WEBVTT - The State of Autonomous Vehicles 0:00:04.240 --> 0:00:07.240 Welcome to tech Stuff, a production of I Heart Radios 0:00:07.320 --> 0:00:13.680 How Stuff Works. Hey there, and welcome to tech Stuff. 0:00:13.720 --> 0:00:16.279 I'm your host, Jonathan Strickland. I'm an executive producer with 0:00:16.360 --> 0:00:18.479 I Heart Radio and I love all things tech. And 0:00:18.520 --> 0:00:21.160 if you haven't noticed already, yeah, my voice is all 0:00:21.239 --> 0:00:23.960 sorts of jacked up because I'm getting over a cold. 0:00:24.040 --> 0:00:29.560 I apologize for that, but the tech must go on now. 0:00:29.680 --> 0:00:32.879 Just a couple of years ago, the tech world in 0:00:32.960 --> 0:00:37.720 general was pretty optimistic about autonomous cars, and I include 0:00:37.760 --> 0:00:42.120 myself in that group. I remember seeing the remarkable progress 0:00:42.159 --> 0:00:45.680 that had come out from the first DARPA Grand Challenge 0:00:46.000 --> 0:00:49.800 up to about I don't know ten or so, and 0:00:49.800 --> 0:00:51.680 and it seemed like we were just on the verge 0:00:51.720 --> 0:00:55.360 of having fleets of robo taxis at our back and call. 0:00:55.960 --> 0:00:59.040 But now we've gone on for several more years and 0:00:59.040 --> 0:01:01.520 we're still at a point are only a handful of 0:01:01.560 --> 0:01:05.920 companies are conducting limited tests. Plus there have been some 0:01:06.160 --> 0:01:11.039 high profile cases of accidents involving vehicles operating under autonomous 0:01:11.120 --> 0:01:15.560 or semi autonomous modes that ended in tragedy. So in 0:01:15.560 --> 0:01:18.039 this episode, we're going to take a look at autonomous 0:01:18.080 --> 0:01:21.479 cars and where we stand today. Now, let's start with 0:01:21.840 --> 0:01:24.399 I think it helps if we run through the levels 0:01:24.520 --> 0:01:28.080 of autonomy, and not everyone uses these levels to talk 0:01:28.080 --> 0:01:31.360 about autonomy, and to be honest, the barriers between levels 0:01:31.560 --> 0:01:34.520 are a bit fuzzy, and sometimes we're not really able 0:01:34.560 --> 0:01:38.520 to say where we are at as far as levels 0:01:38.560 --> 0:01:43.119 of autonomy. We can look back at previous developments and say, 0:01:43.120 --> 0:01:45.680 all right, well, judging on where we are now, we'd 0:01:45.680 --> 0:01:48.480 say that this falls into level two or level three, 0:01:48.800 --> 0:01:51.120 But it can be a little difficult to see what 0:01:51.240 --> 0:01:57.160 level we are currently in without you know, truly remarkable evidence. 0:01:57.600 --> 0:01:59.720 But in general, this is a useful way to talk 0:01:59.720 --> 0:02:03.840 about how far along we are uh as far as 0:02:03.880 --> 0:02:08.959 getting fully autonomous cars. So technically the levels range from 0:02:09.040 --> 0:02:13.520 zero to five, so that means there are really six levels. However, 0:02:13.840 --> 0:02:17.040 level zero really means there is no autonomy at all. 0:02:17.520 --> 0:02:19.920 So with that type of vehicle, the human driver is 0:02:19.960 --> 0:02:24.720 responsible for all operations of the vehicle. Every driving task 0:02:25.040 --> 0:02:28.520 is handled by the driver alone. So some folks will 0:02:28.520 --> 0:02:32.120 say that they're really just five levels of autonomy. Zero 0:02:32.160 --> 0:02:36.359 would refer to vehicles that really, honestly, they don't exist 0:02:36.400 --> 0:02:39.880 that much anymore. Now you might be thinking but hey, 0:02:40.120 --> 0:02:43.640 Mr smarty Pants podcast person. I drive a car and 0:02:43.680 --> 0:02:46.919 it doesn't have any autonomous vehicle features, but depending upon 0:02:46.919 --> 0:02:50.760 whom you ask, features like power steering or anti lock 0:02:50.880 --> 0:02:56.000 breaks or cruise control and other pretty common features fall 0:02:56.120 --> 0:02:59.360 into the low level autonomous range. It doesn't mean your 0:02:59.440 --> 0:03:02.240 car is at on this, but it has some of 0:03:02.280 --> 0:03:07.600 the components that are identified with this concept of autonomy. 0:03:07.639 --> 0:03:12.079 So most cars today are actually above level zero if 0:03:12.120 --> 0:03:15.519 we go by that definition there level one are higher. 