Welcome to tech Stuff, a production from I Heart Radio. Hey there, and welcome to tech Stuff. I'm your host, Jonathan Strickland and how the tech are you? Now? Before we get started with today's episode, I wanted to let y'all know about a new way to get in touch with this podcast, and I figured it'd be best to do it up front, because I suspect some of you folks turn off the show when you hear me wrapping up. Shame, shame, Shane. Now, if you download the I Heart Radio app and navigate to tech Stuff, which you can do by searching text stuff and then selecting the entry that's under podcasts, you will see that there's a little microphone icon on that page. And similarly, if you click into any individual episode, you'll see a microphone icon there. Now, if you tap on that icon, you can leave a voice message of up to thirty seconds for the show. So the microphone on the main tech Stuff page goes into a general folder. This might be like if you wanted to suggest an episode about a specific topic. If you leave a message on a specific episode, well it tags that episode and I'll still see it right away, but it means that I'll know which episode you're referring to, and maybe you want to leave a comment that's specific to that episode. Maybe it's an addition, it could be a correction, it could be a cool I didn't know that? Did you also know this? It could be anything like that. Now, future episodes might include voice clips unless you specifically ask me not to use it, in which case I will not use it. But if it's something where you're leaving a fun message that could prompt an episode, I would love to be able to start including those in tech stuff. It's a great way to send suggestions for topics, to comments on episodes, to corrections I mean, I do sometimes get things wrong, or maybe you just want to say howdy. Also, in case you're curious, the only people who have access to these messages are myself and my super producer Tari, so there aren't any other people listening to these in advance. Also, please, if you do use this, be civil, even with the corrections, because I do have feelings and I look forward to hearing from you. Podcasting is a fun medium because listeners can often feel like the host is talking directly to them. I know a lot of people who listen with headphones on and that's a very intimate experience, right, I'm speaking directly into your ears in that case. I know I have experienced this in the past when I've listened to podcasts. In fact, there are podcasts I listened to where I do feel like I'm just a silent person in a room, which honestly isn't very much like me. But you get what I say. And now this tool lets you talk back to me. And because it's just you on the microphone, I'm not capable all of interrupting you, which is a big bonus over having a real life conversation with me. Anyway. Once again, that is the I Heart radio app with the talk back feature, that little microphone icon that you can just tap. But let's get on with today's episode. And I really wanted to cover this topic and give a bit of an update because I feel it's one that is often misunderstood and frequently misrepresented. Um and that is the different levels of autonomous driving and the state of autonomous driving where we're at. I occasionally do these updates because I think it's very easy to get lost in marketing and pr speak and even some you know, political movements that suggests we might be further along than where we really are now. First of all, let's get this out of the way. Developing autonomous vehicle technology is a good idea. Here in the United States, the National Highway Traffic Safety Administration or n h t s A tracks the number of motor vehicle traffic fatalities year over year. The number of deaths due to vehicle accidents is always in the thirty thousand range or higher. So that means more than thirty thousand people here in the United States die every year due to traffic accidents, and if you just skim recent figures, that really reveals something counterintuitive. The year with the most traffic fatalities over the last decade is twenty twenty. Now I should add that I could not find an estimate for one, not for the full year, only for part of the year, so one very well might have taken the title away from twenty twenty. But yeah, the first year of the pandemic saw the most traffic fatalities in more than a decade. And you might imagine that fatalities for that year would actually be down due to more people staying at home, But you're wrong. The total that year was thirty eight thousand, six hundred eighty estimated deaths. The next highest year would be back in sixteen that had thirty seven thousand, eight hundred six deaths, almost a thousand fewer. The n h T s A also figures out that the fatality rate per one hundred million vehicle miles traveled is a good metric. Uh. This is v MT, which essentially tells you how frequent do these accidents occur, how many millions of miles are are happening in between these fatalities. Keeping in mind that's across the entire United States, so it's taking into account all the vehicles traveling across the country, and again had the highest frequency over the last decade with an astonishing one point three seven fatalities per one hundred million vehicle miles traveled. The next highest on that goes all the way back to two thousand eight with one point to six. So keep in mind like that means more than a decade of innovation and improvements in driver assist features and things of that nature ended up not making a difference because we saw more fatalities per miles traveled than ever well at least since two thousand eight, Because that's as far back as as the the sheet I was looking at wind So people were driving less in but they were dying in accidents more frequently. And it didn't get better in one either. During the first nine months of one, traffic fatalities rose by twelve percent. So within those first nine months there were an estimated thirty one thou seven twenty deaths. Again just nine months, not a full year. Now, I do not have more recent data to pull from, so I'm not certain how the year ended out, but that statistic is terrible and sobering. Here's another bad statistic. The c d C says that road traffic crashes are a leading cause of death here in the United States for people aged one to fifty four. Now, these deaths clearly have a massive impact. Beyond the obvious tragedy of losing someone or multiple someone's in a crash. There's the impact on that person's family and friends. That's an incredible emotional impact on them. If that someone who died in the crash were employed, well, there's an impact on their job and on the economy as well. There's this overall economic impact due to fatalities and accidents in general. The National Safety Council estimates that the average economic cost of a traffic fatality is one point seven five million dollars per incident. Now, this takes into account everything from medical expenses, to motor vehicle damage, to insurance costs, to wage and productivity losses, and more. So, if we take this average figure of one point seven five million dollars and we multiply it by the thirty eight thousand, six eight estimated deaths in twenty twenty, we get an economic impact of sixty seven point seven billion dollars. That's the cost of those fatal traffic accidents. When you take in all traffic accidents, by the way, the amount gets over two billion dollars. So there are a lot of reasons we would want to reduce or eliminate those traffic fatalities. I mean, first and foremost, obviously, we don't want people to die. We would rather not see that happen. We would rather those people go on to lead healthy, productive lives, surrounded by friends and family, and we would not want to see that end. Less importantly, we would not want to see that negative economic impact. And we also would rather not have the thousand other problems that come with traffic accidents, such as the effect on traffic congestion and that sort of thing. And there's one more important fact to consider. According to research, at least of all traffic accidents are caused primarily by human error. In other words, it's not a mechanical failure. It's not that a vehicle suddenly lost functionality in some way and that that led to the accident. More than are caused by someone making a mistake. And that's at least Some research pushes that even higher. The highest I have seen suggests that eight per cent of all accidents are call by human error. Um, so somewhere in that range is probably the truth. Right, That's still enormous, and that's where the sales pitch for autonomous vehicles comes in. Because what if you had a computer controlled vehicle capable of traveling at a safe speed and a safe distance from other vehicles. It would be able to pay equal amounts of attention to all areas around the vehicle. It wouldn't just be paying attention to wherever its eyes were, because it would have eyes all over. They could react much more quickly than any human could, and it could take action that could avoid an accident. You would remove human error as a contributing factor for accidents, and since we know human error accounts for more than of accidents, that would, by extension, eliminate the vast majority of accidents on the road. If you had enough of these vehicles on the road, perhaps not replacing all of them, but having a significant percentage of vehicles being autonomous, you would virtually eliminate traffic fatalities. That is the dream scenario. But there are a lot of things standing between us and that vision. I will go into that after we take this quick break. Okay. To get to that vision of the future where autonomous vehicles are taking us everywhere we need to go and traffic fatalities are a distant memory, we have to get the tech right first, and we also have to get enough vehicles out on the road to make a difference. One autonomous vehicle isn't going to change the world. And I've seen a lot of different research about this as well, some researchers saying that we might need to reach a saturation point of about one third autonomous vehicles to two thirds human driven vehicles to start seeing massive changes and things like accidents and traffic congestion. And then, of course, beyond that, we'd see even more improvements, not necessarily at the same dramatic pace, like it might level off more where we get sort of a diminishing return situation. But when you're talking about saving lives, diminishing return doesn't really have the same meaning, right, Every life saved is significant. Now, when it comes to autonomy, at least for vehicles, there is a spectrum, you know. There. Autonomy is more like guidelines, as the pirates would say. The Society of