Hey guys, this is Jonathan. I have a quick announcement before we get to today's episode, which is sort of a capper on the autonomous car episodes we listened to last week, and that is starting this week, Tech Stuff is going back to publishing two new episodes and one classic episode each week, so we're cutting back a little bit. This is so that I can spend more time putting these episodes together, making sure they are the absolute best I can make them, as well as making time to work on some other shows like The Brink. If you're not listening to The Brink, you should go check that out and some more shows that might be coming down in the near future once we get through pilot season. Here at How Stuff Works, and I just want to let you guys all be aware of that nothing bad is happening. These are good things. It means that I get to spend more time on this stuff and I'm not quite as a as rushed as I have been for the last half year. So enjoy and I'll see you guys again on Wednesday. Get in touch with technology with tech Stuff from how stuff works dot com. Hey there, and welcome to tech Stuff. I'm your host Jonathan Strickland. I'm an executive producer with How Stuff Works, and I heart radio and I love all things tech, and we are continuing our long journey with autonomous cars. Actually we are concluding it at this point, even though as I get to the end of the episode you will hear about how I could cover even more. But out of consideration for you, my dear listeners, I'll move on to other topics for a while. But last week I did a whole series of episodes about the history of autonomous cars, and we started with the science fiction visions of the nineteen twenties and radio controlled vehicles up to the upcoming launch of the Weymo one service in Phoenix, Arizona, which maybe in operation by the time you hear this. And I mentioned how over the last several years, in particular, a few accidents, a couple of them fatal accidents, have brought a lot more critical attention to driverless cars, even as numerous services prepared to unleash thousands of them on the streets of several cities across the world. In this episode, I'm going to talk about the arguments for and against autonomous cars, or at least against early autonomous car deployment. I'm gonna talk about how these things are complicating factors. It's not just a technological problem. In fact, I don't really get into it in this episode, but there are entire issues with regulation and legislation. Like many other technologies, autonomous cars has outpaced the law, and so there are a lot of places that are trying to take into account what driverless cars in the streets would actually mean, and how do you legislate that. Who do you find at fault in the event of an accident? These are still largely open questions. There are other ones that I haven't touched on in this episode either, For example, the trolley problem, the idea that if you have to build into your car system a decision making process to follow in the event that an injury is unavoidable, you've got and and a scenario where no matter what, someone is going to get hurt. How does the car make the choice of who gets hurt. That's a huge question in ethics, in artificial intelligence in general, and in autonomous cars in particular. I don't really go into that in this episode, so there are still many topics in this field that I will have to come back to at some future point. However, that being said, there's still a ton of stuff to talk about outside of that, and I'm going to start off by exam and some of the arguments for autonomous cars, as well as the criticisms leveled at those arguments now before we jump into this, because it's going to sound like I'm totally down on the concept, like I'm like, I'm really against the idea of driverless cars. That's what it might sound like to you as we go through this episode. But that is not the case. And I'm gonna be real with you, guys. I don't drive. I I do not have a driver's license. And I don't talk about this very much because it's hard to talk about because driving in the United States is such a a common experience among many many people. I mean, unless you're living in a city that has incredible public transportation and driving is incredibly expensive, then you probably drive a car, Especially in the Southeast where I live. Owning a car and driving a car is very much tied with the thought of independence. But I don't have a driver's license. I actually have a phobia about driving. I end up locking up when I get behind the wheel of a car and I become a danger on the road, and I'm aware of that, and then that also feeds into my fear and it gets worse and worse. So you see where I'm going. I I have realized that I am a danger to myself and others if I get behind the wheel of a car. I am not a good driver. So this is an incredibly frustrating condition for me, and it's difficult for me to talk about because there's a big stigma against this. But it means that I frequently have to rely upon other people to help me get around, and that means I often feel like a burden, even when people go out of their way to dismiss that idea, saying no, no, no, no, I want to give you a ride. I want you to be involved in this thing. I can't help but feel otherwise. So I very much want a future in which we have a reliable, safe, autonomous car service where I can call upon a car anytime need and it will show up. And I'm not putting anyone out right, I'm not making it inconvenient for somebody else. It's a service. I'm paying for it. Everything's cool. I'm eager to see that kind of thing across the world. That being said, I also recognize that I can't allow my desire for that technology to come to maturity to guide all my thinking on the matter to bias my point of view. The responsible thing to do is to hold up the facts to critical thought and consideration. After all, we're ultimately talking about enormous vehicles traveling at potentially very high speeds, and that's a lot of force. So it's only right that we look at all the facts before we leap to any conclusions. And knowing how badly I want autonomous cars to become a thing has to be a warning sign to me. I have to say, well, I need to be extra careful when I'm looking at the facts so that I don't just become latch on to the most optimistic outlooks and then just, you know, deny that there are any problems. So the first big argument for autonomous cars in favor of them tends to be that, assuming you have cars that perform safely and consistently, you would cut way back on the number of fatalities and injuries caused by car accidents. And the general line of reasoning is that machines can react in a fraction of the time that people can, which is absolutely true, and a well designed vehicle, one that has the proper sensors, could maintain a three D sixty degree awareness at all times, whereas a human can only focus on a small part of their overall field of view, which is locked to whichever way they happen to turn their face at any given time. That's also true assuming you've designed the sensors and the processing system properly so that it can accept that data and analyze it in real time and be able to react to it in real time. Now, this argument that autonomous cars will save lives also assumes that the software the algorithms guiding the car's behaviors have been properly designed and tested and they're reliable. Whether that means programming the car for specific scenarios or using a process like machine learning in which an autonomous system is quote unquote taught how to handle various driving scenarios. Now, that assumption is particularly important, and it also is one that we should be particularly critical of. Now, I don't mean to suggest that it is impossible to design a sufficiently adept system or a sufficiently a depth set of algorithms but rather we should be as certain as we can and be as we can reasonably expect us to be that such systems meet are required levels of reliability before we start getting into driverless vehicles for any sort of you know, given trip. Now, all that being said, statistics do show that humans are far from perfect drivers. The U s Department of Transportation produces reports from the Fatality Analysis Reporting System or FARS, and yes, that's pretty grim, but it's very important work. The two thousand sixteen numbers, which were some of the more recent numbers that I could get hold of, say that in the United States in two thousand and sixteen, there were thirty four thousand, four hundred thirty nine recorded traffic accidents that had one or more fatalities. So if you look at all the actual deaths, you're looking at thirty seven thousand, four hundred sixty one people who lost their lives from car accidents. So when you take into a account the population of the United States, that averages out to eleven point six deaths due to car accidents per one hundred thousand people. When you break it out state by state and you look at their statistics, you see where the concentration is highest. Mississippi had the highest fatality rate by this metric, at twenty three point one fatalities per one hundred thousand people. The report also takes into account the number of vehicle miles traveled. It's very important as well, because that tells you how frequently people were on the roads and how far they went, and how much time they were spending on the roads. So in sixteen, the estimated total miles traveled in vehicles in the United States was three trillion, two hundred twenty billion, six hundred seventy seven million miles. It's a really important number that we're going to come back to in just a second. But another important number is that in the United States, experts estimate that human error causes somewhere between and of all car accidents. Uh, there's discrepancy because it all depends upon which study you're looking at, but the the message there is the overwhelming majority of car accidents in the United States are caused by human error, as opposed to say, mechanical failure or other extenuating circumstances like a tree falling on a car. So it's easy to see how the thought of removing the human element from driving could dramatically reduce the number of crashes. But again, this assumes that the driver list systems are