0:03:16.400 --> 0:03:19.440 So level one autonomy would apply to cars where the 0:03:19.520 --> 0:03:22.399 driver still controls the vehicle. The vehicle is still under 0:03:22.480 --> 0:03:26.400 driver control, but the car has some driver assistance features 0:03:26.440 --> 0:03:29.639 like power steering or antilock brakes. The car might have 0:03:29.720 --> 0:03:33.480 what is called an advanced driver Assistance system or a 0:03:33.639 --> 0:03:37.240 d a S, and the word advanced makes it sound 0:03:37.280 --> 0:03:40.320 a bit fancier than it is at this particular level 0:03:40.360 --> 0:03:44.360 of autonomy. The car might have systems that help people steer, 0:03:44.880 --> 0:03:48.640 or it might have systems that help accelerate and or break, 0:03:49.120 --> 0:03:52.560 but the steering and accelerating or steering and breaking can't 0:03:52.600 --> 0:03:56.800 happen simultaneously. Either one or the other can be taken 0:03:56.840 --> 0:03:59.800 over by these systems, but not both. At the same time, 0:04:00.080 --> 0:04:02.320 not with level one. If we get up to level 0:04:02.320 --> 0:04:06.480 to autonomy, then we're talking about partial automation. The A 0:04:06.600 --> 0:04:08.760 d A s on these cars can do stuff like 0:04:08.840 --> 0:04:12.520 control steering and breaking, or steering and accelerating at the 0:04:12.560 --> 0:04:16.040 same time, at least under certain circumstances. But even in 0:04:16.080 --> 0:04:20.000 those cases, the car's driver still remains primarily in control 0:04:20.040 --> 0:04:22.880 of the vehicle. With this level of autonomy, a driver 0:04:23.000 --> 0:04:25.800 would still not remove their hands from the wheel, as 0:04:25.800 --> 0:04:28.320 the car would need the humans participation to you know, 0:04:28.880 --> 0:04:31.919 works safely. So with level to autonomy, you still have 0:04:32.000 --> 0:04:34.000 to have your attention on the road, you still have 0:04:34.000 --> 0:04:35.719 to have your hands on the wheel. It's just that 0:04:35.760 --> 0:04:40.720 the car can occasionally kick in and assist in various scenarios, 0:04:40.960 --> 0:04:45.720 typically in very restricted cases. Now at level three autonomy, 0:04:45.800 --> 0:04:49.960 we're getting up to conditional automation. These cars would still 0:04:49.960 --> 0:04:52.960 require a human driver, but there can be times when 0:04:52.960 --> 0:04:55.600 the car systems can operate the vehicle on its own 0:04:55.760 --> 0:04:59.799 and the driver is essentially a passenger. During those moments. 0:05:00.200 --> 0:05:02.360 The driver is still supposed to monitor the environment. They're 0:05:02.400 --> 0:05:04.320 still supposed to be prepared to take over the car 0:05:04.520 --> 0:05:07.400 should the vehicle indicate it needs to hand over control 0:05:07.480 --> 0:05:10.240 to the person behind the wheel. So ideally there would 0:05:10.279 --> 0:05:13.400 be a system where the car would identify a situation 0:05:13.440 --> 0:05:15.960 in which the driver needs to take over, and then 0:05:16.200 --> 0:05:20.000 well in advance of that situation becoming imminent, it would 0:05:20.040 --> 0:05:23.640 alert the driver to take over control of the car. Uh. 0:05:23.720 --> 0:05:27.279 This is trickier, right. This is harder to do than 0:05:27.360 --> 0:05:29.719 it than to say, because the car would have to 0:05:29.800 --> 0:05:32.839 know far enough in advance to be able to send 0:05:32.880 --> 0:05:34.960 that alert to the driver, and the driver would have 0:05:35.000 --> 0:05:37.400 to be able to respond to that. And while we 0:05:37.480 --> 0:05:42.840 feel like our response time is really fast, in computational terms, 0:05:43.440 --> 0:05:47.920 we are snails. We move super slow. So this is 0:05:47.920 --> 0:05:51.479 actually pretty tricky, especially if you're talking about a dynamic 0:05:51.880 --> 0:05:55.160 situation where things are changing very rapidly. At level three, 0:05:55.200 --> 0:05:58.800 autonomous cars are supposed to uh do this seamlessly, and 0:05:58.880 --> 0:06:01.800 as I said, that is a pretty tricky thing to 0:06:01.880 --> 0:06:06.679 do technically. Most vehicle systems were looking at now, especially 0:06:06.720 --> 0:06:10.359 the ones like Tesla's autopilot, falls somewhere in level three. 