Automotive Engineers, which is a quote globally active professional association and standards developing organization end quote, has identified six levels of autonomy, ranging from zero to five, and it is ascending in autonomy. So let's go through those now. At levels zero, you have no driving automation really of any significance. I used to think that level zero just meant that you had complete manual control of the vehicle and there were no driver assist features at all. That's not the case. That was my mistake. That was my misunderstanding of this designation. Uh, you could have cruise control and it could still fall into a level zero category. So basic cruise control would not be enough to move a vehicle out of level zero designation to level one because basic cruise control will maintain a set speed, but it won't change it. Right, if there is a vehicle that's ahead of you, the standard cruise control isn't enough to detect that vehicle and change the speed. That would have to be adaptive cruise control. That's different basic cruise control. You're still at level zero. Back in the day, I used to think that that would at least push you to level one. No, not, according to the s A The s A E says that any driver assist features are in a level zero car are limited to quote providing warnings and momentary assistance end quote. So you can have like a lane correct feature, but if it's just momentarily shifting the car, that still is considered a level zero autonomous vehicle. Automatic emergency braking, lane departure warnings, all that kind of stuff I'll just count as minimal driver assist. So any vehicle having those features but nothing more extensive than that, would still be level zero. So that's a lot of the vehicles that are on the road today, the vast majority of them, in fact. But let's go on to level one, because we do have level one and even level two autonomous vehicles on the road. So a level one autonomous vehicle has the ability to take over some duty that the driver would normally do, but only one of those duties, and the human driver would have to handle everything else. So, in other words, the feature might be able to handle something like acceleration and braking, or it might be able to handle steering, but it would not be able to handle both of those, So whichever one it handles, the human driver would have to deal with the other. One adaptive cruise control that I mentioned earlier, in which a vehicle handles acceleration and it will ease off if it detects that it's approaching the rear of a vehicle in front of it. That's an example of a level one autonomous vehicle feature. It's handling acceleration and braking, but it's not handling steering. The human driver still still has to steer the vehicle in that scenario. Now, if we move up to level to autonomy, we're now at a phase where the vehicle can handle both the acceleration and breaking activities as well as steering the vehicle at the same time. However, a human driver must supervise these and also intervene when necessary, as the vehicle will not always be capable of avoiding an accident when something out of the ordinary happens, or even something remotely unexpected happens. So this would be a vehicle that can do both adaptive cruise control and lane centering operations at the same time. And honestly, this is where we're at with most of the really sophisticated passenger vehicles that are on the market right now. Tesla, even with its so called full self driving feature, which is not full self driving, and Tesla darnwell knows it. I have a longstanding issue with Tesla and it's naming conventions. Calling its first driver assist support sweet as autopilot when it's not really autopilot really upset me. Full self driving is even worse because it the name implies that the car just completely drives itself. That is not the case. It still requires human supervision, and Tesla says this to drivers. It says that you still have to supervised the vehicle, that you can't just take your hands off and go to sleep behind the wheel. You have to maintain supervision and be ready to intervene. But yet they insist on calling their features full self driving. I just feel like that's misleading. Anyway, that is still a level to autonomous feature. It means Tesla is on the level two on the autonomous scale. It hasn't hit level three yet. And from levels zero to two, the job of monitoring the driving environment falls on the human in the car. They are responsible for making sure that the environment is safe. Uh that for at least whatever operation the car is in at that time, and you cannot offload that to the vehicle. The vehicle is not capable of shouldering that accountability. So ultimately the person behind the wheel is still responsible for their own safety and for avoiding accidents. But now let's talk about levels three through five, in which the vehicle's automated system is in charge of monitoring the driving environment. So in this case, the car is responsible or the vehicle I should say not just car is responsible for monitoring what's going on around it and then using that information to make important decisions. So level three is called conditional automation, and that kind of clues you into what's going on. So the vehicle is capable of operating autonomously under certain conditions, but not all conditions. So unless all autonomous conditions are met, the vehicle will not operate an autonomous mode. A level three autonomous vehicle will be able to make certain environmental decisions, such as accelerating past a slower moving vehicle or maneuvering through a traffic jam, which typically involves lots of starts and stops and potentially lane changes. There are a couple of vehicles that have received international acknowledgement as attaining level three autonomy. There's a car called the Honda Legend, which is only available as a least vehicle in Japan, so unless you're in Japan, you're not likely to see one or experience it. Mercedes Benz has received a regulatory approval to manufacture level three autonomous vehicles that have its drive pilot system installed. It's recently been updated and has met international standards to be considered a level three autonomous system. So with a level three vehicle, human intervention isn't needed for most situations. It may still be required in edge cases where things are unusual. Now, at level four autonomy, an autonomous vehicle would in theory be able to react to most circumstances without the need for human intervention. It would not ask for human intervention, it would just react. The vehicle would respond properly, able to discern what was happening and then take the appropriate response. For example, such a car would presumably be able to distinguish between a trail of gosling's attempting to cross the road versus some leaves being blown by the wind, and either it would slow down to let the little goslings get across, or it would just continue without worry because it's just some leaves. And yeah, just saying that filled me with anxiety because I think goslings are cute and the thought of a car just driving through a crossing animal really really upsets me on a deep level. However, a level for autonomous vehicle would still be a conditionally autonomous vehicle, so there would still be some conditions under which this vehicle would not operate autonomously. And because this is the level that the s a E considers to be appropriate for robo taxis, it would mean that under a certain con asitions you wouldn't be able to get a ride in a robotaxi. Because if if conditions are wrong for autonomous operation, then they won't they won't pick you up, right, They won't work. Those conditions could include things like weather events like let's say it's really pouring down rain and because of the heavy precipitation, there's a fear that the vehicle's sensors won't operate properly so it's unavailable. Or it could involve things like geo fencing. That is, the vehicles can only operate within a certain geographic area. Typically we're talking about a specific town or specific city, or specific lane between two cities, and the cars systems will not allow the vehicle to venture beyond certain borders. In fact, we're gonna talk about how geo fencing is one way that companies are looking to tackle very tough engineering challenges in an effort to bring robotaxis to various places. Also, at this level, things like a steering wheel and or pedals like an acceleration and brake pedal. Those might not even be installed in a level for autonomous vehicle, so there would be no controls for you to take over should you feel the need to intervene. You would not have that option. You might have a button to press that was like an emergency stop feature where the car would then pull off to the side and come to a halt, but that might be it. So a level for autonomous vehicle should be able to handle pretty much any situation that could pop up under its operational conditions. And finally, we have level five autonomy. Now at this level of vehicle under any conditions would be able to operate autonomously. So this is the level where a vehicle could go anywhere and operate safely no matter what conditions might be present. Such a vehicle might have no other controls in it and handle absolutely everything by itself. We are nowhere close to achieving level five autonomy right now. So when we come back, we'll talk about some of the things that are holding us back from full autonomy, and some of them are technical and some of them are social and political. But first let's take a quick break. Okay, let's talk limitations, and we'll start with the technical. We've seen some pretty dramatic improvements in sensor technology over the years, but the fact is that under some conditions sensors can have trouble detecting the environment effectively, particularly when you're talking about the sensors being mounted to a platform that's moving at several dozen miles per hour or kilometers per hour. So this can include conditions like dense fog or heavy precipitation. Like I mentioned earlier, it would be pretty dangerous to ride in an autonomous vehicle that ends up mistaking fog for a solid surface. For example, let's say that whatever technology the sensors are using, the signals are bouncing back because it's hitting the fog or the precipitation. A car like that might start breaking suddenly, like applying the brakes suddenly, and that could pose a hazard for vehicles that are behind you. And since their limitations aren't a hypothetical either, we've seen some tragic cases in which driver assist systems failed to detect a danger and that led to loss of life fatalities. I mean back in famously, a Tesler driver named Joshua Brown was on a Florida