On this special episode of Data Couture, Jordan is recording while driving from Colorado back to the Midwest. The 12 hour drive has inspired the topic this week: Self-Driving Cars!
Specifically, we talk about three important issues facing the implementation of autonomous vehicles globally. The first is the ability for the artificial intelligence and deep learning algorithms to accurately see the world to be able to drive and avoid obstacles in our complicated roadways.
The second problem exacerbates the first when it comes to computer vision experiencing inclement weather. The current technology makes it nearly an impossible feat. The final problem is our (humans) unpredictable relationship with roadways. How many of us have glued our eyes on our phone while crossing a street?!?
Get ready for some roadway self-driving problems, solutions, and problems to the solutions!
To keep up with the podcast be sure to visit our website at datacouture.org, follow us on twitter @datacouturepod, and on instagram @datacouturepodcast. And, if you’d like to help support future episodes, then consider becoming a patron at patreon.com/datacouture!
Music for the show: Foolish Game / God Don’t Work On Commission by spinmeister (c) copyright 2014 Licensed under a Creative Commons Attribution (3.0) license. http://dig.ccmixter.org/files/spinmeister/46822 Ft: Snowflake
Welcome to data couture the podcast about data culture at work at home. And on the go. I’m your host, Jordan Bohall. To stay up to date with everything data controller, be sure to like and subscribe down below. Furthermore, be sure to follow us around the internet at data to her pod on Twitter, at data couture podcast on Instagram, and at data couture pod on Facebook. Of course, if you’d like to help keep the show going, then consider becoming a patron at patreon. com forward slash data couture. Now, no onto the show,
Welcome to data couture. I’m your host Jordan. And today we have a very special episode. Namely, I’m coming to you live from my car and as I drive back from Colorado to the Midwest. And given that I’m going to be on the road for 12 to 13 hours today, I thought what better time to talk about something that I currently wish I had, namely a self driving car? Because why should I be driving when the car can do it for me? Well, turns out there are a lot of challenges posed for any automaker trying to actually implement self driving vehicles because currently, they rely on humans as emergency backstops. But, frankly, if there’s an emergency, chances are I’ll be lying down.
And I don’t know I’m not sure if we’ll call it a driver’s seat. But I’ll be lying down watching a movie or sleeping or listening to music or listening to podcasts. And I probably won’t be paying attention to the road. So today we’re going to talk about some of the data problems involved with self driving car tech algae will start in the first act by talking about some of these problems, then in the second act, we’re going to go on to the automakers solutions to these problems before the third act, talking about maybe why these solutions won’t work. Stay tuned.
Welcome to the first deck, let’s talk about what a self driving car even is, for those that don’t know, self driving cars pretty well named, it’s a car that drives itself. However, it’s much more interesting than that not only can it drive itself, but it can avoid obstacles that can actually go where it’s intending to go by the maps that you plug it in. So instead of looking at your GPS, and then trying to figure out how to get there yourself, the car can get there for you.
Beyond that, it can avoid any sort humans in the road. For example, think of a cross stop and somebody crossing it, you didn’t see him coming, well, you might have an accident, that might not be so great, but a self driving car, theoretically will be able to notice that person crossing the street and be able to stop in time to avoid that accident. Similarly, self driving cars have the promise of potentially being able to completely revolutionize the way we think about cars. For example, instead of thinking about air travel to get somewhere, we can just happen or self driving car or the self driving car that we share with who knows how many other people plug in the coordinates and, you know, take a nap and by the time that we wake up from our very restful nap will be in that location.
So this technology has the opportunity to disrupt not just playing travel, but train travel, or buses or any other sort of common public transportation because it can do it for we can automate this task in our lives. Now, what are some problems with self driving cars? Well, for vehicles to drive by themselves, they need to understand the world around them. And they need to understand the world around them, maybe better than human drivers so that they can navigate their way through streets or be able to stop at stop signs and traffic lights in order to avoid hitting those obstacles. Like other cars and pedestrians, like I mentioned earlier. And frankly, the closest technology that will allow these sorts of cars to make sense of their surroundings, is a branch of artificial intelligence known as computer vision. In computer vision is, again a part of AI that allows software to understand contents of images, videos, that kind of thing. And so, frankly, modern computer computer vision, it’s really made some leaps and bounds.