0:06:10.920 --> 0:06:13.760 Level four autonomy is at a point where a vehicle 0:06:13.800 --> 0:06:17.320 can automatically operate itself at least under certain conditions, but 0:06:17.400 --> 0:06:22.480 not necessarily all driving conditions, the vehicle would likely include 0:06:22.520 --> 0:06:25.680 the option for a human driver to take over operations, 0:06:25.960 --> 0:06:29.600 but under normal, you know, conditions that the car would 0:06:29.600 --> 0:06:32.840 pretty much drive itself. So with level for autonomy, you 0:06:32.880 --> 0:06:36.400 would have self driving cars that could act as a 0:06:36.400 --> 0:06:39.840 self driving car for most of the time, but also 0:06:39.960 --> 0:06:42.240 allow a human driver to take over if the human 0:06:42.320 --> 0:06:46.800 driver wanted to. UM. Level five autonomy is a fully 0:06:46.839 --> 0:06:51.200 autonomous car. The car can operate itself under all driving conditions, 0:06:51.240 --> 0:06:53.640 So any condition where a human would be driving a car, 0:06:54.200 --> 0:06:57.160 a level five autonomous car should be able to operate 0:06:57.160 --> 0:07:00.719 in that same situation. There may not be any steering 0:07:00.720 --> 0:07:04.200 wheel or any controls at all in a vehicle, meaning 0:07:04.200 --> 0:07:07.120 there's no option for a human driver to take over. 0:07:07.360 --> 0:07:10.760 Now that's not our prerequisite. You can have a level 0:07:10.880 --> 0:07:14.400 five fully autonomous car that would still have controls and 0:07:14.440 --> 0:07:18.760 still allow humans to take over manually if they if 0:07:18.800 --> 0:07:22.440 they chose to do so. It's just that it's an option, 0:07:22.560 --> 0:07:25.040 it's not. It's no longer a mandatory thing to have 0:07:25.200 --> 0:07:28.800 those human based controls with a level five autonomous car. 0:07:29.440 --> 0:07:32.880 UM So we don't have any of these yet, So 0:07:33.000 --> 0:07:36.760 really talking about this is purely in the hypothetical. Arguably, 0:07:36.840 --> 0:07:40.080 we have some that are in the level four range, 0:07:40.600 --> 0:07:44.120 but there will get to that they're under very strict parameters, 0:07:44.320 --> 0:07:47.960 all right, So most experts agree that the versions of 0:07:48.000 --> 0:07:50.960 autonomous cars we've seen so far are mainly in the 0:07:51.040 --> 0:07:54.960 level three and level four categories, uh, creeping more toward 0:07:55.160 --> 0:07:57.679 a firm level four. We're kind of in the early 0:07:57.800 --> 0:08:00.760 stages of that, and there's several tests programs that are 0:08:00.960 --> 0:08:04.760 operating almost as if we're at level five. But there's 0:08:04.920 --> 0:08:08.360 disagreement about whether or not technology is really sophisticate enough 0:08:08.400 --> 0:08:12.440 to warrant us calling any existing vehicle a level four 0:08:12.520 --> 0:08:16.600 or level five autonomous car. And so, while we have 0:08:16.960 --> 0:08:19.280 some examples of cars and I'll talk about a couple 0:08:19.320 --> 0:08:23.360 of them that lack control systems for human drivers, they 0:08:23.400 --> 0:08:26.800 are almost all prototypes and concept vehicles or in very 0:08:26.920 --> 0:08:31.240 limited testing situations. Uh. And so therefore they don't really 0:08:31.320 --> 0:08:34.240 rank as level five autonomous cars because while they lack 0:08:34.280 --> 0:08:38.640 the controls, they cannot operate in every situation an environment 0:08:38.640 --> 0:08:42.880 that humans drive in. So it's it's too early for 0:08:43.000 --> 0:08:45.120 us to talk about deploying cars that have no way 0:08:45.160 --> 0:08:48.960 to hand over control to human driver in all regions 0:08:49.000 --> 0:08:52.880 and in all you know, driving situations. Okay, so let's 0:08:52.880 --> 0:08:55.959 do a very quick rundown on the history of autonomous 0:08:55.960 --> 0:08:58.920 cars up to say, like or so, and to see 0:08:58.920 --> 0:09:01.880 why sub folks like like yours truly we're so bullish 0:09:02.040 --> 0:09:05.560 on the future of autonomous cars. So the history stretches 0:09:05.760 --> 0:09:08.600 back a good long ways, particularly if we're looking at 0:09:08.679 --> 0:09:12.360 stuff like power steering. But that's getting way too granular. 