What does it mean for AI to be in the automotive industry? How does this impact future software developments and integration? What are the misconceptions about AI in this context? These are just some of the questions our seasoned co-hosts, James Carter and David Fidalgo, discuss with Roger Ordman - Executive VP of Marketing at Aurora Labs. Want to hear what they had to say ? Sit back, plug in your headphones, and enjoy their takes on one of the newest aspects of the automotive industry.
This podcast is powered by Y-Mobility.
What does it mean for AI to be in the automotive industry? How does this impact future software developments and integration? What are the misconceptions about AI in this context? These are just some of the questions our seasoned co-hosts, James Carter and David Fidalgo, discuss with Roger Ordman - Executive VP of Marketing at Aurora Labs. Want to hear what they had to say ? Sit back, plug in your headphones, and enjoy their takes on one of the newest aspects of the automotive industry.
This podcast is powered by Y-Mobility.
Hi, and welcome everyone to another episode of Byte Off. This is actually episode number four for us. And we've, David and I have really enjoyed putting these podcasts together. And welcome David, it's great to see you again for number four. So how are you, my friend?
David:Not too bad, sir. Looking forward to have a conversation today, I think is we have a good run. And I think we have a really excellent guest today. So looking forward for the conversation.
James:Yeah, you're absolutely right, we should do have a very interesting guest today. And speaking of him, I would like to introduce everyone to Roger Ordman, who is from Aurora labs. So Roger, thank you very much for coming along and talking with us and informing us a lot more about AI today, we really, really appreciate you being here. So with that, tell us a little bit about yourself a little bit about Aurora Labs and what you've been up to and as a kickoff to our AI discussion.
Roger:Hello, James, and David, thank you for having me. My name is Roger Ordman, I'm actually based out of Israel, if the accent might be deceiving, I was born in England. And myself, I've been 20 year career, starting with an engineering degree. And then through working product management and marketing, from network devices through mobile and focusing mostly on automotive for the past seven or eight years. Main focus of what I've been doing, and Aurora labs particular is around the challenges of managing software. So from over the updates of software, securing software, developing and certifying software. And it's getting more complex, it's getting more challenging. And we think there might be some technology out there maybe an artificial intelligence technology that can help give us the tools we need to become more efficient, and improve the quality and security of the software going to Cass. And that's what Aurora labs is all about.
James:Very, very neat. Now, before we jump in, maybe a two second summary of what AI does with Aurora labs, what you do with that?
Roger:we use AI to recognize patterns in the relationships and the behavior of the software. So is that interesting, we think it's very interesting, because the way the software behaves, indicates the way the software is running in the vehicle. And if you can identify patterns of misbehavior, the software, you can essentially predict software which might fail before it fails. So if we do our job right and use the right AI tools, we might be able to help the car manufacturers find problems in the cars before they cause failures, and have them focus on improving the quality instead of running around trying to fix problems.
David:In your experience Roger, how the industry is the industry fully understand AI? Or is still a little bit of work to help them to understanding what is AI and how AI should be used?
Unknown:I think, I think we can all agree and probably everyone listening can agree that there's a lot of confusion on what AI is all about. What is AI in general, what it can do. And obviously, focusing specifically in the automotive industry. I mean, there's, it seems to be at the moment people using the term AI as a silver bullet that will solve all of their challenges. If we haven't got our own intelligence, we use artificial intelligence, and that will tell us everything we need to know. But David, we know that's not the case, there's some tasks that don't even need AI, you do not need necessarily to analyze patterns and predict outcomes. Sometimes you just need to have algorithms that will take inputs, well known inputs and do an action. AI is really for areas where you the inputs are unknown, the variables are very great, and you cannot define and write a definitive algorithm that takes inputs and creates outputs. AI is really for places where the inputs are variable. And that's why AI is often associated with autonomous driving. Because if you're looking at a road, every car and every dog and every tree looks different. So you have to start identifying patterns to be able to see something new and still recognize it based on what you learned in the past. That's where AI is fantastic for but it's getting better and better, maybe fantastic is optimistic, is improving and getting better and is necessary for autonomous driving, but not everywhere requires AI. I think that's what the challenge is is identifying when AI is required when you need to have learning mechanisms, learning algorithms that can then based on what they learn to decode words, recognize unrecognizable patterns in the future.
David:And you will say that for example, AI coming is this, we have this conversation before, right? Basically a lot of there's a lot of a type of vision, right? One is like AI is a tool. And then AI is a philosophy. Right? That basically is a suffering in itself. What is your thinking on that? Why you think this? Is that, as you said, Everybody's looking at a silver bullet. But what is kind of like the the main context to develop AI and what it should be used for?