There’s an area of AI called Deep Learning, which we’ll talk on or talk about in a future episode. And deep learning allows it enables the computer, it allows the algorithm to recognize different objects and images through a process of both examining and comparing millions and millions of examples. And then determining those patterns was visual cues that defined each object. So think of when you’re doing a capture on a various type of website to prove that you’re not a robot, you and the caption ask you, Oh, where’s the stop sign in this picture, or point out cars in this picture or point out crosswalks in this picture. We are effectively crowdsourcing ways to enable computer vision to be able to be trained so that the deep learning algorithm is able to recognize those different objects. And so while this is very efficient, this type of classification tasks, namely using deep learning, deep learning suffers from a ton of serious limits. And it can fail in lots of unpredictable ways. What does this mean for you, this means that your self driving car can crash into a semi truck and broad daylight, perhaps accidentally hit that pedestrian.
And frankly, the current computer vision tech used in autonomous vehicles is also very vulnerable to adversarial attacks, where, you know, hackers can manipulate the API’s channels so that it forces it to actually make a mistake. And so, as you can see, that’s a major problem. The second problem is actually creating these maps and making it available enough to all self driving cars that they can operate effectively, so they can drive by themselves. And so what does that mean? That means on those perfectly sunny days, maybe the visual software, the visual inputs that have been used to train the self driving car works flawlessly. But I don’t know about you, because I live in an area where it snows and it snows a lot.
What happens when it snows a lot, visibility goes way down, it goes down to the point where you can’t see 10 feet in front of your face, whiteout conditions, so to speak. other situations include heavy rain falls, and in those rainfalls computer vision might fail. Because why the camera that’s used to see the road to recognize the various patterns of objects so that that algorithm, that car has the right cues in order to avoid those obstacles? Well, it’s not going to have those cues, therefore, the self driving car is going to fail. And that’s a huge problem, the problem of truly putting a map allowing the algorithm to see the world and way that humans can see the world so that we can avoid serious issues. The last problem that I’m going to be talking about today isn’t necessarily a problem with the car or with the self driving aspects of the algorithms or the computer vision or as we’ll talk about it in a moment, namely, the usage of LIDAR, various other types of radar technologies. know it’s humans, humans are a problem for self driving cars. Why are humans a problem for self driving cars?
Well, I don’t know about you, but have you ever been in a city or a small town or the country? And have you seen somebody cross the road and look down at their phone the whole time, not paying attention to anything around them? Well, of course, just being glued to our phones isn’t a problem. But the larger problem of rules that we have for interacting with the street as pedestrians interacting with active roadways, and really that’s a problem for self driving cars. Because if humans are unpredictable, it’s unclear that any sort of learning that we can do for a deep learning or neural net algorithm is actually going to be effective. And so our third problem is humans for self driving cars. Now in this next section, we’ll talk about some solutions that the industry has come up with for these three problems.
Welcome to the second part of the show, welcome back to this drive time episode. Now in this second section, we’re going to talk about a few solutions that the auto industry is using towards the first three problems that I mentioned earlier in the show. And to remind the listener, those issues are one being able to to see the road and understand the conditions around the road as well if not better than we as humans can do, as well as to avoid any sort of malicious intent from hackers or other nefarious actors. The second one is understanding the road when these computer vision technologies and LIDAR technologies can’t see the road be that through inclement weather, or perhaps just your bumper or the camera itself is dirty, right. And the third problem is us humans being very unpredictable for the self driving cars. Now, before I get into the solution to the first and second, because one solution is or at least seems to be very promising for knocking out the first two problems. Let me first explain what LIDAR is LIDAR, which is used from various companies like Uber and Google means light direction and raging ranging Aryan g i n g. LIDAR is an evolving domain. various companies, like I said, are using the technologies to patent it.