0:09:12.480 --> 0:09:14.760 I'm not going to do that. And the history is 0:09:14.800 --> 0:09:18.960 also really complex, and that involves lots of different disciplines 0:09:19.040 --> 0:09:23.800 converging into the autonomous car form factor. You have stuff 0:09:23.800 --> 0:09:29.480 like robotics, you know, sensored development, artificial intelligence, computational processing, power, 0:09:29.720 --> 0:09:34.040 range finding technology, lots of things that all have to 0:09:34.040 --> 0:09:37.520 come together. And to really dive into the complex history 0:09:37.559 --> 0:09:40.360 of all the technologies that are coming together to make 0:09:40.400 --> 0:09:44.520 autonomous cars possible would require a whole mini series of episodes. 0:09:44.559 --> 0:09:46.840 So we're not going to jump into all of that 0:09:46.920 --> 0:09:49.760 in this one. Instead, I want to focus on things 0:09:49.800 --> 0:09:52.920 like the DARPA challenges that were created in the mid 0:09:52.960 --> 0:09:55.959 two thousand's. The first one was in two thousand four 0:09:56.480 --> 0:09:59.280 and DARPA, as you'll recall, is the research and Development 0:09:59.360 --> 0:10:00.840 Arm of the depart I'm in the defense in the 0:10:00.880 --> 0:10:06.480 United States, so it's technically an organization that funds various 0:10:06.920 --> 0:10:10.719 other groups to do R and D in technologies that 0:10:10.840 --> 0:10:14.920 ultimately stand to benefit the defense of the United States. 0:10:15.320 --> 0:10:19.400 So while there are other uses for those technologies that 0:10:19.600 --> 0:10:24.640 don't directly relate to defense or military systems, that's the 0:10:24.640 --> 0:10:28.720 primary purpose for DARBA. So in two thousand four they 0:10:28.760 --> 0:10:32.200 created this this challenge. They called for teams to build 0:10:32.760 --> 0:10:37.040 or convert vehicles into autonomous cars that were capable of 0:10:37.120 --> 0:10:41.000 navigating a long distance desert course is more than a 0:10:41.080 --> 0:10:43.559 hundred miles long, and there it needs to be no 0:10:43.640 --> 0:10:46.560 human operators, so it could not be remotely controlled, nor 0:10:46.640 --> 0:10:48.920 would there be a driver in the vehicle. And the 0:10:49.000 --> 0:10:51.880 idea was design a car that would be capable of 0:10:52.160 --> 0:10:57.360 traveling a predesignated route from beginning to end. For that 0:10:57.400 --> 0:11:00.000 two thousand four challenge, no team was able to complete 0:11:00.000 --> 0:11:04.000 eat the challenge. Uh cars failed. Some of them went 0:11:04.040 --> 0:11:06.880 off road and got stuck, some of them just got 0:11:06.920 --> 0:11:11.560 confused and stopped. So no one completed it within the 0:11:11.600 --> 0:11:14.600 time frame that DARPA had set. But it's set the 0:11:14.640 --> 0:11:18.360 stage for subsequent competitions. In two thousand five, DARPA held 0:11:18.400 --> 0:11:21.480 another Grand Challenge again with the desert course. This one 0:11:21.559 --> 0:11:24.440 was a hundred thirty two miles long and this time 0:11:24.520 --> 0:11:27.520 five teams were able to complete the route and the 0:11:27.520 --> 0:11:30.959 winning team was from Stanford Race. Stanford University was the 0:11:31.000 --> 0:11:34.880 Stanford Racing team. They clocked the shortest time on the course. 0:11:35.280 --> 0:11:38.400 By shortest time, I'm still talking about a long time. 0:11:39.240 --> 0:11:42.760 It took them six hours fifty three minutes to make 0:11:42.800 --> 0:11:46.360 the one thirty two mile journey. That would mean that 0:11:46.400 --> 0:11:49.320 the average speed for the vehicle when taken across the 0:11:49.360 --> 0:11:52.520 whole course was somewhere around eighteen miles per hour or 0:11:52.520 --> 0:11:56.400 approximately twenty nine kilometers per hour, which isn't exactly tearing 0:11:56.480 --> 0:11:59.200 up the track, but it was still a very impressive achievement. 0:11:59.240 --> 0:12:01.640 I don't want to take away from what they achieved. 