in fact more reliable more safe than human drivers. Now they likely are, but it's hard for us to say we've definitively proven this. In fact, that would be misleading to say we have definitively proven these cars are more reliable and more safe under automated control than human control, And the reason for that is because of the amount of data. This is where we start to really get scientific with our thinking. So the safety argument says, imagine a future in which all cars are autonomous and can detect one another. They can react quickly to changes in the environment much much faster than humans can. So then imagine how drastic the reduction of fatalities would be. Those nearly forty thou people who died in would still be alive in a world where those cars were the way we got around. But this presupposes that the autonomous cars would perform as we would like them too, and that still remains to be seen. Google's way MOW, which is the company that I think is the the foremost authority on driverless cars as far as public awareness is concerned. At least, they like to point out that their self driving vehicles have passed the ten million mile mark. In other words, autonomous vehicles have driven more than ten million miles and there have been no accidents in Weymo vehicles that resulted in a fatality from their operation in autonomous mode. And ten million is a big number. Ten million miles. That's a lot of distance. But remember that in humans in the United States drove three trillion miles in one year. You know, Weymo is looking at the entire history of their autonomous car program in one year. Because of the number of drivers and how much we drive, humans racked up more than three trillion miles. So we have more human drivers on the streets, which also makes it difficult to draw any sort of comparison between human drivers and autonomous drivers. There are more human drivers out there and they're driving more miles than autonomous cars are, so it's not an apples to apples kind of comparison. We cannot say definitively that autonomous cars are safer. They just haven't been driving enough and there aren't enough of them out there for us to make that kind of general conclusion. Does that mean it's wrong? Not necessarily. It just means we lack the information to be able to say it definitively, and without that information, there's doubt, and when there's doubt, you have problems. In the last couple of years, there have been some pretty high profile incidents involving autonomous cars and accidents, and these accidents don't necessarily mean that the safety argument for autonomous cars is invalid. We haven't seen a rash of driver list car accidents. We've only seen a couple like a few dozen, and many of those aren't even the fault of the autonomous system, but rather the fault of human drivers who happened to be on the road at the same time. However, even if we were to sift through all of the data and prove definitively, statistically speaking, that driverless cars are safer than one driven by an average person, it might not matter at the end of the day. Now, this is because we humans are not entirely rational, and I am guessing this does not come as a shocking revelation to you guys. Psychologically, it can be easier for us to accept that driving a car on roads dominated by other human drivers is risky. We understand this because we know that driving, while often a mundane task that we can kind of do without even devoting our full attention to it, can sometimes be punctuated by moments of unique circumstances, and that in those circumstances, the wrong reaction could cause an accident. This is something we can wrap our minds around. We can accept this. We know accidents happen, sometimes we are in them, sometimes we cause them, and that is something that we can accept. But there's a bit of a leap for an awful lot of people when you shift that two vehicles that are controlled not by human drivers but by machines and algorithms. Suddenly the same people who accept that driving can be inherently dangerous in certain situations find it much more difficult to accept when humans are no longer directly responsible. The journal Risk Analysis published the results of a survey in which people were asked to determine how safe an autonomous vehicle would have to be before they would feel comfortable with the idea of them being on the road. The results were really interesting. They showed that, on average, people felt that autonomous cars would have to be proven to be four to five times safer than human operated vehicles for them to feel comfortable four to five times safer. So let that sink in because by this argument, even if the driverless car was twice as safe as the average human driver, which by definition would mean that if you had enough of these replacing human drivers out on the road, would mean you would have fewer injuries, fewer cases of vehicle or property damage, fewer fatalities. But that wouldn't be enough for these people to feel comfortable even if they knew that, as a result, fewer people were being hurt or killed, there was less damage. And when you turn this argument on its side, it's kind of like saying it's okay if a certain number of people are hurt or killed because we didn't shift to autonomous cars. It's only when autonomous cars are far more safe and capable four to five times than human drivers that people start to feel better about them. And I think you could argue that it's a pretty extreme double standard. We're not requiring humans to undergo driving tests more frequently to prove that they are just as safe or not if not safer, than they were the last time they took a test. We're not requiring people to do that. But it is the sort of stuff we're demanding from the machines, and it's also something I can kind of understand because it means taking the humans out of the equation, or seeming to, since these cars and algorithms are ultimately designed by humans and that's scary. The idea of entrusting lives to machines so is a bit of a paradox. So to summarize the safety part of the argument, we don't yet have enough information to really know how safe autonomous cars are compared to human drivers, although I suspect they are as a whole safer, but again hard to tell because you're extrapolating a lot from a very small data sample size. We don't know because we lack that information. There just aren't enough driverless cars in the road that haven't driven nearly enough miles and enough different environments in order for us to draw any firm conclusions. We also have to remember that different companies are taking very different approaches to driverless cars, so we can't make any sweeping statements because one approach might be better in some situations and worse than others, with no single approach being the best for all scenarios or regions. We do know human drivers are fallible and that we are at fault in the overwhelming majority of car accidents, and we suspect that eventually will have autonomous cars that can operate more safely than people. We just don't know if we're there already or if we still have a ways to go. Now. I'll have more to say about the arguments for and the arguments against autonomous cars in just a second, but first let's take a quick break to thank our sponsor. So, the safety argument also presents a couple of other possibilities. Besides an improvement in safety, assuming that everything is working properly. Uh, it also presents the possibility that autonomous cars might require less material and less energy to operate. So here here's the thought. If your vehicles are proven to be safe and you don't have to worry about them getting into accidents, let's say that you have reached a point where you're ready to really deploy these, then you don't have to build your vehicles quite as sturdily. Right. One of the reasons why cars and trucks and other you know, consumer vehicles way as much as they do is because the materials they're made out of are in part there to protect the people inside in case of accident. But if you've removed the possibility of accident, then you don't have to build them quite as Hardy will say, so it means you could go to lighter materials to make your vehicles. Well, lighter vehicles need less energy to move around than a heavier vehicle does. You don't have to put as much force to get the vehicle moving or to stop it. So if you reduce the weight of vehicles on the road, you also reduce the amount of energy you need to move them around. That also means less fuel consumption, which could be in the vehicle itself if it's a gasoline powered vehicle or hydrogen powered vehicle, or in a power plant that produces the electricity that is going to charge those electric vehicles batteries. And that sounds pretty good, except I would argue we can't really afford to get fully behind switching over to light vehicles until we reach a real tipping point of the percentage of driverless cars versus human operated ones, because even if we assume robot cars are perfect, which again this is just like hypothetical, just assume that they work exactly the way they're supposed to. You still have humans on the road and they could still cause accidents. And for that reason, we can't really afford to essue safety features that add weight to vehicles just yet. We need to still have those considerations in place. Another big argument for autonomous cars is that they could have a positive influence on traffic, meaning a reduction in traffic and traffic jams and congestion. But the studies on this subject of conflicting answers. The National Science Foundation funded a study to look into this matter, and the study consulted with auto industry experts and government officials. And in that study, the researchers created a simulation in which one autonomous vehicle, a computer controlled simulated vehicle, circled a track with at least twenty vehicles driven by humans. It was a simulation that used VR, so these were VR cars driven by human beings and not just you know, twenty simulated human beings, because then you're talking about two different types of autonomous cars, and that wouldn't tell you anything anyway. This experiment saw that humans create a stop and go oscillation effect in traffic, but that autonomous cars don't