divided highway. So that's a highway that has a couple of lanes going in one direction. Then you typically have a divider, a median of some sort, and lanes going in another direction. Uh and back in ten, Joshua Brown's on a Florida divided highway. He had autopilot in his test engaged and as he was traveling down there was a semi truck that was pulling out of a driveway that was perpendicular to the highway. The semi truck was crossing Brown's lanes of traffic in an effort to make a left hand turn to go the opposite direction down the highway, so it was pulling across the lanes that Brown's vehicle was in, uh the Tesla. Brown's Tesla failed to detect that there was an obstacle in the way, so the Tesla did not slow down and ended up colliding with the truck and Brown died in that car accident. Now, back then, Tesla was relying on systems that were provided by a company called Mobile I, but Tesla subsequently ended that partnership and then began to develop new sensor technology in house. So the autopilot of today's Tesla's works on a totally different techno logical platform then the one back in two thousand sixteen. Because Tesla recognize that this is not something that it can it can support and still expect to be able to sell vehicles that have this autopilot feature in them. Now that being said, we saw a near identical tragedy occur in two thousand nineteen when Jeremy Banner, also going down a divided highway, also in Florida, also driving a Tesla that was in autopilot mode, also collided with a semi truck that had pulled out to make a left hand turn. It was, in many ways an identical scenario to what we saw before. So that leads to a question, how could two different autopilot systems make the same fatal error? And a lot of this potentially has to do with the types of technologies we depend upon to do certain things. So a lot of the anti collision technology we find in vehicles real lies on Doppler radar. Let's talk about what that means for just a second. First radar, Now, that involves sending out a radio signal and then detecting the echoes of that signal as they bounce back to the source. So your radar has an emitter and it has a sensor. The admitter sends out a radio signal, the sensor detects the returning radio signal after it's bounced off something. And when you measure the amount of time it took for a signal to go out and then bounce back to you, that tells you how far away you are from whatever it is that you're beaming signals at, which is pretty simple, right, But let's talk about the Doppler part. Now, if you've got two objects that are standing still, well, that means the signal you send out and the signal you get back are going to be pretty much the same wavelength and frequency because both you and the object are at a standstill. But what if one of you is moving toward the other, Well, in that case, the signal that comes back to you is going to be compressed. It's going to be shorter in wavelength and higher in frequency than the signal you sent out. It's being pushed by whatever it is that's moving towards you, or, in the case of you moving toward it, the fact that you're getting closer. So this tells you not only are you within a certain distance of another object, it tells you that you are getting closer to that object or that object is getting closer to you, depending on your point of reference. Now, for those of us who are are able to hear, this is something that we can experience out in the real world without any radar at all. If you've ever heard and approaching emergency vehicle that's running its siren. You probably noticed the sound of the siren changes when the vehicle passes you, So as the vehicle is approaching you, the pitch of the sirens sound is higher, and as it moves away from you, the pitch goes lower. And that's because the actual sound wave is being compressed as the vehicles moving towards you, and it's expanding as it's moving away from you. The vehicle pushes those sound waves closer together as it's coming at you and stretches amount as it's going away. That's when why the pitch is higher one way and lower the other way. The same thing happens with stuff like radar and radio signals, and obviously it gets more complicated if both you and the other objects are in motion at the same time. But these systems work best when everyone is moving in the same direction of travel, right, So it works really well if you're behind another vehicle and that vehicle slows down, then the system is very good at detecting that. But in a case where a truck is pulling across lanes of traffic, where the truck is moving perpendicular to your direction of travel, it doesn't work as well. It doesn't work as well on stationary objects, and a truck would appear to be stationary as it was moving laterally across your lane of traffic. Now, depending upon the lighting and the time of day and autonomous vehicles, camera system might not do much good either. It might misidentify a truck as being like a road sign, and a road sign hangs over the highway, it's not in the way of the highway, so the system wouldn't think of it as being a potential threat. Or it might interpret the side of the truck to be the sky, depending on those lighting conditions, and that would mean that the vehicle safety features would not initiate, and that could lead to tragedies like we've mentioned here on this episode. Now, I don't mean to suggest that these technical problems are impossible to solve. I don't think that's true, but I do think it's going to take a lot of work, and it might require technologies that for the moment are prohibitively expensive, meaning that yeah, we could create a more reliable autonomous system, but then the cost of the vehicle would be so high that it wouldn't be practical from an economic standpoint. Some of these systems. These components like LDAR, some of them can cost as much as a brand new car does all by itself. And I'm not talking about the cheap car either. So when you start looking at that, you have to then take the economic factors into consideration and say, doesn't make economic sense to develop a system that relies on these components, because will we ever make our money back? Like would you be able to sell the vehicle at a cost that would make sense, or would no one buy it because it would be far too expensive. Even if you're talking about, say a taxi service, maybe a taxi service wouldn't buy it either because they'd say, the likelihood of us making our money back with cars that are that expensive before we have to replace those cars is so low that it doesn't make sense to go into it. That could be a real obstacle, and as you know, more to do with the cost than the actual technology. I also think that a major problem with autonomous vehicles has been in messaging and in hype companies in the autonomous vehicle business, not just Tesla I. I heap a lot of abuse on Tesla for this, but it is by far not the only company to engage in these kinds of things. But I feel like a lot of the leaders in that space and made some really aggressive promises that in hindsight were impossible to follow through on. At least they were impossible given the time frames that were being tossed around. There were folks suggesting that we would have cities filled with autonomous vehicles able to operate in all conditions by well, it's two and obviously that didn't happen. And no, it wasn't just because a pandemic really messed things up. Though that definitely didn't help because it disrupted everything from workflow to supply chains. But even with sophisticated sensors, you then have to consider the decision making factor and take add into account. Now, at the moment, autonomous vehicles are for the most part independent computing islands. It's like a personal computer that isn't connected to the internet. So all the technology used to keep the car in autonomous operation typically is just on the car itself. So the car is relying on its own sensors and its own processor to detect the environment and make decisions. That is a tremendous amount of responsibility to put on technology. Then again, it's also a tremendous amount of responsibility to put on a teenager too, but that's a different discussion. There are some proposals that suggest the real future for autonomous vehicles is one in which vehicles are in constant communication with one another and potentially with the infrastructure itself, with the road system that a vehicle would be able to stay in communication with, like a smart city, The road system and the cars on it would both be capable of adapting to different situations. So, as an example, a smart city might have intersections that could proactively change the timing on traffic lights to accommodate changes in traffic patterns and smooth out the experience and avoid long traffic jams. Cars in communication with the smart city infrastructure could plot out the most efficient routes that will save time and energy and adapt to changes that happened on the fly. It might even be able to adapt to routes that cause less wear and tear on the actual infrastructure, giving city planners the chance to address things before they become really big problems. I say that as someone who lives in Atlanta and is very used to the site of giant metal plates laid across the road because potholes have developed and the city has not had time to address it. If you had really smart infrastructure, you could start to detect things before they got to that point of of a problem. So these are really cool ideas, and the improvements in quality of life in those kinds of cities would be enormous. But to make this happen a lot of other stuff has to fall into place. First. Vehicles and cities would need to agree upon a common set of languages and protocols in order to communicate effectively. That alone is going to require tons of work. Typically, we see car companies develop their own in house systems which may or may not be compatible with the systems that are used by other car companies, which could lead to real delays. And seeing a fleet of truly autonomous vehicles navigate the roads, it would be kind of like putting a bunch of people who don't speak the same language into a maze and have them try and navigate it together. The cars would only understand other vehicles that were from the same manufacturer, and that would have limited usefulness unless we all just migrated to the same vehicle manufacturer. But that would turn that company into a global monopoly and uh I humbly suggest we don't do that. But that future in which vehicles talk to each other as well as two road systems into cities will require tons of other work as well. Obviously, cities are going to have to build out smart infrastructure, and that's