But LIDAR works by sending millions of laser pulses, and an array of directions of various very slightly different directions to create a 3d representation of the area surrounding the car. Based on the time it takes for the posts to hit an object and return think of echolocation but much fancier right. Now, LIDAR, as well as the direct computer vision used by companies like Tesla, they all fall down to the similar sort of problems that I mentioned for the first and second problem, namely, being able to see the road and the conditions well enough to predict what’s coming down the road and let the car avoid or continued on a normal path. And then the case of inclement weather, that doesn’t work so well. So what is the solution that’s proposed? Well, it’s one interestingly enough, proposed by Ford, I think, and that is to use the blockchain as a way to effectively push the 3d maps of every environment to the car itself and a very cheap and efficient way. Now, we’ll be talking about blockchain and a future episode.
But blockchain and kind of quick and dirty way to explain it is a simple but very, very clever way of passing information from A to B, and a fully automated and safe manner. So what does that mean? That means a blockchain is a network has no central authority. It is therefore a democratized system.
It is a timestamp series of records of data managed by clusters of computer routers that are not owned by any single entity. And so this blockchain technology, one gets around this problem of nefarious actors, potentially driving down the road next to you and causing your car to crash because the data is stored on hundreds, thousands, millions of different nodes around the world. And so there’s no one way to attack a single specific self driving car going down the road. And so it prevents that nefarious actor problem.
And the second is that the blockchain allows all of this data with very low latency and very low costs both economic costs, but also computationally to be fed to the computer. So that the computer on the car, the algorithm in that car is not relying solely on the vision of the cameras placed in the car, or the various radar systems or LIDAR systems to be more exact that exists on the car. So if there’s inclement weather, then this blockchain technology can effectively eliminate the need for the car to see as far as it would have to otherwise. Now, the second solution, the solution of people, this is much more difficult. I don’t know about you, but given the I am in the data profession, and I am obsessed with tracking basically, everything I do, and all the gadgets and technology around me what they do.
I of course, I’m glued to my phone, my wife regularly asked me what I’m doing. And I tell her that I’m checking my data, checking my stats, making sure that I know what’s happening in my environment at all time, whether with me with this podcast, with my cars, whatever it is, it’s all connected to me. And it gets fed to me very neatly through my phone, which very rarely leaves my side. Now, in the case of self driving cars, the solution to solving this human problem is more or less automating the rules for human involvement and roadways, that means creating very strict rules for how we approach roads, how we cross roads, and, unfortunately, making it a carrot and stick scenario. And so, carrot being well, we get to have all this awesome technology, and isn’t that great for humanity and society at large. But the stick part is, if you get out of line, and if you jaywalk, like we all do need to be harsh punishments.
So while it’s not the prettiest solution, and it’s not the most socially friendly solution, because you can just imagine all of the various problems that come with us. The way that we can have self driving cars, automated vehicles, whether that be semis delivering our online friends purchases, our cars taking us from A to B are automated vehicles, autonomous vehicles taking us from a across the country to be like I wish I was doing right now sleeping in my self driving car. That means we’re going to have to give up a little bit of our liberty when it comes to our freedom when it comes to interacting with roadways. Now, that doesn’t sound like the happiest solution. So let’s move to the third section where we talk about issues with these solutions to the problems posed in the first see you soon.
Welcome to the third act of data tour on the road.
I guess that’s what I’m calling it now. Now in the first section, we mentioned three problems for self driving cars, namely, computer vision causes issues when it comes to mapping out the world and normal conditions, and also causes issues when mapping out the world and income at weather conditions as well as the problems of pedestrians crossing the road. So those three issues solved in the second section by the use of blockchain as well as by having more stringent rules on how pedestrians interact with the roadways and byways of this country or any other country. Those seem great. But of course, there are problems with these solutions. Let’s first talk about the problem with LIDAR that say, Google and Uber are employing light art has a fantastic technology and extraordinarily accurate and safe from malicious actors. However, LIDAR is extraordinarily expensive and its current form. So unless there are advances in science that come very quickly, LIDAR will make autonomous vehicles out of reach for the vast, vast majority of the population around the world.