Unknown:I think the first part of that question is AI
Roger:And how you think its been using in automotive philosophy a tool, is a fantastic question. To me, AI is definitely a tool. You know, I was talking to a vice president of a car manufacturer just yesterday. And I said, you still because there's a whole discussion in the car industry about, you know, 10s of 1000s of engineering jobs that are yet to be fulfilled. So we still need a lot more engineers. And I was saying he, you know, he was saying that's his biggest pain, he hasn't got enough engineers. And I agree with him, essentially, we need to have more people focusing their big brains on the challenges of mobility, mobility. But what we as an industry need to do, as managers need to do is give our engineers the best tools possible to help them be successful. So I don't think AI is replacing engineers, I don't think AI is the threats that some people go well AI will, will take and get us all out of jobs, I don't see that I think there's still we need people in many, many places, in many jobs in many industries. But all of these people can can do better with better tools. I have a friend, a colleague who said his industry. I mean, it's a lot of conversations of AI for talks and compares it to when the tractor was introduced to the farmer, we still need farmers. But instead of plowing by hand, they now have tractors, they can do things more when combined, they can do things more efficiently. The seamstress has a sewing machine, you still need a seamstress. I think the same thing here, we need engineers, but we need to give them better tools to help them make better give them more information makes they can make more informed decisions. So to me AI is definitely a tool. autonomy. Right? But I think I think it for me as well as I want to take out this stigma now. Because AI has been using in automotive industry before, it's not been using AI; it's not been calling AI hashtag right. But basically is how is that understanding in the industry in automotive of what AI means, and basically how it's been using how actually can help in the future with autonomous you mentioned a little bit before. But what is your vision on that, in terms of software too, right?
Unknown:I think in the industry, the AI is still very much associated with autonomous driving. It's probably the most mature use of AI in automotive in automotive. even have that mature is a difficult term to say about internet technology is only a few years old. But it's still that is the substance of the area, we have the most curious references. What we're seeing in the industry is a term called shift left, which is basically the earlier in the process that you identify a problem and fix it in any process, the cheaper the solution is. So if you if you're building a building, and you find a problem in the in the windy construction of the walls, and you identify them earlier on, it's much cheaper to fix. And after the house, the house has finished being built. So shift left in the development world is move a lot of your quality tools, a lot of your insights early in the process and don't leave all the testing and all the insights nor the analysis to the end of the process. Again, fixing things later on, for example, fixing things with an over the air update on the road. Or even worse, fixing things with a recall are incredibly expensive. So if you can find problems in the quality, early on in the process, as is being quality, this is being tested, it's going through QA QC, as it's being developed, the cost of fixing those problems becomes much, much cheaper. So there's a big shift in the industry of how they can move shift left, how they can be more tools early in the process to help improve the quality and the safety and security of software going into vehicles. And that is where AI is coming into play is there's a starting to be an openness to that to bringing those kinds of tools in earlier on. Another thing that's happening there's two other trends that are influencing this. One is the move towards introducing CI CD and agile software development methodology. Also in the automotive world, it's taking some time, they're not the most natural people to go CICD but it does enable if the car was being designed over 567 year period in the past Now the software is sorted, the cars have been designed in a much shorter period and the software is evolved constantly evolving. Which leads me to the third trend, which is motor pushing all of this, which is a software defined vehicle as the software is being disconnected from the hardware platform. And as a lifecycle cycle, the software is being disconnected from the car year model. And it's going software is going across models and throughout the lifecycle, even when the vehicles on the roads that those disconnect that focus on software, that understanding of a continuous evolution of software, that's shortening the development cycle, the CIC, D, the agility, these are all driving an openness for new tools. And some of these tools are the ones we're talking about the AI based tools that can give insights much earlier in the process, and help improve the quality, much of anyone at a much lower cost.
James:Actually, that's interesting, Roger, because one of the things that we see on a consumer experience side, is when customers want it, probably the biggest complaint that the customer has had is is they don't understand why they can't operate their vehicle software, or that it's buggy, or, you know, some controls within the vehicle aren't working properly. And certainly it the ability to figure that out early on leads to much better, you know, IQs quality solution when the vehicles are out on the road.