likely to be expensive and time consuming. We also need really good wireless technologies in place to five G has promise, but we would need it to be more dense and more widespread than it is currently. For some industries like trucking, five G is not a great solution because on long hauls between cities, you're gonna hit routes that are going to have little to know five G support along those routes, and it would be foolish to build out a business that depends upon five G connectivity if that business involves navigating through places that don't have five G support. Meanwhile, there has to be a strong enough business case for the telecommunications industry to build out five G availability to these regions. That's hard to imagine if there just aren't that many people living along those routes, right, if it's not densely populated, then the telecommunications companies are gonna say, well, it's really expensive to build out the infrastructure. We have to put antennas all over the place. If we want to have high frequency five G, the stuff that's really high through put, low latency, that stuff, the five G that we think of when we think, oh, this can replace fiber Internet. You have to have very dense antenna deployment for that to work. Well, that's expensive and if a place is not densely populated, it's a really tough thing to tell a telecommunications company you need to put out the expense to build out that infrastructure even if you only have a few thousand customers in these area, is right, Like, that's that's a tough sales pitch to make to these telecommunications companies. So it's not likely that we're going to see intense five G coverage along some of these places. So five G could work in cities, and it might be a really good solution for the future of smart cities, but that would likely mean we would see autonomous vehicles largely confined to a specific region that geo fencing approach. They wouldn't necessarily be able to travel outside of those regions, not without having a robust enough system on board. The vehicle itself to handle everything when it, you know, kind of loses connectivity. And that's a really big part of it too. The fact is, all sorts of weird stuff happens on the roads, and engineers are not able to anticipate all of them. None of us are. There's just no practical or even possible way to preprogram the right response to every possible thing that could happen on the roads. This is another reason why geo fencing is an important solution, or at least a temporary one, because it limits the operational range of an autonomous vehicle. That also means it limits the number and variety of weird stuff that that vehicle could potentially encounter. It's not gonna eliminate weird stuff. It'll just limited to mostly a subsection of all the weird stuff. Right, Eliminating variables makes the whole problem less complicated. I was gonna say easier, but that's the wrong implication. Less complicated is probably the better wording, because it's still wicked complicated. Then we have to get through all these social components to making driverless vehicles an acceptable reality. According to Alex Kopatinski of Policy Advice around fort of people in the United States do not feel safe in a driverless car. They do not like the idea of getting into a driverless vehicle. I actually think for being uncomfortable is pretty low. I would have expected that more than half of people in the United States worry about their safety if they were to get into a driverless vehicle. Now that's not based on any hard data. That's just my my feeling, which we all know is not really that's not evidence, it's not worth while. I just feel forty feels uh really generous. And that's because whenever a driverless vehicle gets in an accident, we see a disproportionate amount of coverage about that accident. Now, we all know fatal accidents due to human error happen far too frequently. They're happening every day, lots of them, with thirty eight thousand, uh nearly thirty nine thousand happening in lots are happening every single day, and they typically do not receive nearly as much coverage as an autonomous vehicle accident beyond maybe local reports. Your local news might cover it, but and sometimes it might just be something like a traffic report, but you don't get the coverage that you would get if it were an autonomous vehicle. When an autonomous vehicles at fault, that becomes national or maybe even global news. So there is a disparity going on here. Right. You could argue that the autonomous vehicles in in aggregate are far more safe than human driven vehicles, but the news reporting has an amazing disparity there. There's going to be way more reporting on autonomous vehicle accidents than on your typical human caused accident. Still, even at you're still looking at a ton of people who would feel unsafe getting into a driverless car, and whether that worry is justified or not, it poses as a challenge when it comes to policy. See another really big component in the adoption of drive list car technology is getting politic Aisians to support it and to develop laws that create the guidelines around it. Many places lack laws that would cover autonomous driving vehicles. That gap is something to be concerned about. It means that if something happens, like if an autonomous vehicle gets into an accident in one of those places, you would have no framework to establish things like accountability. That's just one tiny slice of how it's important to establish policy so that there is a legal infrastructure around autonomous vehicle operation. That legal structure would in turn inform manufacturers as to certain requirements their vehicles would have to meet, and that would mean that these policies would also shape technology. Wouldn't just be how can we make this technology do the thing we wanted to do, It would be how can we make this technology do the thing we wanted to do within the legal framework. So that's an important thing that has to happen too, And in some places we're seeing that. In other places there's not really any official policy. If the policies don't match each other, that also creates a complication, because what do you do in a case where an autonomous vehicle that's legal to operate in one place ventures into a place where it's not legal to operate because it doesn't fully meet all the requirements. These are tough questions now. Way back in the day, if you listen to tech stuff, you know, I was really super hyped up for autonomous vehicles. I was so excited about it, and that's because I recognized the fallibility of humans. I recognize that we make mistakes, and the fact that technology can detect and react at a speed that is impossible for us to even imagine, meant the potential for this technology to save thousands of lives every year. And based off that very limited perspective, yeah, autonomous cars are phenomenal, but it also ignores all the other things that you have to take into consideration that driving, say fleet, is about way more than just seeing something and then reacting to it in time. And that's really where I was being blind. I bought into the hype and was eagerly looking forward to one when I'd be able to hop into a driverless car and go wherever I wanted. These days, i'm I would say, I'm more cautiously optimistic. I do still think we're going to get to a future where autonomous vehicles will play a more central role in how we get around locally, and in fact, it might it might even cut way back on the number of vehicles we have on the road in the long run, But I think that's going to take ages to play out. Maybe we'll see those visions of a future where parking lots are reclaimed and turned into stuff like parks and whatnot. Maybe we'll see that come true, Because maybe we will see a future where we have a fleet of autonomous vehicles that just kind of are there on demand whenever we need them, and then go somewhere when we don't. That somewhere is a big question mark by the way, that we haven't answered, and that we don't own our own car, right, we just use driverless vehicles as an on demand service whenever we need to get somewhere. It will be priced at a point where it would make sense to do that, where you know it would be at least uh no more expensive than owning and maintaining your own vehicle, and preferably less expensive than that. For that to happen, a lot of stuff has to fall into place, and it goes well beyond the scope of this episode, so I'm not going to go into it. But let's say that all these things are actually able to come true. If that's in fact a possibility, it's gonna take decades, if not longer, for us to get there. We we've just got we've got so much inertia built up that we have to get through that in order to make that future reality. And we have to remind ourselves that to achieve that vision, it's going to require a ton of work and the solution to a lot of non trivial problems. That is not a reason to give up, but it is a good reason to check our expectations and to look at things from a more critical point of view, and to approach things that way. That way, when we're really being critical, when we're really acknowledging the challenges that face us, were more equipped to meet those challenges, to find solutions, and to work through it. If we fall into the trap of being overhyped, like I was back in the day, then we're not gonna realize those challenges until they are undeniable, and when we've wasted time, money, and resources on things that were trivial in the in the long term. So still, I'm still up on autonomous vehicles. I still I'm eager to see them become a thing, not entirely sure that they're going to be an effective thing within my lifetime. I mean, they will exist, they will be out there. I just don't know at what sort of density it will be. But I do think it's a worthwhile pursuit. I think the benefits are undeniable if we are able to solve the problems. All Right, That wraps up this catch up on autonomous vehicles. Hope you enjoyed it as a reminder, if you listen to this show on the I heart Radio app, really, if you just download the i heart Radio app, you can use that talkback feature to leave comments either on this episode or on the show in general, and it can be up to thirty seconds and I'll be able to hear it and maybe even use it in a future episode if you would like that. Also, if you would prefer some other method, the other tried and true way to get in touch with the show is through Twitter. The handle for the show is text stuff h s W and I'll talk to you again really soon. Text Stuff is an I heart Radio production. For more podcasts from I Heart Radio, visit the i heart Radio app, Apple Podcasts, or wherever you listen to your favorite shows.