cause this kind of oscillation effect, and this can happen in things like when cars change lanes or they need to exit the highway or into the highway, and this can lead to traffic jams. So if you've ever been on highway and you hit a super slow section of traffic and then the traffic starts to clear up, and you never see any reason for the traffic jam, right, there's no accident, there's no reason, nothing to rubber neck at. You might wonder what the heck happened, Well, that's this phenomenon, this oscillating effect. It probably means that someone somewhere did this stop and go approach and it ended up perpetuating throughout the traffic from that point behind the car. So the researchers concluded that if even only five percent of the cars on the road were to be autonomous, it would cut down on traffic problems and it would reduce fuel consumption because you wouldn't be doing the stop and go, you'd be using fuel more efficiently, and that in turn would mean less pollution admitted as a result. So it has this ripple effect, which sounds great. It's kind of similar to the safety argument. And then the fact that that would mean less fuel consumption, less volution. You start to see real promise here. However, there's a flip side to this argument that we cannot ignore. While various studies have argued that concentrations of between five to of autonomous cars on the road would really help with traffic problems, and a full conversion to autonomous cars would be presumably even better, this often presupposes that the number of cars on the road would remain pretty much the same as they are today, but some studies suggest that just isn't necessarily the case. For example, the World Economic Forum and the Boston Consulting Group partnered to conduct a study about how driverless cars might impact traffic in and around Boston. According to their research, they found that driverless cars would actually increase traffic in downtown Boston. Their findings suggested that people would opt to use driverless cars over other alternatives such as public transportation. So maybe some people would give up driving and let the robots do it. And then in that case, you're just changing a human driven car for a robot driven car. No big deal, it's a one for one swamp. But there would be other people who would give up taking the bus or the train in favor of more direct door to door service using these driverless car ride hailing services, And so it's possible that for some areas, driverless cars would actually increase congestion in certain municipal corridors like downtown Boston. Now that's not an isolated study, by the way, it's not just the one outlier. AUDI conducted a similar research project in Europe. The company focused on Ingelstadt, Germany. Now that happens to be where AUDI headquarters are, so I suppose that has something to do with their choice for location. But beyond that, Audi said that this city represented a really good case study because it's a medium sized city and it has very limited public transportation options. And those studies suggest that a more important component than adding driverless cars might be just to increase the number of people riding in each vehicle. According to Audi's research, if the average car worked out to holding one point three passengers, so in other words, you just average it all out, you get one point three passengers per per car instead of one point one, commute times would drop as much as even if you increase the number of people taking cars, so true ride sharing would make a bigger difference than autonomous cars. Whether it's a person driving the vehicle or a robot driving the vehicle, the more important part is that you have more people sharing that car. But the company was also quick to point out that what applies to one city may not apply to other cities, so it's not like it's a one size fits all approach. The message seems to be that driver less cars alone are not enough to help solve traffic problems, particularly in cities that have lots of cars in them Atlanta, Houston, Los Angeles. Cities need to invest in public transportation and alternative means to get around, particularly for short distances, and that will reduce the demand for car services as well as just cars in general. But without that investment, we should expect to see a lot more cars hitting the streets, human driven or otherwise. I should point out, though, that these studies do pretty much universally conclude that one thing we would need less of with a proliferation of driverless cars would be parking spaces. Because these studies all presume that we're going to rely on ride haling services with these autonomous cars rather than owning our own autonomous vehicle, we're not likely to go out and buy one. They're going to be way too expensive. It makes way more sense to make use of it in an as needed basis. So if we're using them just when we need them, then we don't have to park them anywhere. So the cars will just drop us off at destinations and then whisk us whisk off on their own to go pick up somebody else while we go and do whatever it is we need to do. There'll be no need to park, so a lot of the space currently dedicated to parking in various cities could be repurposed for other things, which is kind of cool. One great argument for driver list cars is accessibility. People who cannot drive for whatever reason could potentially reclaim a lot of agency in their day to day lives. The elderly people who have vision issues, people like me who have anxiety issues and fear issues. That's not just a benefit for those people, but for everyone. Those people would otherwise never meet. It's benefit to society because it allows them, these people who otherwise would have trouble integrating and contributing to society to do so. I mean, it may not be a big positive with me. If I show up, you might roll your eyes and wonder where the closest eggs it is. But for other people it could be really cool. And that's a pretty darned compelling argument. Now. One of the arguments against Thomas Cars that I have heard is that if driverless cars are safe, and if they reduce accidents, that would mean personal injury attorneys will get less business. But while that might also be true, I think it's pretty grim to suggest a reduction in accidents is a bad thing. So I acknowledge that it would create a hardship on people who are in that profession, I do not necessarily agree that in the end, that is the worst of all the alternatives. Then there are the arguments that driverless cars would free up passengers to do other stuff rather than drive, so, in other words, you don't have to pay attention to driving, so you can do whatever you need to do, like work, or you watch movies, or you catch up on social media. The biggest cautionary warning I think we should consider is that driverless cars may not be effective in all the environments and weather conditions that we encounter. Heavy rain, fog, other weather events could damage or mislead sensors. And since many autonomous vehicles like Waymost cars depend at least in part on pre mapped information. In other words, the company has already gone out and mapped out the regions in which the cars are going to operate. That presents a bit of a challenge. So on the good side, the cars have a baseline to work from. It makes them more accurate, more precise, at least as long as the environment continues to match the data inside the database. But things can change. Let's say there's a catastrophic event. Maybe there's a fire or there's an earthquake. You know, we've had both recently in the news as I record this. In December two thou eighteen, you had the massive earthquake in Alaska. You had the fires in California. These can be transformative events, and it can change the terrain significantly from what it used to be. And if a car is at least partly dependent upon drawing information that has been stored in a database to know where it is within its environment, that can produce a real problem, especially if there are now obstacles in the way. And if you're talking about an event where you might need to evacuate people, that could cause even more issues. Whether the driver list cars are carrying anyone at the time or not, they might be obstacles. So conceivably it could be more dangerous to be in a driverless car in that kind of situation than a human operated car. But here's another crazy thing. We have to consider, what do we even mean when we talk about driverless cars. I'll explain more in just a bit, but let's take a quick break to thank our sponsor. In two thousand fourteen, s a E International, which is an organization that creates standards for the automotive industry, published a document designated J three zero one six and had the title Taxonomy and Definitions for terms related to on road motor vehicle automated driving systems. Other organizations, such as the United States National Highway Traffic Safety Administration had likewise classified autonomous cars in several ways, but ultimately the world has largely adopted this s a E classification, which got an update in ten. And here's how they classify the concept of autonomous cars. They define it in varying levels of autonomy. You start at level zero and you work up to level five, that is least to most autonomous. So a level zero automobile can have automatic warnings and may intervene very briefly, but there's no ongoing control of the vehicle. A human driver maintains control through the vast majority of any trips. So you might have like an emergency braking system that immediately comes on and turns off, but that's it. It's not going to continuously control the car. A level one autonomous car is one in which the car and the human driver share control. A car with adaptive cruise control is an example of a level one autonomous car. The human driver continues to steer the vehicle while adaptive cruise control is in is in effect, but the car itself is what's establishing and maintaining the speed, so it's cooperative. A Level two vehicle can take full control of all driving operations, including steering, accelerating, and breaking, but the human driver is supposed to monitor everything closely and be ready to take over control if necessary. Google's driverless cars in the