0:12:01.960 --> 0:12:06.520 It was incredible, especially for the time, but it's not 0:12:06.679 --> 0:12:08.920 the sort of speed you would look at and think, oh, well, 0:12:08.960 --> 0:12:12.439 this is the replacement for the modern car. The next 0:12:12.559 --> 0:12:15.480 challenge would happen in two thousand seven, and it switched 0:12:15.520 --> 0:12:18.600 things up by requiring teams to design a car capable 0:12:18.600 --> 0:12:23.040 of navigating through a simulated urban environment complete with traffic 0:12:23.160 --> 0:12:26.160 and traffic laws like you know, traffic lights and stop 0:12:26.240 --> 0:12:31.000 signs and simulated pedestrians. It wouldn't be enough to design 0:12:31.040 --> 0:12:33.520 a car that could detect a road and follow it, 0:12:33.840 --> 0:12:36.960 or even a car capable of managing stuff like how 0:12:37.000 --> 0:12:41.080 to how to send torque two different wheels. In order 0:12:41.080 --> 0:12:43.440 to get out of a tricky situation, the cars would 0:12:43.520 --> 0:12:47.920 need advanced collision detection and decision making capabilities. They have 0:12:47.960 --> 0:12:50.800 to obey traffic laws, they'd have to be able to 0:12:50.920 --> 0:12:53.960 adapt to potentially changing situations, the kind of stuff you 0:12:54.040 --> 0:12:58.280 might find if you're driving around a city. So in 0:12:58.320 --> 0:13:02.320 that case, six teams were able to finish the course, 0:13:02.640 --> 0:13:06.000 Stanford Racing would actually take second place that time. They 0:13:06.080 --> 0:13:08.880 clocked in at just under four and a half hours. 0:13:09.320 --> 0:13:12.360 First place went to a group called Tartan Racing from 0:13:12.440 --> 0:13:17.839 Carnegie Melon University and they finished in four hours ten minutes. Now, 0:13:17.840 --> 0:13:20.360 the purpose of these competitions wasn't just to find out 0:13:20.520 --> 0:13:24.280 which groups of smarty pants engineers were able to build 0:13:24.320 --> 0:13:27.200 the best car. It was an attempt to kick start 0:13:27.320 --> 0:13:32.160 serious development in the various fields related to making autonomous 0:13:32.200 --> 0:13:37.400 cars a possibility. Engineers worked on all sorts of different designs. 0:13:37.440 --> 0:13:41.719 Some incorporated lots of optical cameras. Some used lie dar, 0:13:42.000 --> 0:13:44.760 which is a type of laser based range finding technology 0:13:44.880 --> 0:13:47.719 similar to radar. So it works by zapping out a 0:13:47.800 --> 0:13:51.600 laser and then detecting any reflections coming back from that 0:13:51.720 --> 0:13:55.840 laser light. It uses an array of sensors looking for 0:13:55.880 --> 0:13:59.000 any evidence of that laser light coming back to the sensor, 0:13:59.520 --> 0:14:02.360 and then measures the time difference between when the laser 0:14:02.720 --> 0:14:07.439 went out and when it picked up the reflection, and then, 0:14:07.559 --> 0:14:10.320 working with some math, it can figure out how far 0:14:10.400 --> 0:14:13.960 away an obstacle is from the vehicle. Not only that, 0:14:14.000 --> 0:14:16.240 it can also figure out whether or not that obstacle 0:14:16.679 --> 0:14:19.080 is moving, or if it's stationary, or if it's moving 0:14:19.120 --> 0:14:21.640 away from or toward the vehicle. It can figure out 0:14:21.640 --> 0:14:25.080 all that. Uh, and I've talked about that in past episode, 0:14:25.080 --> 0:14:27.320 so I won't get into the whull technical details here, 0:14:27.960 --> 0:14:30.920 but it was one of those key components that's used 0:14:30.960 --> 0:14:35.360 in some, but not all vehicles that are following under 0:14:35.360 --> 0:14:38.960 this autonomous card development. It's interesting because there are lots 0:14:38.960 --> 0:14:41.880 of different companies that are working on autonomous cars. They 0:14:41.880 --> 0:14:45.800 are not all relying on exactly the same technologies to 0:14:45.880 --> 0:14:48.680 achieve that goal. Some of them are much more heavily 0:14:48.720 --> 0:14:52.040 focused on optical cameras. Some of them are more focused 0:14:52.040 --> 0:14:55.160 on things like lidar and other sensors. Some of them 0:14:55.200 --> 0:14:59.720 involve a whole, you know, slew of different technologies that 0:14:59.760 --> 0:15:02.800 