Welcome to tech Stuff, a production from I Heart Radio. Hey there, and welcome to tech Stuff. I'm your host, Jonathan Strickland and how the tech are you? Now? Before we get started with today's episode, I wanted to let y'all know about a new way to get in touch with this podcast, and I figured it'd be best to do it up front, because I suspect some of you folks turn off the show when you hear me wrapping up. Shame, shame, Shane. Now, if you download the I Heart Radio app and navigate to tech Stuff, which you can do by searching text stuff and then selecting the entry that's under podcasts, you will see that there's a little microphone icon on that page. And similarly, if you click into any individual episode, you'll see a microphone icon there. Now, if you tap on that icon, you can leave a voice message of up to thirty seconds for the show. So the microphone on the main tech Stuff page goes into a general folder. This might be like if you wanted to suggest an episode about a specific topic. If you leave a message on a specific episode, well it tags that episode and I'll still see it right away, but it means that I'll know which episode you're referring to, and maybe you want to leave a comment that's specific to that episode. Maybe it's an addition, it could be a correction, it could be a cool I didn't know that? Did you also know this? It could be anything like that. Now, future episodes might include voice clips unless you specifically ask me not to use it, in which case I will not use it. But if it's something where you're leaving a fun message that could prompt an episode, I would love to be able to start including those in tech stuff. It's a great way to send suggestions for topics, to comments on episodes, to corrections I mean, I do sometimes get things wrong, or maybe you just want to say howdy. Also, in case you're curious, the only people who have access to these messages are myself and my super producer Tari, so there aren't any other people listening to these in advance. Also, please, if you do use this, be civil, even with the corrections, because I do have feelings and I look forward to hearing from you. Podcasting is a fun medium because listeners can often feel like the host is talking directly to them. I know a lot of people who listen with headphones on and that's a very intimate experience, right, I'm speaking directly into your ears in that case. I know I have experienced this in the past when I've listened to podcasts. In fact, there are podcasts I listened to where I do feel like I'm just a silent person in a room, which honestly isn't very much like me. But you get what I say. And now this tool lets you talk back to me. And because it's just you on the microphone, I'm not capable all of interrupting you, which is a big bonus over having a real life conversation with me. Anyway. Once again, that is the I Heart radio app with the talk back feature, that little microphone icon that you can just tap. But let's get on with today's episode. And I really wanted to cover this topic and give a bit of an update because I feel it's one that is often misunderstood and frequently misrepresented. Um and that is the different levels of autonomous driving and the state of autonomous driving where we're at. I occasionally do these updates because I think it's very easy to get lost in marketing and pr speak and even some you know, political movements that suggests we might be further along than where we really are now. First of all, let's get this out of the way. Developing autonomous vehicle technology is a good idea. Here in the United States, the National Highway Traffic Safety Administration or n h t s A tracks the number of motor vehicle traffic fatalities year over year. The number of deaths due to vehicle accidents is always in the thirty thousand range or higher. So that means more than thirty thousand people here in the United States die every year due to traffic accidents, and if you just skim recent figures, that really reveals something counterintuitive. The year with the most traffic fatalities over the last decade is twenty twenty. Now I should add that I could not find an estimate for one, not for the full year, only for part of the year, so one very well might have taken the title away from twenty twenty. But yeah, the first year of the pandemic saw the most traffic fatalities in more than a decade. And you might imagine that fatalities for that year would actually be down due to more people staying at home, But you're wrong. The total that year was thirty eight thousand, six hundred eighty estimated deaths. The next highest year would be back in sixteen that had thirty seven thousand, eight hundred six deaths, almost a thousand fewer. The n h T s A also figures out that the fatality rate per one hundred million vehicle miles traveled is a good metric. Uh. This is v MT, which essentially tells you how frequent do these accidents occur, how many millions of miles are are happening in between these fatalities. Keeping in mind that's across the entire United States, so it's taking into account all the vehicles traveling across the country, and again had the highest frequency over the last decade with an astonishing one point three seven fatalities per one hundred million vehicle miles traveled. The next highest on that goes all the way back to two thousand eight with one point to six. So keep in mind like that means more than a decade of innovation and improvements in driver assist features and things of that nature ended up not making a difference because we saw more fatalities per miles traveled than ever well at least since two thousand eight, Because that's as far back as as the the sheet I was looking at wind So people were driving less in but they were dying in accidents more frequently. And it didn't get better in one either. During the first nine months of one, traffic fatalities rose by twelve percent. So within those first nine months there were an estimated thirty one thou seven twenty deaths. Again just nine months, not a full year. Now, I do not have more recent data to pull from, so I'm not certain how the year ended out, but that statistic is terrible and sobering. Here's another bad statistic. The c d C says that road traffic crashes are a leading cause of death here in the United States for people aged one to fifty four. Now, these deaths clearly have a massive impact. Beyond the obvious tragedy of losing someone or multiple someone's in a crash. There's the impact on that person's family and friends. That's an incredible emotional impact on them. If that someone who died in the crash were employed, well, there's an impact on their job and on the economy as well. There's this overall economic impact due to fatalities and accidents in general. The National Safety Council estimates that the average economic cost of a traffic fatality is one point seven five million dollars per incident. Now, this takes into account everything from medical expenses, to motor vehicle damage, to insurance costs, to wage and productivity losses, and more. So, if we take this average figure of one point seven five million dollars and we multiply it by the thirty eight thousand, six eight estimated deaths in twenty twenty, we get an economic impact of sixty seven point seven billion dollars. That's the cost of those fatal traffic accidents. When you take in all traffic accidents, by the way, the amount gets over two billion dollars. So there are a lot of reasons we would want to reduce or eliminate those traffic fatalities. I mean, first and foremost, obviously, we don't want people to die. We would rather not see that happen. We would rather those people go on to lead healthy, productive lives, surrounded by friends and family, and we would not want to see that end. Less importantly, we would not want to see that negative economic impact. And we also would rather not have the thousand other problems that come with traffic accidents, such as the effect on traffic congestion and that sort of thing. And there's one more important fact to consider. According to research, at least of all traffic accidents are caused primarily by human error. In other words, it's not a mechanical failure. It's not that a vehicle suddenly lost functionality in some way and that that led to the accident. More than are caused by someone making a mistake. And that's at least Some research pushes that even higher. The highest I have seen suggests that eight per cent of all accidents are call by human error. Um, so somewhere in that range is probably the truth. Right, That's still enormous, and that's where the sales pitch for autonomous vehicles comes in. Because what if you had a computer controlled vehicle capable of traveling at a safe speed and a safe distance from other vehicles. It would be able to pay equal amounts of attention to all areas around the vehicle. It wouldn't just be paying attention to wherever its eyes were, because it would have eyes all over. They could react much more quickly than any human could, and it could take action that could avoid an accident. You would remove human error as a contributing factor for accidents, and since we know human error accounts for more than of accidents, that would, by extension, eliminate the vast majority of accidents on the road. If you had enough of these vehicles on the road, perhaps not replacing all of them, but having a significant percentage of vehicles being autonomous, you would virtually eliminate traffic fatalities. That is the dream scenario. But there are a lot of things standing between us and that vision. I will go into that after we take this quick break. Okay. To get to that vision of the future where autonomous vehicles are taking us everywhere we need to go and traffic fatalities are a distant memory, we have to get the tech right first, and we also have to get enough vehicles out on the road to make a difference. One autonomous vehicle isn't going to change the world. And I've seen a lot of different research about this as well, some researchers saying that we might need to reach a saturation point of about one third autonomous vehicles to two thirds human driven vehicles to start seeing massive changes and things like accidents and traffic congestion. And then, of course, beyond that, we'd see even more improvements, not necessarily at the same dramatic pace, like it might level off more where we get sort of a diminishing return situation. But when you're talking about saving lives, diminishing return doesn't really have the same meaning, right, Every life saved is significant. Now, when it comes to autonomy, at least for vehicles, there is a spectrum, you know. There. Autonomy is more like guidelines, as the pirates would say. The Society of