testing phase fell into this category, as does still to this day. Tesla's autopilot system is a is essentially a level two um, or at least it's marketed as such. Level three autonomous cars are one sophisticate enough that it is no longer necessary for a human operator to monitor the situation and keep eyes on everything. You can take your eyes off the road, in other words, so you can turn your focus to something else, at least temporarily. But the car also can give an alert to the person in the car that indicates they need to take over at some given time, preferably not right the heck now, but more like in five seconds, we're going to hand control back over to you. Level four would allow for total vehicle control to the extent that a person inside the car could take a nap if they wanted to, but only within specific geographical areas. In other words, there would be a geo fencing feature built in, so if the vehicle were to approach the boundary of that area, it would alert the person in the car that their control would be necessary. Otherwise the car would take care of everything. This is the way way MOO one would work, because it's working specifically within the city of Phoenix, Arizona. It's been designed for that, it's been optimized and localized for that. Level five would be a car that never needed to seed control to a human operator. A Level five autonomous car would have to be capable of following all traffic laws wherever it may be, and go from city to city to city without any issues. So when we say driverless cars, we have to determine what level we're talking about. It's not just sufficient to say an autonomous car. You need to understand that autonomy is a spectrum. So a level of five car would be ideal. It's a sort of model I assume people tend to talk about when they're discussing things like the reduction and accidents and the improvement of traffic flow, but no one has quite reached that level. In fact, that's a very difficult problem to solve. Level four is more of the region that companies like Weymo appear to operate within, or at least how they're presenting themselves. The driver less vehicle with no human safety operator would have to work on level four at the very least, or else you're gambling with a person's life, because again, level three would occasionally hand over control back to a human driver. So if you're using a driverless ride hailing service and there's no safety operator in there, it has to be a level four if it's going to be responsible. If the safe operation of the vehicle means that it has to a hand over control to a person on occasion. That doesn't really work for a ride hailing service obviously, unless you've also got a backup driver in the car, at which case you start to ask the question, why are we even bothering with the autonomous part? At least from a service perspective, you might argue it's useful from gathering information to make it better, But when you're talking about an actual ride hailing service, that raises a lot of questions. So level four is where we really need to be At the least, the question is are we really confident that we're at level four now? Now it is possible. I don't mean to suggest that there's no way we're at level four, but there's a lot to a lot of standards we have to meet to truly be level four. There's now there are a lot of very sophisticated technologies that go into these cars, and there's and equally astonishingly sophisticated series of algorithms. Guy eating this technology and way Moo's approach of rolling out in that specific region of Phoenix, Arizona means that the cars can rely upon that localized data to help guide decisions. The car doesn't have to be so versatile that can adapt to different cities with different traffic laws and cultures. And a localized approach means that it's even possible even to program in rules that might only apply to specific parts of Phoenix, like specific intersections. So, for example, to let you know what I'm talking about here, in Atlanta, Georgia, where I live, there are certain communities where it is against the law to make a right hand turn at a red light within that community. There's a community in the Atlanta area called Avondale Estates is against the law to take a right turn on a red light there. It doesn't matter what the traffic situation is. If you're in Avendale, states don't make a right on red. But in other parts of Atlanta, making a right on a red is perfectly legal. So a driver less car service in the Atlanta area would need to have that information built in so it could integrate into existing traffic while obeying the law. It wouldn't hold up traffic in neighborhoods where a right on red was permissible, and it wouldn't break the law by taking a right on red in a place where it wasn't that would be a requirement that might work if the service is truly local. The service might also be bounded by a geo fencing feature like I mentioned that prevents writers from specifying a drop off point beyond a certain distance of the center of operations. So, in other words, if I said I'm in Atlanta, I want to get picked up, and I want to be dropped off in Charleston, in a different state, then if I tried to select that distance, that location, or that destination, the app could tell me I'm sorry, that's not possible. So you might find out you can get a ride with a driverless car, but only across town, not across the state or into another state. And it's also good to remember, just to couch our expectation, what the conditions of Phoenix, Arizona are like. The city has an average annual rainfall of eight point zero four inches for the entirety of the year. In Atlanta, the annual rainfall average is closer to fifty inches. Now, what that tells us is that you get fewer rainy days and Phoenix than you would in Atlanta, and presumably you would have fewer days with very heavy rain, which is important because that could interfere with a car's sensors. The annual low temperature and Phoenix is sixty three point four degrees fahrenheit, which is about seventeen point four degrees celsius. Even in the coldest month of the year for Phoenix, which is December, the average low is forty five degrees fahrenheit or seven point two degrees celsius, which is chilly for someone like me who grew up in the Southeast. I might want to wear a jacket, maybe even a light coat. But it's not freezing, right, you haven't dropped below the temperature of freezing. The likelihood that the driver lest cars and Phoenix, Arizona would encounter conditions like ice or snow is very low, not impossible, but very unlikely. So you could argue that rolling out a driverless car service in Phoenix is like riding a bike with training wheels. But on the other hand, it makes way more sense to me to start off in areas where conditions are pretty consistent so that you can continue to gather information about driving conditions and driver behaviors and work with that knowledge and build it into future algorithms and improvements. That might mean we'll see a very gradual rollout of autonomous car technology in cities that are most amenable to it, let's say, so it might be a very controlled rollout, and it might be a long time before we see it everywhere. Now, I think I'm going to close with one other big point. There are tons more that I could talk about, obviously, but the one I want to talk about here. Another warning is I just mentioned the idea of Weymo gathering useful information from the driverless car service. It's it's launching in Phoenix, but that also ties into information about us as potential customers. Weymo is part of Alphabet. Alphabet's the holding company for Google. Google is ultimately in the business of information. So consider this for a moment. Using a driverless car service, it's just like using any other ride hailing service. It requires that you establish an account with that service. That account is tied to your identity. You have to be able to pay for the service, so ultimately this has to be linked to some legitimate form of payment, which in turn tends to be linked to your specific identity. And you use an app to call for a ride, you designate where you're going to be picked up and where you want to be dropped off. That means part of your payment is not in money, it's actually in the information you're giving the service. The service learns where you're coming from, where you're going, and when you were traveling, and that is valuable information. You can bet it will get tied to other stuff like advertising. So that's a very good thing to remember too. It's not to say don't use the service, but be aware of everything that's going on, be an informed consumer. Driver lest cars might transform our world, but they might do so in ways that aren't entirely to our liking in the future. Like I said, I'll probably do some more episodes about this to touch on some of those other elements that I mentioned at the top of the show but didn't really get a chance to go into. And I didn't really talk about things like the impact that driverless cars might have on employment, for example, so people who do deliveries, you know, postal service, the trucking industry, the taxi industry, right hailing businesses. I haven't talked about how driver lest cars could potentially have a big impact on those those various industries, but I have rambled about autonomous cars for more than a week now, so there's plenty for us to come back to later. For the time being, we're gonna put this to bed. We're gonna switch to other topics. I'm gonna take the air out of an entirely different futuristic idea in our next episode. If you guys have suggestions for topics I should cover in future episodes of tech Stuff, whether it's a technology, a company, Maybe there's a person in tech I should talk about. Maybe there's someone I should interview or have on as a guest host. Let me know. Send me an email the addresses tech Stuff at how stuff works dot com. Head on over to our website that's tech Stuff podcast dot com. You'll find other ways to contact me, plus a link to our store. Remember every purchase you make goes to help the show, and we greatly appreciate it. And I'll talk to you again really so soon for more on this and thousands of other topics. Is that how stuff works dot com.