are meant to be both uh, you know, primary systems 0:15:02.800 --> 0:15:06.120 and redundant systems. So it's really interesting and it was. 0:15:06.360 --> 0:15:08.960 It was really impressive to see these teams complete the 0:15:09.040 --> 0:15:12.080 Urban Challenge, but again, it didn't immediately make everyone think 0:15:12.160 --> 0:15:17.480 driverless cars would be available right away. The challenges, while impressive, 0:15:18.240 --> 0:15:21.680 didn't compare to what the average human driver deals with 0:15:21.800 --> 0:15:24.600 on a on a regular day. The competition times were 0:15:24.600 --> 0:15:28.520 pretty long. The average speeds were all below fifteen miles 0:15:28.520 --> 0:15:31.560 per hour, so they're all below twenty four kilometers per hour. 0:15:31.880 --> 0:15:34.200 At that speed, it was clear that these vehicles were 0:15:34.240 --> 0:15:39.200 just the earliest incarnations of technologies that would power autonomous 0:15:39.240 --> 0:15:41.720 cars in the future. So they were airing on the 0:15:41.720 --> 0:15:45.160 side of caution, which frankly, you want in the first place. 0:15:45.200 --> 0:15:47.040 You don't want to see a lot of people say 0:15:47.600 --> 0:15:50.840 let's take some chances, when when you're talking about vehicles, 0:15:50.880 --> 0:15:55.360 I mean, they're human lives at stake. Meanwhile, another narrative 0:15:55.520 --> 0:15:59.400 drives home pun intended why a lot of folks got 0:15:59.440 --> 0:16:03.560 really hyped up about autonomous cars. It's also a sobering 0:16:04.400 --> 0:16:07.400 line of thought. And I'm talking, of course, about the 0:16:07.480 --> 0:16:11.280 frequency of fatal car accidents and how many of them 0:16:11.520 --> 0:16:15.680 can be traced back to human error. Now, getting global 0:16:16.040 --> 0:16:18.680 statistics is pretty tough on this, so I'm going to 0:16:18.800 --> 0:16:21.160 focus on the United States because we have a lot 0:16:21.200 --> 0:16:25.000 of organizations in the US that track these kinds of numbers, 0:16:25.080 --> 0:16:27.640 and you can kind of get an idea of how 0:16:27.680 --> 0:16:31.800 big the problem is. So in tween, the National Safety 0:16:31.880 --> 0:16:36.040 Council released a report that stated and estimated forty thousand 0:16:36.040 --> 0:16:40.040 people had died in car accidents in the United States. Uh, 0:16:40.160 --> 0:16:43.600 that actually amounted to a decline of one percent from 0:16:43.600 --> 0:16:48.840 two thousand seventeen. That was when forty one people died. 0:16:49.400 --> 0:16:51.960 Another four and a half million people on top of 0:16:52.000 --> 0:16:57.800 that had become seriously injured in car crashes in Meanwhile, 0:16:58.160 --> 0:17:01.600 the National Highway Traffics say D administration in the United 0:17:01.640 --> 0:17:07.359 States said that of serious car crashes result due to 0:17:07.480 --> 0:17:12.119 human error or dangerous choices. So, in other words, mechanical 0:17:12.200 --> 0:17:16.679 failures only contribute a very small percentage to the overall numbers. 0:17:16.680 --> 0:17:20.040 When it comes to serious car accidents, most serious car 0:17:20.080 --> 0:17:25.199 accidents aren't caused by a tire blowout or you know, 0:17:25.640 --> 0:17:29.160 a car failing in some way. They're caused by humans 0:17:29.200 --> 0:17:34.720 doing something wrong, whether it's totally by accident or someone 0:17:34.800 --> 0:17:37.160 just makes a really bad decision, like they think, oh, 0:17:37.200 --> 0:17:40.200 sure there's no there's no dashed line here, but I'm 0:17:40.200 --> 0:17:42.320 gonna go ahead and try and pass this person on 0:17:42.359 --> 0:17:46.399 this windy rural road because I bet nobody's coming the 0:17:46.440 --> 0:17:49.680 other way. That's what we would call a bad decision, 0:17:50.320 --> 0:17:55.040 So says the tech optimist. If you could create autonomous 0:17:55.080 --> 0:17:59.800 cars that operates safely, you could eliminate the vast majority 0:18:00.040 --> 0:18:03.960 of car crashes and thus fatalities on the road. You 0:18:04.040 --> 0:18:07.560 just remove the human error element, and suddenly you're talking 0:18:07.720 --> 0:18:14.120 about a staggering result, and that is an incredibly powerful motivator. 