Automotive Engineers, which is a quote globally active professional association and standards developing organization end quote, has identified six levels of autonomy, ranging from zero to five, and it is ascending in autonomy. So let's go through those now. At levels zero, you have no driving automation really of any significance. I used to think that level zero just meant that you had complete manual control of the vehicle and there were no driver assist features at all. That's not the case. That was my mistake. That was my misunderstanding of this designation. Uh, you could have cruise control and it could still fall into a level zero category. So basic cruise control would not be enough to move a vehicle out of level zero designation to level one because basic cruise control will maintain a set speed, but it won't change it. Right, if there is a vehicle that's ahead of you, the standard cruise control isn't enough to detect that vehicle and change the speed. That would have to be adaptive cruise control. That's different basic cruise control. You're still at level zero. Back in the day, I used to think that that would at least push you to level one. No, not, according to the s A The s A E says that any driver assist features are in a level zero car are limited to quote providing warnings and momentary assistance end quote. So you can have like a lane correct feature, but if it's just momentarily shifting the car, that still is considered a level zero autonomous vehicle. Automatic emergency braking, lane departure warnings, all that kind of stuff I'll just count as minimal driver assist. So any vehicle having those features but nothing more extensive than that, would still be level zero. So that's a lot of the vehicles that are on the road today, the vast majority of them, in fact. But let's go on to level one, because we do have level one and even level two autonomous vehicles on the road. So a level one autonomous vehicle has the ability to take over some duty that the driver would normally do, but only one of those duties, and the human driver would have to handle everything else. So, in other words, the feature might be able to handle something like acceleration and braking, or it might be able to handle steering, but it would not be able to handle both of those, So whichever one it handles, the human driver would have to deal with the other. One adaptive cruise control that I mentioned earlier, in which a vehicle handles acceleration and it will ease off if it detects that it's approaching the rear of a vehicle in front of it. That's an example of a level one autonomous vehicle feature. It's handling acceleration and braking, but it's not handling steering. The human driver still still has to steer the vehicle in that scenario. Now, if we move up to level to autonomy, we're now at a phase where the vehicle can handle both the acceleration and breaking activities as well as steering the vehicle at the same time. However, a human driver must supervise these and also intervene when necessary, as the vehicle will not always be capable of avoiding an accident when something out of the ordinary happens, or even something remotely unexpected happens. So this would be a vehicle that can do both adaptive cruise control and lane centering operations at the same time. And honestly, this is where we're at with most of the really sophisticated passenger vehicles that are on the market right now. Tesla, even with its so called full self driving feature, which is not full self driving, and Tesla darnwell knows it. I have a longstanding issue with Tesla and it's naming conventions. Calling its first driver assist support sweet as autopilot when it's not really autopilot really upset me. Full self driving is even worse because it the name implies that the car just completely drives itself. That is not the case. It still requires human supervision, and Tesla says this to drivers. It says that you still have to supervised the vehicle, that you can't just take your hands off and go to sleep behind the wheel. You have to maintain supervision and be ready to intervene. But yet they insist on calling their features full self driving. I just feel like that's misleading. Anyway, that is still a level to autonomous feature. It means Tesla is on the level two on the autonomous scale. It hasn't hit level three yet. And from levels zero to two, the job of monitoring the driving environment falls on the human in the car. They are responsible for making sure that the environment is safe. Uh that for at least whatever operation the car is in at that time, and you cannot offload that to the vehicle. The vehicle is not capable of shouldering that accountability. So ultimately the person behind the wheel is still responsible for their own safety and for avoiding accidents. But now let's talk about levels three through five, in which the vehicle's automated system is in charge of monitoring the driving environment. So in this case, the car is responsible or the vehicle I should say not just car is responsible for monitoring what's going on around it and then using that information to make important decisions. So level three is called conditional automation, and that kind of clues you into what's going on. So the vehicle is capable of operating autonomously under certain conditions, but not all conditions. So unless all autonomous conditions are met, the vehicle will not operate an autonomous mode. A level three autonomous vehicle will be able to make certain environmental decisions, such as accelerating past a slower moving vehicle or maneuvering through a traffic jam, which typically involves lots of starts and stops and potentially lane changes. There are a couple of vehicles that have received international acknowledgement as attaining level three autonomy. There's a car called the Honda Legend, which is only available as a least vehicle in Japan, so unless you're in Japan, you're not likely to see one or experience it. Mercedes Benz has received a regulatory approval to manufacture level three autonomous vehicles that have its drive pilot system installed. It's recently been updated and has met international standards to be considered a level three autonomous system. So with a level three vehicle, human intervention isn't needed for most situations. It may still be required in edge cases where things are unusual. Now, at level four autonomy, an autonomous vehicle would in theory be able to react to most circumstances without the need for human intervention. It would not ask for human intervention, it would just react. The vehicle would respond properly, able to discern what was happening and then take the appropriate response. For example, such a car would presumably be able to distinguish between a trail of gosling's attempting to cross the road versus some leaves being blown by the wind, and either it would slow down to let the little goslings get across, or it would just continue without worry because it's just some leaves. And yeah, just saying that filled me with anxiety because I think goslings are cute and the thought of a car just driving through a crossing animal really really upsets me on a deep level. However, a level for autonomous vehicle would still be a conditionally autonomous vehicle, so there would still be some conditions under which this vehicle would not operate autonomously. And because this is the level that the s a E considers to be appropriate for robo taxis, it would mean that under a certain con asitions you wouldn't be able to get a ride in a robotaxi. Because if if conditions are wrong for autonomous operation, then they won't they won't pick you up, right, They won't work. Those conditions could include things like weather events like let's say it's really pouring down rain and because of the heavy precipitation, there's a fear that the vehicle's sensors won't operate properly so it's unavailable. Or it could involve things like geo fencing. That is, the vehicles can only operate within a certain geographic area. Typically we're talking about a specific town or specific city, or specific lane between two cities, and the cars systems will not allow the vehicle to venture beyond certain borders. In fact, we're gonna talk about how geo fencing is one way that companies are looking to tackle very tough engineering challenges in an effort to bring robotaxis to various places. Also, at this level, things like a steering wheel and or pedals like an acceleration and brake pedal. Those might not even be installed in a level for autonomous vehicle, so there would be no controls for you to take over should you feel the need to intervene. You would not have that option. You might have a button to press that was like an emergency stop feature where the car would then pull off to the side and come to a halt, but that might be it. So a level for autonomous vehicle should be able to handle pretty much any situation that could pop up under its operational conditions. And finally, we have level five autonomy. Now at this level of vehicle under any conditions would be able to operate autonomously. So this is the level where a vehicle could go anywhere and operate safely no matter what conditions might be present. Such a vehicle might have no other controls in it and handle absolutely everything by itself. We are nowhere close to achieving level five autonomy right now. So when we come back, we'll talk about some of the things that are holding us back from full autonomy, and some of them are technical and some of them are social and political. But first let's take a quick break. Okay, let's talk limitations, and we'll start with the technical. We've seen some pretty dramatic improvements in sensor technology over the years, but the fact is that under some conditions sensors can have trouble detecting the environment effectively, particularly when you're talking about the sensors being mounted to a platform that's moving at several dozen miles per hour or kilometers per hour. So this can include conditions like dense fog or heavy precipitation. Like I mentioned earlier, it would be pretty dangerous to ride in an autonomous vehicle that ends up mistaking fog for a solid surface. For example, let's say that whatever technology the sensors are using, the signals are bouncing back because it's hitting the fog or the precipitation. A car like that might start breaking suddenly, like applying the brakes suddenly, and that could pose a hazard for vehicles that are behind you. And since their limitations aren't a hypothetical either, we've seen some tragic cases in which driver assist systems failed to detect a danger and that led to loss of life fatalities. I mean back in famously, a Tesler driver named Joshua Brown was on a Florida