Hey guys, this is Jonathan. I have a quick announcement before we get to today's episode, which is sort of a capper on the autonomous car episodes we listened to last week, and that is starting this week, Tech Stuff is going back to publishing two new episodes and one classic episode each week, so we're cutting back a little bit. This is so that I can spend more time putting these episodes together, making sure they are the absolute best I can make them, as well as making time to work on some other shows like The Brink. If you're not listening to The Brink, you should go check that out and some more shows that might be coming down in the near future once we get through pilot season. Here at How Stuff Works, and I just want to let you guys all be aware of that nothing bad is happening. These are good things. It means that I get to spend more time on this stuff and I'm not quite as a as rushed as I have been for the last half year. So enjoy and I'll see you guys again on Wednesday. Get in touch with technology with tech Stuff from how stuff works dot com. Hey there, and welcome to tech Stuff. I'm your host Jonathan Strickland. I'm an executive producer with How Stuff Works, and I heart radio and I love all things tech, and we are continuing our long journey with autonomous cars. Actually we are concluding it at this point, even though as I get to the end of the episode you will hear about how I could cover even more. But out of consideration for you, my dear listeners, I'll move on to other topics for a while. But last week I did a whole series of episodes about the history of autonomous cars, and we started with the science fiction visions of the nineteen twenties and radio controlled vehicles up to the upcoming launch of the Weymo one service in Phoenix, Arizona, which maybe in operation by the time you hear this. And I mentioned how over the last several years, in particular, a few accidents, a couple of them fatal accidents, have brought a lot more critical attention to driverless cars, even as numerous services prepared to unleash thousands of them on the streets of several cities across the world. In this episode, I'm going to talk about the arguments for and against autonomous cars, or at least against early autonomous car deployment. I'm gonna talk about how these things are complicating factors. It's not just a technological problem. In fact, I don't really get into it in this episode, but there are entire issues with regulation and legislation. Like many other technologies, autonomous cars has outpaced the law, and so there are a lot of places that are trying to take into account what driverless cars in the streets would actually mean, and how do you legislate that. Who do you find at fault in the event of an accident? These are still largely open questions. There are other ones that I haven't touched on in this episode either, For example, the trolley problem, the idea that if you have to build into your car system a decision making process to follow in the event that an injury is unavoidable, you've got and and a scenario where no matter what, someone is going to get hurt. How does the car make the choice of who gets hurt. That's a huge question in ethics, in artificial intelligence in general, and in autonomous cars in particular. I don't really go into that in this episode, so there are still many topics in this field that I will have to come back to at some future point. However, that being said, there's still a ton of stuff to talk about outside of that, and I'm going to start off by exam and some of the arguments for autonomous cars, as well as the criticisms leveled at those arguments now before we jump into this, because it's going to sound like I'm totally down on the concept, like I'm like, I'm really against the idea of driverless cars. That's what it might sound like to you as we go through this episode. But that is not the case. And I'm gonna be real with you, guys. I don't drive. I I do not have a driver's license. And I don't talk about this very much because it's hard to talk about because driving in the United States is such a a common experience among many many people. I mean, unless you're living in a city that has incredible public transportation and driving is incredibly expensive, then you probably drive a car, Especially in the Southeast where I live. Owning a car and driving a car is very much tied with the thought of independence. But I don't have a driver's license. I actually have a phobia about driving. I end up locking up when I get behind the wheel of a car and I become a danger on the road, and I'm aware of that, and then that also feeds into my fear and it gets worse and worse. So you see where I'm going. I I have realized that I am a danger to myself and others if I get behind the wheel of a car. I am not a good driver. So this is an incredibly frustrating condition for me, and it's difficult for me to talk about because there's a big stigma against this. But it means that I frequently have to rely upon other people to help me get around, and that means I often feel like a burden, even when people go out of their way to dismiss that idea, saying no, no, no, no, I want to give you a ride. I want you to be involved in this thing. I can't help but feel otherwise. So I very much want a future in which we have a reliable, safe, autonomous car service where I can call upon a car anytime need and it will show up. And I'm not putting anyone out right, I'm not making it inconvenient for somebody else. It's a service. I'm paying for it. Everything's cool. I'm eager to see that kind of thing across the world. That being said, I also recognize that I can't allow my desire for that technology to come to maturity to guide all my thinking on the matter to bias my point of view. The responsible thing to do is to hold up the facts to critical thought and consideration. After all, we're ultimately talking about enormous vehicles traveling at potentially very high speeds, and that's a lot of force. So it's only right that we look at all the facts before we leap to any conclusions. And knowing how badly I want autonomous cars to become a thing has to be a warning sign to me. I have to say, well, I need to be extra careful when I'm looking at the facts so that I don't just become latch on to the most optimistic outlooks and then just, you know, deny that there are any problems. So the first big argument for autonomous cars in favor of them tends to be that, assuming you have cars that perform safely and consistently, you would cut way back on the number of fatalities and injuries caused by car accidents. And the general line of reasoning is that machines can react in a fraction of the time that people can, which is absolutely true, and a well designed vehicle, one that has the proper sensors, could maintain a three D sixty degree awareness at all times, whereas a human can only focus on a small part of their overall field of view, which is locked to whichever way they happen to turn their face at any given time. That's also true assuming you've designed the sensors and the processing system properly so that it can accept that data and analyze it in real time and be able to react to it in real time. Now, this argument that autonomous cars will save lives also assumes that the software the algorithms guiding the car's behaviors have been properly designed and tested and they're reliable. Whether that means programming the car for specific scenarios or using a process like machine learning in which an autonomous system is quote unquote taught how to handle various driving scenarios. Now, that assumption is particularly important, and it also is one that we should be particularly critical of. Now, I don't mean to suggest that it is impossible to design a sufficiently adept system or a sufficiently a depth set of algorithms but rather we should be as certain as we can and be as we can reasonably expect us to be that such systems meet are required levels of reliability before we start getting into driverless vehicles for any sort of you know, given trip. Now, all that being said, statistics do show that humans are far from perfect drivers. The U s Department of Transportation produces reports from the Fatality Analysis Reporting System or FARS, and yes, that's pretty grim, but it's very important work. The two thousand sixteen numbers, which were some of the more recent numbers that I could get hold of, say that in the United States in two thousand and sixteen, there were thirty four thousand, four hundred thirty nine recorded traffic accidents that had one or more fatalities. So if you look at all the actual deaths, you're looking at thirty seven thousand, four hundred sixty one people who lost their lives from car accidents. So when you take into a account the population of the United States, that averages out to eleven point six deaths due to car accidents per one hundred thousand people. When you break it out state by state and you look at their statistics, you see where the concentration is highest. Mississippi had the highest fatality rate by this metric, at twenty three point one fatalities per one hundred thousand people. The report also takes into account the number of vehicle miles traveled. It's very important as well, because that tells you how frequently people were on the roads and how far they went, and how much time they were spending on the roads. So in sixteen, the estimated total miles traveled in vehicles in the United States was three trillion, two hundred twenty billion, six hundred seventy seven million miles. It's a really important number that we're going to come back to in just a second. But another important number is that in the United States, experts estimate that human error causes somewhere between and of all car accidents. Uh, there's discrepancy because it all depends upon which study you're looking at, but the the message there is the overwhelming majority of car accidents in the United States are caused by human error, as opposed to say, mechanical failure or other extenuating circumstances like a tree falling on a car. So it's easy to see how the thought of removing the human element from driving could dramatically reduce the number of crashes. But again, this assumes that the driver list systems are