0:18:14.600 --> 0:18:19.960 Tens of thousands of people wouldn't die each year from 0:18:20.000 --> 0:18:23.800 these car accidents. Millions more would never be injured or 0:18:23.880 --> 0:18:27.120 affected by the tragic loss of a loved one from 0:18:27.119 --> 0:18:31.080 an accident. Then you start moving outward, you go out 0:18:31.119 --> 0:18:33.439 another circle. You think of this as a ripple effect 0:18:33.440 --> 0:18:37.320 and you think, imagine all the contributions those people might 0:18:37.440 --> 0:18:40.040 make in the future that they'll get a chance to 0:18:40.119 --> 0:18:44.200 make because they wouldn't have had this terrible car crash. 0:18:44.240 --> 0:18:47.520 These are things we never would see come to fruition 0:18:47.840 --> 0:18:50.119 if they were to get in a fatal car crash, 0:18:50.440 --> 0:18:53.639 and it becomes this butterfly effect issue. And of course, 0:18:54.160 --> 0:18:57.800 we want to make the roads safer for everyone. Now, 0:18:58.080 --> 0:19:00.879 I'm sure all of you have already hit upon the 0:19:00.960 --> 0:19:04.960 major issue here. The whole concept of people being safer 0:19:05.000 --> 0:19:10.359 in autonomous cars is contingent upon those autonomous cars performing 0:19:10.480 --> 0:19:14.119 better than humans already do and in every type of 0:19:14.160 --> 0:19:17.760 situation in which humans find themselves driving in. If we 0:19:17.840 --> 0:19:22.200 can't get that right, then we haven't made things safer 0:19:22.359 --> 0:19:25.439 at all. All we would have done is shifted the 0:19:25.600 --> 0:19:29.760 cause of the accidents from human error to machine error 0:19:29.920 --> 0:19:33.919 or computer error. So we must be absolutely certain that 0:19:33.960 --> 0:19:37.800 the vehicles we make meet a very high standard if 0:19:37.920 --> 0:19:42.600 our goal is to reduce car accidents. So we have 0:19:42.680 --> 0:19:46.400 to prove that these machines operate better than people do 0:19:46.880 --> 0:19:50.120 in all the different situations people find themselves driving in 0:19:50.400 --> 0:19:53.280 before we can make any sort of declarative statement of 0:19:53.359 --> 0:19:56.440 this is the best way forward. Now, when we come back, 0:19:56.720 --> 0:20:00.760 I'll talk about why this gets super tricky, and talk 0:20:00.960 --> 0:20:05.080 about thought experiments and things, and and also some real 0:20:05.119 --> 0:20:08.880 world scenarios that kind of illustrate why this is harder 0:20:08.920 --> 0:20:12.280 than what it sounds. But first, let's take a quick break. 0:20:19.760 --> 0:20:23.240 Before the break, I positive that a future with autonomous 0:20:23.320 --> 0:20:27.280 cars that all but eliminate fatal car crashes hinge upon 0:20:28.040 --> 0:20:31.480 building driverless vehicles that are much better at driving cars 0:20:31.520 --> 0:20:34.520 than humans are in all situations. Now, we could get 0:20:34.760 --> 0:20:38.879 a bit more lucy goosey here, but doing so brings 0:20:38.960 --> 0:20:42.320 up some tough ethical issues. So, for example, what if 0:20:42.359 --> 0:20:46.920 we knew that machines were better. Right, autonomous cars are 0:20:47.040 --> 0:20:51.040 better than human drivers, but they are by no means perfect. 0:20:51.680 --> 0:20:54.359 So what if we could be certain that autonomous cars, 0:20:54.440 --> 0:20:59.400 if widely adopted, would reduce those fatalities by half, for example, 0:20:59.680 --> 0:21:02.320 but they would still be at fault in the case 0:21:02.400 --> 0:21:06.120 of the other half of fatal car accidents. So let's 0:21:06.119 --> 0:21:09.640 say it's you know, I don't know, and we have 0:21:10.680 --> 0:21:16.200 level four autonomous cars that are pretty reliably level four 0:21:16.760 --> 0:21:20.200 and they are better as a whole than human drivers are. 0:21:20.480 --> 0:21:24.280 So we've seen a vast reduction in vehicles operated by humans. 