightly sophisticated to very sophisticated, highly sophisticated. But the question here that I'm seeing in some some some areas here is that is this new technology only going to be available to those who already have highly sophisticated systems? Or is this leapfrog technology? And maybe john, Philip, if you could start us off with that answer.
Yeah, absolutely. So because we really work at early, the very early stage, and also I want to do a bridge with what Paul was mentioning. So a lot of people they don't really know what it is and what it can do. So usually, us we this well, we we help them understanding and so they can make business decisions. It's like we don't want them to be technical experts. So yeah, and that's right. It can be it can be a leap, right. However it is that you need to have like A company with infrastructure to help you go forward. We've met a lot of younger companies with a lot of like great analogies. I'll tell you an example like concrete with the talus group for example.
we work with they work with us for all the young companies, there was a great one here in Montreal. And it was it was called air and basically what they were doing is is really helping the residents into 4040 men really picking up content and things like that, but they were not at all in the defense or no aerospace. So they could they try it but it will for them was awkward, that's that's too big for us. We can warn other interested in when they work with that as well. Basically they they help them to say you know, use our platform to be integrated in the airplanes, for example. So now that was like within six months basically took this what it took for them to be integrated into aircrafts. I think it's doable. It's doable as a new solution to BDT skip ahead, but you need to select the right partner that has the right data and right mindset as well. So, and Tripoli, everybody's sitting here like the, it's kind of the partner you're looking for. And you want to work with them to make sure but I think there's a lot of potential for younger ones, you don't need to use whatever that was built before, because it's kind of a new way of thinking as well. So I will stop there. I want to the others to compliment they want to but
thank you, actually. So Andy, the technology that you were describing, is that something that could be implemented anywhere in the world? Or is there a minimum, let's say requirement that you look for before you recommend that it be used?
Yeah, it's great question. So my view, actually, it looks at it in both ways, because I get major airports saying to me, that they are just about to invest in or have just invested in some of the systems like the radar systems that I showed you. presentation, you know, and they're worried that perhaps that was the wrong way to go. But my answer is simple. whatever data you have is useful. And therefore having those radar systems, that data can be fused and used alongside of the camera data. Likewise, if Apple has existing camera systems, then you can take some of that data and either augment it with with additional cameras, or you can simply use that. So I would suggest that if you have a use case for your operation, and that's a key to make sure that you're focused on a use case, which is valuable to you. Then basically, you then look at what you require. And it could well be that you have a lot of that information already. Some of the stuff we've done with Sarah which very recently has actually kind of looked at a fairly common area, which is Is voice communications as being a something which, which almost gives you a baseline. And again, if you have that voice data, then we're able to go through processing of that, from the point of view of of it being able to track a specific outcomes or support for an air traffic controller. So I think really, it's very much a case of Look, look for what what you need you need a system to do. And then do you have enough enough data? What do you need to supplement it with, but you certainly don't have to start as Heathrow or any other major app or to be able to take this kind of AI technology forward.
Thank you very much. And so I want to pivot a little bit to a question that I see that all has already started to answer online. So I was gonna ask Paul. So we have some questions on jobs. And you know how Job structures gonna change for aviation. It's a bit of an elephant of the room sometimes, but especially for the younger people who are about to get into aviation. Is there any guidance provided by your document where we can kind of see what the jobs of the future of aviation might be a little bit different, or very different from what we see today.
If you if you want me to tackle that question, what we identified in the in the fly report is that certainly in the short term, middle term, we don't see a massive loss of jobs in aviation, because of artificial intelligence. What we do see though, is a shift of the type of profiles that you need. And I think there that's nothing new in aviation. I, part of my career was working in aeronautical information and the whole IQ and iq 15 etc, where we have moved from The paper based AIP and the unreadable notams into much more digital systems, and electronic and digital data sets, etc. Does that mean that we now have much less system people working in AI is? No, this is not what we see, at least not in Europe. What we do see, though, is there are many more people, maybe less people on the pure operational side, and many more with a technical background, all of a sudden, we have much more people walking around with a PhD in an AAS officer than we ever saw before. So no, we don't think there's a major the diminishing or decrease of the number of people certainly not short term, but we do see the need for more profile, more skilled profiles, as well coming into the business.
Thank you very much and not so related, but I think you somehow to mark on if I could just ask you on the questions that I see here. So the question What are the limitations? Or maybe an example of something that you thought would work but then later you found out it didn't work?
Yeah, yeah. I think that's that's a great question and I think it's a great point to make is that and I think Paul has touched on this already this there's a lot of good information out there but there's also some some bad information out there about what AI is going to do and some of the stories you know, the the Skynet stories and it's going to take everybody's job so I think it's important to see where we are right now with AI and where we're trying to go and again, it's not to replace the human controllers it is it augment. And, and what I want is, for example, an example of where we have thought everything was going to work flawlessly is any mentioned the voice communication. So everybody looks at their Google Home, Amazon Alexa says, Well, you know, we're there and we can understand everybody's speech. Well, let's just let's just automate all of the RT in traffic control. Well, turns out that the communication in SFP controllers is quite hard to understand, even for us He was a human from me coming into system engineering nor at your school background at all, listening to some of the the speech, especially accident speeches, it's quite difficult, not only from a noise quality perspective, but also from the jargon accents all around the world. You have pilots from Dubai being trained in the UK, then you have Spanish pilots being trained in UK, they all have different accents. So what we thought was going to be relatively easy as quite complicated because there's so many parameters, so much variability in the data. And so I think that that is an example where we thought everything was going to be easy. You're looking at the big tech firms and then realizing that the aviation problem is much more difficult.