in fact more reliable more safe than human drivers. Now they likely are, but it's hard for us to say we've definitively proven this. In fact, that would be misleading to say we have definitively proven these cars are more reliable and more safe under automated control than human control, And the reason for that is because of the amount of data. This is where we start to really get scientific with our thinking. So the safety argument says, imagine a future in which all cars are autonomous and can detect one another. They can react quickly to changes in the environment much much faster than humans can. So then imagine how drastic the reduction of fatalities would be. Those nearly forty thou people who died in would still be alive in a world where those cars were the way we got around. But this presupposes that the autonomous cars would perform as we would like them too, and that still remains to be seen. Google's way MOW, which is the company that I think is the the foremost authority on driverless cars as far as public awareness is concerned. At least, they like to point out that their self driving vehicles have passed the ten million mile mark. In other words, autonomous vehicles have driven more than ten million miles and there have been no accidents in Weymo vehicles that resulted in a fatality from their operation in autonomous mode. And ten million is a big number. Ten million miles. That's a lot of distance. But remember that in humans in the United States drove three trillion miles in one year. You know, Weymo is looking at the entire history of their autonomous car program in one year. Because of the number of drivers and how much we drive, humans racked up more than three trillion miles. So we have more human drivers on the streets, which also makes it difficult to draw any sort of comparison between human drivers and autonomous drivers. There are more human drivers out there and they're driving more miles than autonomous cars are, so it's not an apples to apples kind of comparison. We cannot say definitively that autonomous cars are safer. They just haven't been driving enough and there aren't enough of them out there for us to make that kind of general conclusion. Does that mean it's wrong? Not necessarily. It just means we lack the information to be able to say it definitively, and without that information, there's doubt, and when there's doubt, you have problems. In the last couple of years, there have been some pretty high profile incidents involving autonomous cars and accidents, and these accidents don't necessarily mean that the safety argument for autonomous cars is invalid. We haven't seen a rash of driver list car accidents. We've only seen a couple like a few dozen, and many of those aren't even the fault of the autonomous system, but rather the fault of human drivers who happened to be on the road at the same time. However, even if we were to sift through all of the data and prove definitively, statistically speaking, that driverless cars are safer than one driven by an average person, it might not matter at the end of the day. Now, this is because we humans are not entirely rational, and I am guessing this does not come as a shocking revelation to you guys. Psychologically, it can be easier for us to accept that driving a car on roads dominated by other human drivers is risky. We understand this because we know that driving, while often a mundane task that we can kind of do without even devoting our full attention to it, can sometimes be punctuated by moments of unique circumstances, and that in those circumstances, the wrong reaction could cause an accident. This is something we can wrap our minds around. We can accept this. We know accidents happen, sometimes we are in them, sometimes we cause them, and that is something that we can accept. But there's a bit of a leap for an awful lot of people when you shift that two vehicles that are controlled not by human drivers but by machines and algorithms. Suddenly the same people who accept that driving can be inherently dangerous in certain situations find it much more difficult to accept when humans are no longer directly responsible. The journal Risk Analysis published the results of a survey in which people were asked to determine how safe an autonomous vehicle would have to be before they would feel comfortable with the idea of them being on the road. The results were really interesting. They showed that, on average, people felt that autonomous cars would have to be proven to be four to five times safer than human operated vehicles for them to feel comfortable four to five times safer. So let that sink in because by this argument, even if the driverless car was twice as safe as the average human driver, which by definition would mean that if you had enough of these replacing human drivers out on the road, would mean you would have fewer injuries, fewer cases of vehicle or property damage, fewer fatalities. But that wouldn't be enough for these people to feel comfortable even if they knew that, as a result, fewer people were being hurt or killed, there was less damage. And when you turn this argument on its side, it's kind of like saying it's okay if a certain number of people are hurt or killed because we didn't shift to autonomous cars. It's only when autonomous cars are far more safe and capable four to five times than human drivers that people start to feel better about them. And I think you could argue that it's a pretty extreme double standard. We're not requiring humans to undergo driving tests more frequently to prove that they are just as safe or not if not safer, than they were the last time they took a test. We're not requiring people to do that. But it is the sort of stuff we're demanding from the machines, and it's also something I can kind of understand because it means taking the humans out of the equation, or seeming to, since these cars and algorithms are ultimately designed by humans and that's scary. The idea of entrusting lives to machines so is a bit of a paradox. So to summarize the safety part of the argument, we don't yet have enough information to really know how safe autonomous cars are compared to human drivers, although I suspect they are as a whole safer, but again hard to tell because you're extrapolating a lot from a very small data sample size. We don't know because we lack that information. There just aren't enough driverless cars in the road that haven't driven nearly enough miles and enough different environments in order for us to draw any firm conclusions. We also have to remember that different companies are taking very different approaches to driverless cars, so we can't make any sweeping statements because one approach might be better in some situations and worse than others, with no single approach being the best for all scenarios or regions. We do know human drivers are fallible and that we are at fault in the overwhelming majority of car accidents, and we suspect that eventually will have autonomous cars that can operate more safely than people. We just don't know if we're there already or if we still have a ways to go. Now. I'll have more to say about the arguments for and the arguments against autonomous cars in just a second, but first let's take a quick break to thank our sponsor. So, the safety argument also presents a couple of other possibilities. Besides an improvement in safety, assuming that everything is working properly. Uh, it also presents the possibility that autonomous cars might require less material and less energy to operate. So here here's the thought. If your vehicles are proven to be safe and you don't have to worry about them getting into accidents, let's say that you have reached a point where you're ready to really deploy these, then you don't have to build your vehicles quite as sturdily. Right. One of the reasons why cars and trucks and other you know, consumer vehicles way as much as they do is because the materials they're made out of are in part there to protect the people inside in case of accident. But if you've removed the possibility of accident, then you don't have to build them quite as Hardy will say, so it means you could go to lighter materials to make your vehicles. Well, lighter vehicles need less energy to move around than a heavier vehicle does. You don't have to put as much force to get the vehicle moving or to stop it. So if you reduce the weight of vehicles on the road, you also reduce the amount of energy you need to move them around. That also means less fuel consumption, which could be in the vehicle itself if it's a gasoline powered vehicle or hydrogen powered vehicle, or in a power plant that produces the electricity that is going to charge those electric vehicles batteries. And that sounds pretty good, except I would argue we can't really afford to get fully behind switching over to light vehicles until we reach a real tipping point of the percentage of driverless cars versus human operated ones, because even if we assume robot cars are perfect, which again this is just like hypothetical, just assume that they work exactly the way they're supposed to. You still have humans on the road and they could still cause accidents. And for that reason, we can't really afford to essue safety features that add weight to vehicles just yet. We need to still have those considerations in place. Another big argument for autonomous cars is that they could have a positive influence on traffic, meaning a reduction in traffic and traffic jams and congestion. But the studies on this subject of conflicting answers. The National Science Foundation funded a study to look into this matter, and the study consulted with auto industry experts and government officials. And in that study, the researchers created a simulation in which one autonomous vehicle, a computer controlled simulated vehicle, circled a track with at least twenty vehicles driven by humans. It was a simulation that used VR, so these were VR cars driven by human beings and not just you know, twenty simulated human beings, because then you're talking about two different types of autonomous cars, and that wouldn't tell you anything anyway. This experiment saw that humans create a stop and go oscillation effect in traffic, but that autonomous cars don't