0:21:24.480 --> 0:21:27.159 And let's even assume that most cars are now controlled 0:21:27.160 --> 0:21:31.160 by computers, but let's also assume they're not perfect now. 0:21:31.400 --> 0:21:35.080 Using the numbers from if humans were still in control, 0:21:35.280 --> 0:21:38.640 we would expect to see another forty thousand fatalities due 0:21:38.680 --> 0:21:41.239 to human error. And I'm just using that number as 0:21:41.240 --> 0:21:43.600 an example. I realized that in reality we'd be talking 0:21:43.600 --> 0:21:48.159 about nine of forty thousand. But now that cars are 0:21:48.200 --> 0:21:52.320 in control, it means half of those accidents are totally prevented. 0:21:52.920 --> 0:21:57.440 But we still see fatal accidents that claim twenty thousand lives. 0:21:57.480 --> 0:22:00.000 On the one hand, we could look at that scenari 0:22:00.000 --> 0:22:02.920 are you and say, based upon what we know from 0:22:02.960 --> 0:22:06.360 past experience, we would have seen many more people die 0:22:06.520 --> 0:22:10.359 in accidents if humans were actually still operating cars. But 0:22:10.440 --> 0:22:12.760 on the other hand, that's all hypothetical, right, I mean, 0:22:13.200 --> 0:22:17.160 we can only know anything based on what actually happened, 0:22:17.320 --> 0:22:20.160 not on what might have happened if things had gone 0:22:20.200 --> 0:22:24.160 a different way. We can't be sure. But more than that, though, 0:22:24.200 --> 0:22:28.480 we're still talking about twenty thousand people losing their lives 0:22:28.520 --> 0:22:32.199 and all the ripple effects that that makes throughout society, 0:22:32.280 --> 0:22:35.840 and moreover, we have machines that are at the fault 0:22:36.080 --> 0:22:38.760 for those twenty thousand lives being lost, and the idea 0:22:38.840 --> 0:22:42.600 that people have built machines that, through a failure of 0:22:42.640 --> 0:22:46.760 some sort or another, resulted in deaths is a very 0:22:46.840 --> 0:22:51.159 difficult proposition to accept. Also, it's just a key to 0:22:51.280 --> 0:22:54.400 think of in terms like that. I mean, clearly, one 0:22:54.560 --> 0:22:58.000 death is too many. We don't want to see anyone 0:22:58.160 --> 0:23:01.680 die in a car accident. Having a discussion in which 0:23:01.720 --> 0:23:05.000 you compare a fewer number of deaths and referring to 0:23:05.040 --> 0:23:08.840 it as quote unquote better is something that's pretty hard 0:23:08.880 --> 0:23:11.600 for us to process. It's easier to do it the 0:23:11.640 --> 0:23:14.960 other way, right, I mean, it's obvious that forty thousand 0:23:14.960 --> 0:23:18.960 people dying is worse than twenty thou people dying, But 0:23:19.040 --> 0:23:22.120 it's hard to view it the other way because anyone 0:23:22.320 --> 0:23:25.879 dying at all is awful. Now. Part of this also 0:23:25.920 --> 0:23:29.640 really boils down to a fear of handing over control 0:23:29.720 --> 0:23:32.679 to a machine. I know a lot of people bulk 0:23:33.000 --> 0:23:35.760 at that idea. They don't like the idea of not 0:23:35.920 --> 0:23:40.520 being the actual entity making decisions behind the wheel. Confronting 0:23:40.560 --> 0:23:44.440 them with statistics showing how human error leads to catastrophe, 0:23:44.880 --> 0:23:47.840 doesn't tend to sway them. I mean a lot of 0:23:47.840 --> 0:23:52.560 people think, well, yeah, that's other people. I am not 0:23:52.720 --> 0:23:55.760 that person. Also, to be fair, we don't have the 0:23:55.840 --> 0:23:59.600 evidence to show that computers would necessarily be better, so 0:23:59.760 --> 0:24:03.240 they're something to that right now. Okay, let's let's get 0:24:03.240 --> 0:24:05.399 back to where we were in our history. The Grand 0:24:05.440 --> 0:24:08.679 Challenges helped set the stage for the next phase of development, 0:24:09.040 --> 0:24:13.200 which was mostly the realm of startups and some big companies, 0:24:13.280 --> 0:24:17.680 namely Google, would hire participants from the Grand and Urban 0:24:17.760 --> 0:24:21.000 Challenges to come and work in new divisions dedicated to 0:24:21.040 --> 0:24:25.520 creating driverless cars. The early pioneering work was now shifting 0:24:25.760 --> 0:24:30.399 pun intended into a phase of rapid iteration, as engineers 0:24:30.400 --> 0:24:34.960 and computer scientists and mechanics began to refine technologies to 0:24:34.960 --> 0:24:38.840 help make them better. So going from the first sort 0:24:38.880 --> 0:24:42.240