divided highway. So that's a highway that has a couple of lanes going in one direction. Then you typically have a divider, a median of some sort, and lanes going in another direction. Uh and back in ten, Joshua Brown's on a Florida divided highway. He had autopilot in his test engaged and as he was traveling down there was a semi truck that was pulling out of a driveway that was perpendicular to the highway. The semi truck was crossing Brown's lanes of traffic in an effort to make a left hand turn to go the opposite direction down the highway, so it was pulling across the lanes that Brown's vehicle was in, uh the Tesla. Brown's Tesla failed to detect that there was an obstacle in the way, so the Tesla did not slow down and ended up colliding with the truck and Brown died in that car accident. Now, back then, Tesla was relying on systems that were provided by a company called Mobile I, but Tesla subsequently ended that partnership and then began to develop new sensor technology in house. So the autopilot of today's Tesla's works on a totally different techno logical platform then the one back in two thousand sixteen. Because Tesla recognize that this is not something that it can it can support and still expect to be able to sell vehicles that have this autopilot feature in them. Now that being said, we saw a near identical tragedy occur in two thousand nineteen when Jeremy Banner, also going down a divided highway, also in Florida, also driving a Tesla that was in autopilot mode, also collided with a semi truck that had pulled out to make a left hand turn. It was, in many ways an identical scenario to what we saw before. So that leads to a question, how could two different autopilot systems make the same fatal error? And a lot of this potentially has to do with the types of technologies we depend upon to do certain things. So a lot of the anti collision technology we find in vehicles real lies on Doppler radar. Let's talk about what that means for just a second. First radar, Now, that involves sending out a radio signal and then detecting the echoes of that signal as they bounce back to the source. So your radar has an emitter and it has a sensor. The admitter sends out a radio signal, the sensor detects the returning radio signal after it's bounced off something. And when you measure the amount of time it took for a signal to go out and then bounce back to you, that tells you how far away you are from whatever it is that you're beaming signals at, which is pretty simple, right, But let's talk about the Doppler part. Now, if you've got two objects that are standing still, well, that means the signal you send out and the signal you get back are going to be pretty much the same wavelength and frequency because both you and the object are at a standstill. But what if one of you is moving toward the other, Well, in that case, the signal that comes back to you is going to be compressed. It's going to be shorter in wavelength and higher in frequency than the signal you sent out. It's being pushed by whatever it is that's moving towards you, or, in the case of you moving toward it, the fact that you're getting closer. So this tells you not only are you within a certain distance of another object, it tells you that you are getting closer to that object or that object is getting closer to you, depending on your point of reference. Now, for those of us who are are able to hear, this is something that we can experience out in the real world without any radar at all. If you've ever heard and approaching emergency vehicle that's running its siren. You probably noticed the sound of the siren changes when the vehicle passes you, So as the vehicle is approaching you, the pitch of the sirens sound is higher, and as it moves away from you, the pitch goes lower. And that's because the actual sound wave is being compressed as the vehicles moving towards you, and it's expanding as it's moving away from you. The vehicle pushes those sound waves closer together as it's coming at you and stretches amount as it's going away. That's when why the pitch is higher one way and lower the other way. The same thing happens with stuff like radar and radio signals, and obviously it gets more complicated if both you and the other objects are in motion at the same time. But these systems work best when everyone is moving in the same direction of travel, right, So it works really well if you're behind another vehicle and that vehicle slows down, then the system is very good at detecting that. But in a case where a truck is pulling across lanes of traffic, where the truck is moving perpendicular to your direction of travel, it doesn't work as well. It doesn't work as well on stationary objects, and a truck would appear to be stationary as it was moving laterally across your lane of traffic. Now, depending upon the lighting and the time of day and autonomous vehicles, camera system might not do much good either. It might misidentify a truck as being like a road sign, and a road sign hangs over the highway, it's not in the way of the highway, so the system wouldn't think of it as being a potential threat. Or it might interpret the side of the truck to be the sky, depending on those lighting conditions, and that would mean that the vehicle safety features would not initiate, and that could lead to tragedies like we've mentioned here on this episode. Now, I don't mean to suggest that these technical problems are impossible to solve. I don't think that's true, but I do think it's going to take a lot of work, and it might require technologies that for the moment are prohibitively expensive, meaning that yeah, we could create a more reliable autonomous system, but then the cost of the vehicle would be so high that it wouldn't be practical from an economic standpoint. Some of these systems. These components like LDAR, some of them can cost as much as a brand new car does all by itself. And I'm not talking about the cheap car either. So when you start looking at that, you have to then take the economic factors into consideration and say, doesn't make economic sense to develop a system that relies on these components, because will we ever make our money back? Like would you be able to sell the vehicle at a cost that would make sense, or would no one buy it because it would be far too expensive. Even if you're talking about, say a taxi service, maybe a taxi service wouldn't buy it either because they'd say, the likelihood of us making our money back with cars that are that expensive before we have to replace those cars is so low that it doesn't make sense to go into it. That could be a real obstacle, and as you know, more to do with the cost than the actual technology. I also think that a major problem with autonomous vehicles has been in messaging and in hype companies in the autonomous vehicle business, not just Tesla I. I heap a lot of abuse on Tesla for this, but it is by far not the only company to engage in these kinds of things. But I feel like a lot of the leaders in that space and made some really aggressive promises that in hindsight were impossible to follow through on. At least they were impossible given the time frames that were being tossed around. There were folks suggesting that we would have cities filled with autonomous vehicles able to operate in all conditions by well, it's two and obviously that didn't happen. And no, it wasn't just because a pandemic really messed things up. Though that definitely didn't help because it disrupted everything from workflow to supply chains. But even with sophisticated sensors, you then have to consider the decision making factor and take add into account. Now, at the moment, autonomous vehicles are for the most part independent computing islands. It's like a personal computer that isn't connected to the internet. So all the technology used to keep the car in autonomous operation typically is just on the car itself. So the car is relying on its own sensors and its own processor to detect the environment and make decisions. That is a tremendous amount of responsibility to put on technology. Then again, it's also a tremendous amount of responsibility to put on a teenager too, but that's a different discussion. There are some proposals that suggest the real future for autonomous vehicles is one in which vehicles are in constant communication with one another and potentially with the infrastructure itself, with the road system that a vehicle would be able to stay in communication with, like a smart city, The road system and the cars on it would both be capable of adapting to different situations. So, as an example, a smart city might have intersections that could proactively change the timing on traffic lights to accommodate changes in traffic patterns and smooth out the experience and avoid long traffic jams. Cars in communication with the smart city infrastructure could plot out the most efficient routes that will save time and energy and adapt to changes that happened on the fly. It might even be able to adapt to routes that cause less wear and tear on the actual infrastructure, giving city planners the chance to address things before they become really big problems. I say that as someone who lives in Atlanta and is very used to the site of giant metal plates laid across the road because potholes have developed and the city has not had time to address it. If you had really smart infrastructure, you could start to detect things before they got to that point of of a problem. So these are really cool ideas, and the improvements in quality of life in those kinds of cities would be enormous. But to make this happen a lot of other stuff has to fall into place. First. Vehicles and cities would need to agree upon a common set of languages and protocols in order to communicate effectively. That alone is going to require tons of work. Typically, we see car companies develop their own in house systems which may or may not be compatible with the systems that are used by other car companies, which could lead to real delays. And seeing a fleet of truly autonomous vehicles navigate the roads, it would be kind of like putting a bunch of people who don't speak the same language into a maze and have them try and navigate it together. The cars would only understand other vehicles that were from the same manufacturer, and that would have limited usefulness unless we all just migrated to the same vehicle manufacturer. But that would turn that company into a global monopoly and uh I humbly suggest we don't do that. But that future in which vehicles talk to each other as well as two road systems into cities will require tons of other work as well. Obviously, cities are going to have to build out smart infrastructure, and that's