Thank you. And perhaps Andy, if, if you have some information, share with the group on, you know, how long did it take to train some of this stuff? Yeah, how, how much data was needed, and it was, were you able to predict, let's say, the development cycle
Yeah. development cycles were sets, as we do whenever we do some of this sort of innovation type work, either in the UK or elsewhere as fairly short turnaround. And in fact, that was done because of the fact that we had the software engineers and the end users working directly with each other. So that certainly helps with regards to reducing cycles from sort of traditional industry approach to development of systems. But I think from your point of view, the model training particularly if you have it in a safe environment, like in the laboratory, where you are proving then, you know, it actually can turn around such that it can be a matter of weeks, within, you know, sort of starting to actually having your first model trained, and then just literally test to see how it performs in initially a fairly high level number of different scenarios. Yeah, sort of daylight, low visibility, what have you. And then once you've got that there can be further refinements on the model. But I think the other thing also to bear in mind when you're looking at success is not to immediately take a failure as for what it is. So again, going back to when Marco was discussing about voice, you know, when we actually we looked at the performance of systems set up trained to take voice and convert that to text. directly, were some times where we we didn't get a successful result. And then we put a human traffic controller to listen to that transmission that had been received. And the human couldn't, in fact, make sense of what had been said. But the human air traffic controller on a regular basis during the working day, quite often will say to a pilot, so again, giving the pilot your opportunity to rephrase say more clearly include a full call sign, whatever it may be, that is the problem that that caused the system, their human or otherwise to have an issue with it. So actually, when you start, they're looking at the system being able to make a similar request of a of a user to repeat, then again, you get an opportunity to rectify that problem. And it's simply a case of determining whether the system is asking for too many repeats, or whether that's kind of an acceptable level of sort of second requests that it makes. So I think you have to kind of work through it with that, with that in mind with end users because again, the end user will often say, Well, I can't I can't understand that I would ask the system to or ask the person to give me a further run. Go at answering that question. So long Likewise, you know then then stepping in telling the software engineer don't worry about that give it the allow it to to make pre programmed response that often that often then ends up with a quite a surprise as to the how many times it's acceptable to to make that kind of request. So I think it's what would nga says and train and test in rapid cycle so development Sprint's are really useful.
Thank you very much. So as a last round of questions, and I think we can ask this to all in terms of the Sustainable Development Goals, you know, improving the lives of people all over the world. Do you have any, any specific say dream that you think that your, what you're involved with, can have a good effect on the people, the societies in the world so a little bit out of your comfort zone? I know because we're very excited. Hear but, but feel free, if you could just share some some vision of, of how you think that what you're working on is going to help us as a community, you know, beyond just never losing our bags again, let's say. So Paul, maybe you could start.
Thank you for that one.
So how could Kota artificial intelligence help us with World world peace basically? Okay, so thinking out loud, I mean, what do you need to do artificial intelligence, you need two things you need first of all digital data, which increasingly become available. And I saw the question as well about Africa and the worries they have to roll out AI, they're etc. I think this is feasible. It's fairly cheap nowadays to turn data, digital etc. So again, as long as the data quality is correct, you have one you have the digital data. Number two, what you need for AI is CPU power, unit computer power. And if I See nowadays, what my boy at university is doing with a Raspberry Pi, I think we are lowering the threshold for lots of people to do AI. And I think with that we're starting to tap into the brains of many more people, certainly youngsters, who have a lot of bright ideas on how to improve the world and specifically aviation in our context, etc. So I would see it as a way Amal, I'm an optimist, maybe naive, but I'm an optimist. And I see that this whole digitalization will make solutions and r&d available to many more people. And I'm sure good things will come out of it.
Thank you very much. And maybe Mark if you want to supplement before I ask the other two.
Yeah, for sure. I mean, I've always cared very deeply about environment and climate change from when I was in high school all the way through unit recipe. So I think one of my dreams would be to to help that Sustainable Development Goal of curbing climate change and even reversing it in a way through the use of AI. So we talked a little bit about Heathrow and then how we can bring some of that land capacity back. And you know, reducing the fuel burn. And that really comes back to the variability in operation. And that is something that is not only important for an airport operator in terms of bottom line and and everybody wants as an airline to get the same service every time but it really reduces the overall fuel burn and really inefficiencies in the system. In aviation. It's like every other system if you run it away from its steady state normal operating procedure using efficiency. So simply if you drive your car then not correct speed or too high of the speed, you're burning more fuel. So what we're trying to do with this is really reduce the variability in the system in order to curb climate change and reduce fuel burn but also noise pollution, things like that. So if you would say that that is my dream in order to help just a little bit with with climate change
Thank you. And do you want to supplement that real quick?