Now, that’s fine. Let’s just use blockchain. You say, Okay, well, blockchain itself also has quite a few issues. And one challenge is, well, to create a black blockchain, the size needed to enable autonomous vehicles in the world across the world. Well, it’s gonna require a massive Internet of Things ecosystem. And again, we’ll be talking about IoT or internet of things in a future episode. But the IoT ecosystem, if it were large enough, would create lots of very strong connections, and it would improve the abilities of autonomous vehicles to the point where they became more than autonomous vehicles are just simply vehicles that are autonomous. However, this requires massive coordination from thousands and thousands different parties, including stakeholders such as regulators, the legal system, vehicle manufacturers, high tech companies, the general population, the electric grid, any sort of payment processing system, the list goes on and on and on. And so the technology, technological advancements, which are currently possible and currently being employed using the blockchain are great. But they’re only effective insofar as they exist, where that Internet of Things ecosystem is deployed.
That means that until we create a massive funding and massive, dedicated effort towards making these IoT devices, these Internet of Things devices, ubiquitous across every roadway. And by the way, across the world, the time was vehicles won’t be ubiquitous, like I personally want them to be currently traveling down I at and I’d love to have all the stuff mise autonomous on the far right side of the tunnel is vehicles going the speed limit, or somehow hopefully, there’ll be a hack. So you can go slightly faster than the speed limit. I’m not saying you should speed be safe. But when they’d be nice and switch, create a very nice flow, and then it pulls up to st electric charging station because of course, the future of cars is electric. Does that for you hop out maybe you know, get some food or use the restroom or maybe go for a walk whatever it is take your dog out. And then you get back in the car and on you go. Fabulous. But like I said, in order for this to work, blockchain has to become ubiquitous all the IoT devices needed to enact blockchain need to become ubiquitous.
And then another issue with blockchain technology is it’s going to raise a ton of legal questions which aren’t answered yet, of course, because of technology as well ahead of any sort of legal process. So for example, one legal issue will certainly be assigning any sort of responsibility for glitches that happen in the computer system or various technical errors. is the problem on the human who might have been a slept or asleep at the time? Or is it the responsibility of the automakers responsibility of the IoT device, which itself might have pushed out or you know, so that’s going to have to be worked out? And of course, the courts take quite a bit of time. Now, what’s wrong with creating stranger more stringent regulations on the way we use humans interact with the world, specifically around the roadways and byways?
Well, it comes down to how much we’re willing to give up for a convenience that maybe not all of us are going to use. So for example, maybe we have to start punishing j walking by a very massive fine, say thoughts, thousand dollars? Well, what if that person jaywalking has never used an autonomous car doesn’t plan on using autonomous car? Maybe they live in a city and don’t need a car at all, for that matter? Is it fair to them, that they should be punished so harshly just because perhaps a relative, few of us want to enjoy the perks of an autonomous vehicle? So you see, that’s just a simple example. But more and more we get involved in this, the fewer and fewer freedoms we have? Well, I could see that as a major issue. Moving forward. Let me know what you think. Leave some comments down below and I’d love to talk to you. See you soon.
That’s it for the show. Thank you for listening. And if you liked what you’ve heard, then consider leaving a comment or like down below. Stay up to date on everything data couture, be sure to follow us on Twitter at data couture pod to consider becoming a firstname.lastname@example.org. forward slash dedicate to music for the podcast. It’s called foolish game. God don’t work on commission by the artist spin Meister used under a Creative Commons Attribution 3.0 license, writing, editing and production of the podcast is by your host, Jordan Bohall.