likely to be expensive and time consuming. We also need really good wireless technologies in place to five G has promise, but we would need it to be more dense and more widespread than it is currently. For some industries like trucking, five G is not a great solution because on long hauls between cities, you're gonna hit routes that are going to have little to know five G support along those routes, and it would be foolish to build out a business that depends upon five G connectivity if that business involves navigating through places that don't have five G support. Meanwhile, there has to be a strong enough business case for the telecommunications industry to build out five G availability to these regions. That's hard to imagine if there just aren't that many people living along those routes, right, if it's not densely populated, then the telecommunications companies are gonna say, well, it's really expensive to build out the infrastructure. We have to put antennas all over the place. If we want to have high frequency five G, the stuff that's really high through put, low latency, that stuff, the five G that we think of when we think, oh, this can replace fiber Internet. You have to have very dense antenna deployment for that to work. Well, that's expensive and if a place is not densely populated, it's a really tough thing to tell a telecommunications company you need to put out the expense to build out that infrastructure even if you only have a few thousand customers in these area, is right, Like, that's that's a tough sales pitch to make to these telecommunications companies. So it's not likely that we're going to see intense five G coverage along some of these places. So five G could work in cities, and it might be a really good solution for the future of smart cities, but that would likely mean we would see autonomous vehicles largely confined to a specific region that geo fencing approach. They wouldn't necessarily be able to travel outside of those regions, not without having a robust enough system on board. The vehicle itself to handle everything when it, you know, kind of loses connectivity. And that's a really big part of it too. The fact is, all sorts of weird stuff happens on the roads, and engineers are not able to anticipate all of them. None of us are. There's just no practical or even possible way to preprogram the right response to every possible thing that could happen on the roads. This is another reason why geo fencing is an important solution, or at least a temporary one, because it limits the operational range of an autonomous vehicle. That also means it limits the number and variety of weird stuff that that vehicle could potentially encounter. It's not gonna eliminate weird stuff. It'll just limited to mostly a subsection of all the weird stuff. Right, Eliminating variables makes the whole problem less complicated. I was gonna say easier, but that's the wrong implication. Less complicated is probably the better wording, because it's still wicked complicated. Then we have to get through all these social components to making driverless vehicles an acceptable reality. According to Alex Kopatinski of Policy Advice around fort of people in the United States do not feel safe in a driverless car. They do not like the idea of getting into a driverless vehicle. I actually think for being uncomfortable is pretty low. I would have expected that more than half of people in the United States worry about their safety if they were to get into a driverless vehicle. Now that's not based on any hard data. That's just my my feeling, which we all know is not really that's not evidence, it's not worth while. I just feel forty feels uh really generous. And that's because whenever a driverless vehicle gets in an accident, we see a disproportionate amount of coverage about that accident. Now, we all know fatal accidents due to human error happen far too frequently. They're happening every day, lots of them, with thirty eight thousand, uh nearly thirty nine thousand happening in lots are happening every single day, and they typically do not receive nearly as much coverage as an autonomous vehicle accident beyond maybe local reports. Your local news might cover it, but and sometimes it might just be something like a traffic report, but you don't get the coverage that you would get if it were an autonomous vehicle. When an autonomous vehicles at fault, that becomes national or maybe even global news. So there is a disparity going on here. Right. You could argue that the autonomous vehicles in in aggregate are far more safe than human driven vehicles, but the news reporting has an amazing disparity there. There's going to be way more reporting on autonomous vehicle accidents than on your typical human caused accident. Still, even at you're still looking at a ton of people who would feel unsafe getting into a driverless car, and whether that worry is justified or not, it poses as a challenge when it comes to policy. See another really big component in the adoption of drive list car technology is getting politic Aisians to support it and to develop laws that create the guidelines around it. Many places lack laws that would cover autonomous driving vehicles. That gap is something to be concerned about. It means that if something happens, like if an autonomous vehicle gets into an accident in one of those places, you would have no framework to establish things like accountability. That's just one tiny slice of how it's important to establish policy so that there is a legal infrastructure around autonomous vehicle operation. That legal structure would in turn inform manufacturers as to certain requirements their vehicles would have to meet, and that would mean that these policies would also shape technology. Wouldn't just be how can we make this technology do the thing we wanted to do, It would be how can we make this technology do the thing we wanted to do within the legal framework. So that's an important thing that has to happen too, And in some places we're seeing that. In other places there's not really any official policy. If the policies don't match each other, that also creates a complication, because what do you do in a case where an autonomous vehicle that's legal to operate in one place ventures into a place where it's not legal to operate because it doesn't fully meet all the requirements. These are tough questions now. Way back in the day, if you listen to tech stuff, you know, I was really super hyped up for autonomous vehicles. I was so excited about it, and that's because I recognized the fallibility of humans. I recognize that we make mistakes, and the fact that technology can detect and react at a speed that is impossible for us to even imagine, meant the potential for this technology to save thousands of lives every year. And based off that very limited perspective, yeah, autonomous cars are phenomenal, but it also ignores all the other things that you have to take into consideration that driving, say fleet, is about way more than just seeing something and then reacting to it in time. And that's really where I was being blind. I bought into the hype and was eagerly looking forward to one when I'd be able to hop into a driverless car and go wherever I wanted. These days, i'm I would say, I'm more cautiously optimistic. I do still think we're going to get to a future where autonomous vehicles will play a more central role in how we get around locally, and in fact, it might it might even cut way back on the number of vehicles we have on the road in the long run, But I think that's going to take ages to play out. Maybe we'll see those visions of a future where parking lots are reclaimed and turned into stuff like parks and whatnot. Maybe we'll see that come true, Because maybe we will see a future where we have a fleet of autonomous vehicles that just kind of are there on demand whenever we need them, and then go somewhere when we don't. That somewhere is a big question mark by the way, that we haven't answered, and that we don't own our own car, right, we just use driverless vehicles as an on demand service whenever we need to get somewhere. It will be priced at a point where it would make sense to do that, where you know it would be at least uh no more expensive than owning and maintaining your own vehicle, and preferably less expensive than that. For that to happen, a lot of stuff has to fall into place, and it goes well beyond the scope of this episode, so I'm not going to go into it. But let's say that all these things are actually able to come true. If that's in fact a possibility, it's gonna take decades, if not longer, for us to get there. We we've just got we've got so much inertia built up that we have to get through that in order to make that future reality. And we have to remind ourselves that to achieve that vision, it's going to require a ton of work and the solution to a lot of non trivial problems. That is not a reason to give up, but it is a good reason to check our expectations and to look at things from a more critical point of view, and to approach things that way. That way, when we're really being critical, when we're really acknowledging the challenges that face us, were more equipped to meet those challenges, to find solutions, and to work through it. If we fall into the trap of being overhyped, like I was back in the day, then we're not gonna realize those challenges until they are undeniable, and when we've wasted time, money, and resources on things that were trivial in the in the long term. So still, I'm still up on autonomous vehicles. I still I'm eager to see them become a thing, not entirely sure that they're going to be an effective thing within my lifetime. I mean, they will exist, they will be out there. I just don't know at what sort of density it will be. But I do think it's a worthwhile pursuit. I think the benefits are undeniable if we are able to solve the problems. All Right, That wraps up this catch up on autonomous vehicles. Hope you enjoyed it as a reminder, if you listen to this show on the I heart Radio app, really, if you just download the i heart Radio app, you can use that talkback feature to leave comments either on this episode or on the show in general, and it can be up to thirty seconds and I'll be able to hear it and maybe even use it in a future episode if you would like that. Also, if you would prefer some other method, the other tried and true way to get in touch with the show is through Twitter. The handle for the show is text stuff h s W and I'll talk to you again really soon. Text Stuff is an I heart Radio production. For more podcasts from I Heart Radio, visit the i heart Radio app, Apple Podcasts, or wherever you listen to your favorite shows.