Yeah. So I would say the same thing, I would say that the experience of people within the system, the traffic management system, so that's both the controllers and pilots, but also the passengers would see benefits in the same way that we've seen those as passengers going through airports and the use of technology, how that's changed over the last five years. It's significant once you get onto the air side and into the air traffic management area. That hasn't been a significant over the last five years. But I think there's a potential for that. And I'd agree with my colleagues that you know, that can have an impact also, from the point of view of being able to control a large number of parameters means that potentially that can do anything from making something more predictable to also being able to manage noise more effectively and, and fuel an emissions.
IQ and john Felipe, we hand back to my room.
Yep, so it's a bit indecisive with the same same sense of all the environmental and climate change basically also as much control it just before to put the plane up, like everything that is on there involves like a lot of a lot of different people at different companies. So there's a lot of movement before even building that plane. So that aircraft so on that end, just up to magga still optimizing that part of it, I think, would make a great change. A lot of it also and especially the course CDs and stuff and more philosophical part of it, it's more to understand like the new generations value and another way I could do that, but with all the social media and all that stuff, maybe if we can learn more value, see how we can actually leverage that the continuously evolving the values of the new generation and take it to profit, basically The two to two different causes. We want to we want to challenge so
thank you very much. Well, my mana with that I'll hand back to you.
Yeah, thank you so much for brilliantly run session of question and answer. Thank you so much URI. So one of the questions that Yuri asked was the role of IQ in supporting the implementation of AI. AI. I think IQ can play a very important future room. Because one of the functions of IQ is prevent sets of standards, which will help regulate aviation across the world in future. It has a role to play in developing high level standards, the role can be implemented by its membership, to help regulate the use of AI. For now, this is not yet mature. There's a lot of discussions going on. There's a lot of research going on, and that he participates in these things like this forum, for instance. were part of the discussions we are we are part of our panels and an order order mechanisms to see how we could promote the use of a good, it's very clear that the use of AI is very important, it improves efficiencies, and it can be very beneficial for any system. So in short, basic features like you can play, and one of these would be to develop high level standards when the time is right. So with that, I will and before closing the panel, I would like to highlight some of the key messages or recommendations that we raise during the discussions. And some of these related to the establishment of data foundations and infrastructures. These are very, very important if you want to actually implement a good AI system. We've had discussions on the need to further explore the potential of AI in aviation and ATM and that the communication element on AI It is important for you to be enhanced, especially when it comes to the issues of demystifying the use of AI we've seen a lot of talk about that particular issue from and also, we've discussed issues related to the aviation community that needs to build up an inclusive AI, ATM partnership, this is very important. And that AI could be used to improve operational efficiency and predictability, especially in the ATM sector, even in flight operation when you come to low visibility operations and so on. And in a lot of areas Well, I think the development of AI in aviation should be pursued in a manner that leaves no country behind. We've seen some of these questions that have come from from from the audience talking about some particular areas in the world like Africa, where you have issues of infrastructure, issues of qualified personnel and so on, but what Whatever the case may be, to make it very useful. You know, we cannot leave no country behind, we have to bring everybody with us. And there are significant uses of cases of AIDS in aviation to help deliver on the SDGs. We've seen that in the IQ video that we have. We've shown as well as the video that was said in the presentation from Andrew as well. So with that, I would like to wrap up now and to say Finally, we have come to the end of our session. I would like to thank again, our eminent and brilliant speakers for joining us today and addressing this very interesting topic. So thank you, Paul, Marco, Andrew and Jan Phillipe for your brilliant presentations. I would also like to thank all the participants for the incredible participation in the TOEFL questions. I quite enjoyed watching the questions come in one Two, three, you know that excitement was interesting because it's a new topic. It's interesting is technology is the future. So it's been quite wonderful on my part, you know, spending this hour or so with you. Please do stay safe. And thank you so much also for thank you so much to the ITU T. And I'll hand over back to you it back to you.
Thank you very much. Hi, Mina. And thank you also to iCal for this session. It was a great pleasure hosting and presenting one of our first un partner sessions. A big thank you also to our panelists as well to all the participants for making it so engaging, and providing so many interesting use cases. Before we wrap up, I would like to highlight a few things that may be of interest to you. This Thursday, the ninth of July, we are very excited to partner with the Association for Computing Machinery ACM to present our first AI for Good keynote professorship patch Patel is the 2018 recipient of the ACM prize in computing, and he will be here on virtually he will be here on Thursday to present novel perspectives for how ubiquitous computing can help improve our approach to healthcare and the COVID-19 pandemic. And next week will be a busy one, we have three webinars next week two will be hosted with the XPrize Foundation, and one with AI for Africa. So be sure to tune in. We're also posting the links to these sessions in the chat. And you can also find out more at any time on our website AI for good.iq dot IMT. With that we have reached the end of the webinar and we would like to once again thank everybody involved, our panel participants, partners sponsors and our co convener Switzerland. Thank you very much and we hope to see you this Thursday.