Unknown:This is a very interesting thing that this conversation I had just this morning with a European OEM, he was talking about features of demand, and how he's enabling features interface. So as a driver, as a user, I could go into the infotainment system. And now BMW have just talked about I believe, enabling you to purchase a subscription for car seat heaters, which would be great where you live, James, and I would not need it at all, where I live, air conditioning be far more useful for me. So it's good. If we're both buying a BMW, I don't need to pay for that feature, you can pay for it in the cold months, which is probably 10 months of your year. And that's, you know, lots of configurations, which is great for you. So if we take that step back and understand the complexity, for quality perspective, if once the car came out with very with much less variation of configurations, and much more set configurations, the testing was far easier. Now, they have to test it with so many permutations of configurations and variations, it becomes incredibly complex. And then if you make a change to one piece of software, and how does that filter through to the rest of the software, you then go into this area of the unknown. And this is why standards, existing testing mechanisms become far more difficult and you need to have - because if you're trying to write test beds, scenarios, for every permutation of variation in every in every configuration, you can very quickly get to a point where there's some that you can't write all these tests, and you certainly don't have enough time to run them. So if you've got algorithms running in the background, AI algorithms are monitoring the behavior of the software as is being run, and picking up on deviations automatically unscripted. And without any manually defined thresholds. But picking up on changes in behavior automatically and saying this was not happening beforehand, this is changing, you need to go and pay attention for that, then those engineers can focus their attention and on what the what is changing and what's something that affects and hopefully, as a user, when you get to choose a new feature, you'll get it straight away. And it will work. It won't cause other systems in the vehicle to fail at the moment that's happening. But I think this will improve over time.
David:Because for me, I want to want to point before right? So is that because we mix, there are a few terminologies that I'll want you to understand with you. Right? That's because I mean, you were saying that AI tools need to be shift left, then we said that you need to be a software defined vehicle architecture one, and that's what AI helping us. So were you referring in terms of AI as a tool is actually when you mean in the Shift left is or because what I understand and sort of the improved quality and cost of integration and valuation of the different types of vehicles. Basically, you need to create a systems engineering framework with a really good process of development. Right? If we're going back to the V cycle, right, that's basically is there so will you say now, it's basically the AI tools basically helping to improve testing and validation of the of the software process or this is something that you've seen that is happening more now because it just to clarification, I mean, I'm my background is systems engineering and software development for automotives, so basically i was trying to understand what you're trying to put this wavelength and maybe there's confusion with the left hand side of the V cycle, yeah?
Unknown:I'll give you an example. So say you're receiving; you've got a complex system is receiving software for many different departments, you've got software from three internal development departments, you've got a software coming from an open source, you've got a software coming from two external suppliers, you have six or seven pieces of software, which are all being integrated to enable the system to work. And what I've just described as scary as it might seem, is very heavy. You seen that quite often in the car. Now, somebody may change, one of the suppliers may change their software. We could take all that software, integrate it onto the board, install it on the board, run it through all of the test suites, at the end, analyze the outcomes of test suites and find out that it failed, and then start rolling back and seeing At what point did it fail, did it fail with a test and it failed the integration to start rolling the ball backwards. And so we eventually find out what caused the failure, you go all the way forward to come all the way back. And this is, this will take a long time. Would it be possible? Yes? Would it take a long time. And like many resources, obviously. Another way of doing it is when you receive all the binary packages from the seven different vendors, analyzing just those binary packages and analyzing the symbols in those packages, and see what has changed between the previous version you received and this new version? And understanding, will they still communicate correctly? Or will there be a problem. And that's information you can get maybe not down to the line of code resolution, because we're talking about binaries. But you can get very good intricate information, the very early stage before you start integrating on the board before you start running the software tests before you go through all that process. And that only thing you can get an indication of maybe not if something is going to break, you can find that out very early on. You might not know exactly what but you found out something's broken, before you've gone through that whole process. And you say guys, before we start doing this whole testing integration, this isn't going to work. There's something's changed in this package here, which is changing the signaling, we're just going to make the whole thing not work. And we can give that insight using the right tools, you can give that insight much earlier on. So you're still gonna have to go back and fix it and so on. But you've saved maybe two weeks of integration and testing to get back that same point where it doesn't work. So okay, we know this will work. But now we saved two weeks before we came to that conclusion. That's the shift left, catch problems earlier. And the cost, therefore is two weeks less of development of testing. So it's much cheaper to find it early
David:Oh, okay. No, I understand. Eh, and
Unknown:however, David, I will add, however, that in that instance, we're not using AI, because you know what the inputs are, and you're calculating it. So that's not an AI Solution. I'm destined to be honest, we said beforehand, not everything needs AI.