cause this kind of oscillation effect, and this can happen in things like when cars change lanes or they need to exit the highway or into the highway, and this can lead to traffic jams. So if you've ever been on highway and you hit a super slow section of traffic and then the traffic starts to clear up, and you never see any reason for the traffic jam, right, there's no accident, there's no reason, nothing to rubber neck at. You might wonder what the heck happened, Well, that's this phenomenon, this oscillating effect. It probably means that someone somewhere did this stop and go approach and it ended up perpetuating throughout the traffic from that point behind the car. So the researchers concluded that if even only five percent of the cars on the road were to be autonomous, it would cut down on traffic problems and it would reduce fuel consumption because you wouldn't be doing the stop and go, you'd be using fuel more efficiently, and that in turn would mean less pollution admitted as a result. So it has this ripple effect, which sounds great. It's kind of similar to the safety argument. And then the fact that that would mean less fuel consumption, less volution. You start to see real promise here. However, there's a flip side to this argument that we cannot ignore. While various studies have argued that concentrations of between five to of autonomous cars on the road would really help with traffic problems, and a full conversion to autonomous cars would be presumably even better, this often presupposes that the number of cars on the road would remain pretty much the same as they are today, but some studies suggest that just isn't necessarily the case. For example, the World Economic Forum and the Boston Consulting Group partnered to conduct a study about how driverless cars might impact traffic in and around Boston. According to their research, they found that driverless cars would actually increase traffic in downtown Boston. Their findings suggested that people would opt to use driverless cars over other alternatives such as public transportation. So maybe some people would give up driving and let the robots do it. And then in that case, you're just changing a human driven car for a robot driven car. No big deal, it's a one for one swamp. But there would be other people who would give up taking the bus or the train in favor of more direct door to door service using these driverless car ride hailing services, And so it's possible that for some areas, driverless cars would actually increase congestion in certain municipal corridors like downtown Boston. Now that's not an isolated study, by the way, it's not just the one outlier. AUDI conducted a similar research project in Europe. The company focused on Ingelstadt, Germany. Now that happens to be where AUDI headquarters are, so I suppose that has something to do with their choice for location. But beyond that, Audi said that this city represented a really good case study because it's a medium sized city and it has very limited public transportation options. And those studies suggest that a more important component than adding driverless cars might be just to increase the number of people riding in each vehicle. According to Audi's research, if the average car worked out to holding one point three passengers, so in other words, you just average it all out, you get one point three passengers per per car instead of one point one, commute times would drop as much as even if you increase the number of people taking cars, so true ride sharing would make a bigger difference than autonomous cars. Whether it's a person driving the vehicle or a robot driving the vehicle, the more important part is that you have more people sharing that car. But the company was also quick to point out that what applies to one city may not apply to other cities, so it's not like it's a one size fits all approach. The message seems to be that driver less cars alone are not enough to help solve traffic problems, particularly in cities that have lots of cars in them Atlanta, Houston, Los Angeles. Cities need to invest in public transportation and alternative means to get around, particularly for short distances, and that will reduce the demand for car services as well as just cars in general. But without that investment, we should expect to see a lot more cars hitting the streets, human driven or otherwise. I should point out, though, that these studies do pretty much universally conclude that one thing we would need less of with a proliferation of driverless cars would be parking spaces. Because these studies all presume that we're going to rely on ride haling services with these autonomous cars rather than owning our own autonomous vehicle, we're not likely to go out and buy one. They're going to be way too expensive. It makes way more sense to make use of it in an as needed basis. So if we're using them just when we need them, then we don't have to park them anywhere. So the cars will just drop us off at destinations and then whisk us whisk off on their own to go pick up somebody else while we go and do whatever it is we need to do. There'll be no need to park, so a lot of the space currently dedicated to parking in various cities could be repurposed for other things, which is kind of cool. One great argument for driver list cars is accessibility. People who cannot drive for whatever reason could potentially reclaim a lot of agency in their day to day lives. The elderly people who have vision issues, people like me who have anxiety issues and fear issues. That's not just a benefit for those people, but for everyone. Those people would otherwise never meet. It's benefit to society because it allows them, these people who otherwise would have trouble integrating and contributing to society to do so. I mean, it may not be a big positive with me. If I show up, you might roll your eyes and wonder where the closest eggs it is. But for other people it could be really cool. And that's a pretty darned compelling argument. Now. One of the arguments against Thomas Cars that I have heard is that if driverless cars are safe, and if they reduce accidents, that would mean personal injury attorneys will get less business. But while that might also be true, I think it's pretty grim to suggest a reduction in accidents is a bad thing. So I acknowledge that it would create a hardship on people who are in that profession, I do not necessarily agree that in the end, that is the worst of all the alternatives. Then there are the arguments that driverless cars would free up passengers to do other stuff rather than drive, so, in other words, you don't have to pay attention to driving, so you can do whatever you need to do, like work, or you watch movies, or you catch up on social media. The biggest cautionary warning I think we should consider is that driverless cars may not be effective in all the environments and weather conditions that we encounter. Heavy rain, fog, other weather events could damage or mislead sensors. And since many autonomous vehicles like Waymost cars depend at least in part on pre mapped information. In other words, the company has already gone out and mapped out the regions in which the cars are going to operate. That presents a bit of a challenge. So on the good side, the cars have a baseline to work from. It makes them more accurate, more precise, at least as long as the environment continues to match the data inside the database. But things can change. Let's say there's a catastrophic event. Maybe there's a fire or there's an earthquake. You know, we've had both recently in the news as I record this. In December two thou eighteen, you had the massive earthquake in Alaska. You had the fires in California. These can be transformative events, and it can change the terrain significantly from what it used to be. And if a car is at least partly dependent upon drawing information that has been stored in a database to know where it is within its environment, that can produce a real problem, especially if there are now obstacles in the way. And if you're talking about an event where you might need to evacuate people, that could cause even more issues. Whether the driver list cars are carrying anyone at the time or not, they might be obstacles. So conceivably it could be more dangerous to be in a driverless car in that kind of situation than a human operated car. But here's another crazy thing. We have to consider, what do we even mean when we talk about driverless cars. I'll explain more in just a bit, but let's take a quick break to thank our sponsor. In two thousand fourteen, s a E International, which is an organization that creates standards for the automotive industry, published a document designated J three zero one six and had the title Taxonomy and Definitions for terms related to on road motor vehicle automated driving systems. Other organizations, such as the United States National Highway Traffic Safety Administration had likewise classified autonomous cars in several ways, but ultimately the world has largely adopted this s a E classification, which got an update in ten. And here's how they classify the concept of autonomous cars. They define it in varying levels of autonomy. You start at level zero and you work up to level five, that is least to most autonomous. So a level zero automobile can have automatic warnings and may intervene very briefly, but there's no ongoing control of the vehicle. A human driver maintains control through the vast majority of any trips. So you might have like an emergency braking system that immediately comes on and turns off, but that's it. It's not going to continuously control the car. A level one autonomous car is one in which the car and the human driver share control. A car with adaptive cruise control is an example of a level one autonomous car. The human driver continues to steer the vehicle while adaptive cruise control is in is in effect, but the car itself is what's establishing and maintaining the speed, so it's cooperative. A Level two vehicle can take full control of all driving operations, including steering, accelerating, and breaking, but the human driver is supposed to monitor everything closely and be ready to take over control if necessary. Google's driverless cars in