David:Yeah, you're talking about software in the loop. Yeah. Because that is I think what you were talking about when you're doing these particular tests and so forth, testing that you can have algorithms actually, there are AI deep learning and machine learning base that can actually understanding deviations from signalling functionalities when supplier has to provide software that they need to be mandated requirements that actually the the OEMs provided them. And that I think is a good utilization. I think my point it was a good decision, or they have what he called it using AI based tools to understanding failure in the systems basically, or in actually the software code. And I think it's one of the things we joke we all talk about thing and we were talking about. We do Roger, there's a lot of applications of AI. And one of the things I actually love the you will bring it to the table is you need to put context on what we use AI for point to what problem we're solving. Right. And how we've been utilizing artificial intelligence to actually solve that particular problem, I think is a good approach. Because we we mentioned before there's I think James as well, this is another conversation you I think I have with you that when a lot of people do autonomous, okay, which is the context of autonomous driving, why? What is the use cases and basically how AI could be much to the context and the the situation you need to do to do autonomous driving. But generally, we are shifting to autonomous driving as well. What do you guys think, basically, which level of maturity you think that that we are now on understanding the AI that requires to actually put in autonomous driving based on? And this is a question for both of you, James and draw your right to understanding what is your view on that one?
James:I'll let Roger go first. Roger,
Unknown:I think we we mentioned at the beginning that you need AI when it's a huge fit with that then there's more unknown variables. If we look at the world, that the amount of unknown variables in driving conditions are too humid, too, too, too, too large. The difference of where the three of us live and the environment around where we live and the driving patterns that are driving and the weather, whether it's the height of the sun, so many different variables would make a system that could work seamlessly in all three of our locations. And we're not even talking extreme locations like New Delhi in the middle of the day that we're not, we're not there yet. And it's going to take a long time to reach that. That being said, on dedicated paths with dedicated use cases, obviously things like motorways, I think we're much, much closer to that than we were. Obviously, the original optimistic forecasts were way off. But I think we are getting much closer to campus on campus on motorways in more constrained environments, fully autonomous level five, all over the world, in any environment, I still hoped to see it in my lifetime.
James:Yeah, I would agree with that. I think it's really, you know, as we roll out, autonomous driving, it's really about slowly introducing different use cases and keeping them very narrow. And I know, for instance, the truck developers, that's what they're about. They're just focused on doing trucks on the highway, they don't bother about the cities that don't bother about any of those problems. It's just highway. So yes, totally agree with what you're saying, Roger. I also know that we've only got a couple of minutes left. So Roger, perhaps you want to wrap up with some comments on, on, on our thoughts, or from your side, the future of AI for automotive, particularly, in what you've been focused on?
Unknown:I think the AI is very important. And people need to understand it has great power for creating great insights. But people need to understand it's not a silver bullet, it is a tool. You need to be first of all, as David said, you need to very clearly define what is the use case? What is it you're trying to solve. And then you need to understand what inputs you have and what the outputs you need, but also what inputs you've got. And then you find the right tool for that. In some cases, it will be AI. In some cases, it will be machine learning or deep learning and good strong algorithms. And there's nothing wrong with that. That is actually certain sometimes that just works properly works well repeatedly. That's the best. So understanding what it is you're trying to achieve understanding what information you have to feed into the into the algorithms. That's the key. And again, AI being a tool will make life easier for the engineers, the engineers are not going anywhere. We need stronger engineers, we need more people to come into our industries and to be focusing on new AI use cases and how they can make our lives better and safer. And AI I see as a way that will help us get there.
James:Awesome. David, what's your wrap up comments and thoughts?
David:Well, I think it's a lot of things we need to actually teach and learn in the industry to fully understand AI again, today we have the first one is AI I think we need to be composing it as well understanding how we can use it. And I think what is clear is this, there's a lot of learning in teaching, we need to do at the executive level to fully understand what is the value? What is the context, and basically how we can help the industry to capitalize as much as possible, the power of artificial intelligence? And what's your thought James? What is your your view on the on the topic?
James:Yeah, look, you know, I think what it is, is we're seeing really a fundamental shift in the way that vehicles will be used that the way that people will communicate or operate with their vehicles, and really the desire of customers, what they want in their vehicles and when they want it. And, you know, as Roger said, as our, as we've started to be ruled by our phones and our computers, and whatever else, just as much, we're going to have that same thing with our cars as well, in terms of what we want out of them and how we want to interact with them. And I certainly think that, you know, using technologies like AI and thinking of them as tools is certainly a really important way to improve the customer experience and really, the direction in which, you know, ultimately vehicles will will end up going in future mobility. So, you know, I think it's absolutely important conversation to have and thank you, Roger. Really appreciate your time today. And, you know, being able to educate us more on AI. And David, thanks again, this was another great discussion that we've had.
David:Thank you, Roger, very much.
Roger:Thank you for having me.
David:A pleasure to have you and hopefully we can do for all of our listeners. Hopefully we can bring you all of these discussions, interesting topics and discussions in our next episode of Byte Off. I hope you like, if you like what we're doing, and then click the button to download. Follow us to more insight content on technology and the future of mobility. Thank you very much.
Roger:Thank you