the testing phase fell into this category, as does still to this day. Tesla's autopilot system is a is essentially a level two um, or at least it's marketed as such. Level three autonomous cars are one sophisticate enough that it is no longer necessary for a human operator to monitor the situation and keep eyes on everything. You can take your eyes off the road, in other words, so you can turn your focus to something else, at least temporarily. But the car also can give an alert to the person in the car that indicates they need to take over at some given time, preferably not right the heck now, but more like in five seconds, we're going to hand control back over to you. Level four would allow for total vehicle control to the extent that a person inside the car could take a nap if they wanted to, but only within specific geographical areas. In other words, there would be a geo fencing feature built in, so if the vehicle were to approach the boundary of that area, it would alert the person in the car that their control would be necessary. Otherwise the car would take care of everything. This is the way way MOO one would work, because it's working specifically within the city of Phoenix, Arizona. It's been designed for that, it's been optimized and localized for that. Level five would be a car that never needed to seed control to a human operator. A Level five autonomous car would have to be capable of following all traffic laws wherever it may be, and go from city to city to city without any issues. So when we say driverless cars, we have to determine what level we're talking about. It's not just sufficient to say an autonomous car. You need to understand that autonomy is a spectrum. So a level of five car would be ideal. It's a sort of model I assume people tend to talk about when they're discussing things like the reduction and accidents and the improvement of traffic flow, but no one has quite reached that level. In fact, that's a very difficult problem to solve. Level four is more of the region that companies like Weymo appear to operate within, or at least how they're presenting themselves. The driver less vehicle with no human safety operator would have to work on level four at the very least, or else you're gambling with a person's life, because again, level three would occasionally hand over control back to a human driver. So if you're using a driverless ride hailing service and there's no safety operator in there, it has to be a level four if it's going to be responsible. If the safe operation of the vehicle means that it has to a hand over control to a person on occasion. That doesn't really work for a ride hailing service obviously, unless you've also got a backup driver in the car, at which case you start to ask the question, why are we even bothering with the autonomous part? At least from a service perspective, you might argue it's useful from gathering information to make it better, But when you're talking about an actual ride hailing service, that raises a lot of questions. So level four is where we really need to be At the least, the question is are we really confident that we're at level four now? Now it is possible. I don't mean to suggest that there's no way we're at level four, but there's a lot to a lot of standards we have to meet to truly be level four. There's now there are a lot of very sophisticated technologies that go into these cars, and there's and equally astonishingly sophisticated series of algorithms. Guy eating this technology and way Moo's approach of rolling out in that specific region of Phoenix, Arizona means that the cars can rely upon that localized data to help guide decisions. The car doesn't have to be so versatile that can adapt to different cities with different traffic laws and cultures. And a localized approach means that it's even possible even to program in rules that might only apply to specific parts of Phoenix, like specific intersections. So, for example, to let you know what I'm talking about here, in Atlanta, Georgia, where I live, there are certain communities where it is against the law to make a right hand turn at a red light within that community. There's a community in the Atlanta area called Avondale Estates is against the law to take a right turn on a red light there. It doesn't matter what the traffic situation is. If you're in Avendale, states don't make a right on red. But in other parts of Atlanta, making a right on a red is perfectly legal. So a driver less car service in the Atlanta area would need to have that information built in so it could integrate into existing traffic while obeying the law. It wouldn't hold up traffic in neighborhoods where a right on red was permissible, and it wouldn't break the law by taking a right on red in a place where it wasn't that would be a requirement that might work if the service is truly local. The service might also be bounded by a geo fencing feature like I mentioned that prevents writers from specifying a drop off point beyond a certain distance of the center of operations. So, in other words, if I said I'm in Atlanta, I want to get picked up, and I want to be dropped off in Charleston, in a different state, then if I tried to select that distance, that location, or that destination, the app could tell me I'm sorry, that's not possible. So you might find out you can get a ride with a driverless car, but only across town, not across the state or into another state. And it's also good to remember, just to couch our expectation, what the conditions of Phoenix, Arizona are like. The city has an average annual rainfall of eight point zero four inches for the entirety of the year. In Atlanta, the annual rainfall average is closer to fifty inches. Now, what that tells us is that you get fewer rainy days and Phoenix than you would in Atlanta, and presumably you would have fewer days with very heavy rain, which is important because that could interfere with a car's sensors. The annual low temperature and Phoenix is sixty three point four degrees fahrenheit, which is about seventeen point four degrees celsius. Even in the coldest month of the year for Phoenix, which is December, the average low is forty five degrees fahrenheit or seven point two degrees celsius, which is chilly for someone like me who grew up in the Southeast. I might want to wear a jacket, maybe even a light coat. But it's not freezing, right, you haven't dropped below the temperature of freezing. The likelihood that the driver lest cars and Phoenix, Arizona would encounter conditions like ice or snow is very low, not impossible, but very unlikely. So you could argue that rolling out a driverless car service in Phoenix is like riding a bike with training wheels. But on the other hand, it makes way more sense to me to start off in areas where conditions are pretty consistent so that you can continue to gather information about driving conditions and driver behaviors and work with that knowledge and build it into future algorithms and improvements. That might mean we'll see a very gradual rollout of autonomous car technology in cities that are most amenable to it, let's say, so it might be a very controlled rollout, and it might be a long time before we see it everywhere. Now, I think I'm going to close with one other big point. There are tons more that I could talk about, obviously, but the one I want to talk about here. Another warning is I just mentioned the idea of Weymo gathering useful information from the driverless car service. It's it's launching in Phoenix, but that also ties into information about us as potential customers. Weymo is part of Alphabet. Alphabet's the holding company for Google. Google is ultimately in the business of information. So consider this for a moment. Using a driverless car service, it's just like using any other ride hailing service. It requires that you establish an account with that service. That account is tied to your identity. You have to be able to pay for the service, so ultimately this has to be linked to some legitimate form of payment, which in turn tends to be linked to your specific identity. And you use an app to call for a ride, you designate where you're going to be picked up and where you want to be dropped off. That means part of your payment is not in money, it's actually in the information you're giving the service. The service learns where you're coming from, where you're going, and when you were traveling, and that is valuable information. You can bet it will get tied to other stuff like advertising. So that's a very good thing to remember too. It's not to say don't use the service, but be aware of everything that's going on, be an informed consumer. Driver lest cars might transform our world, but they might do so in ways that aren't entirely to our liking in the future. Like I said, I'll probably do some more episodes about this to touch on some of those other elements that I mentioned at the top of the show but didn't really get a chance to go into. And I didn't really talk about things like the impact that driverless cars might have on employment, for example, so people who do deliveries, you know, postal service, the trucking industry, the taxi industry, right hailing businesses. I haven't talked about how driver lest cars could potentially have a big impact on those those various industries, but I have rambled about autonomous cars for more than a week now, so there's plenty for us to come back to later. For the time being, we're gonna put this to bed. We're gonna switch to other topics. I'm gonna take the air out of an entirely different futuristic idea in our next episode. If you guys have suggestions for topics I should cover in future episodes of tech Stuff, whether it's a technology, a company, Maybe there's a person in tech I should talk about. Maybe there's someone I should interview or have on as a guest host. Let me know. Send me an email the addresses tech Stuff at how stuff works dot com. Head on over to our website that's tech Stuff podcast dot com. You'll find other ways to contact me, plus a link to our store. Remember every purchase you make goes to help the show, and we greatly appreciate it. And I'll talk to you again really so soon for more on this and thousands of other topics. Is that how stuff works dot com.