WEBVTT - TechStuff Classic: The Great Google Car Crash of 2016 0:00:04.360 --> 0:00:12.319 Welcome to tech Stuff, a production from iHeartRadio. Hey there, 0:00:12.400 --> 0:00:15.840 and welcome to tech Stuff. I'm your host Jonathan Strickland. 0:00:15.880 --> 0:00:19.800 I'm an executive producer with iHeartRadio and how the Tech Area. 0:00:20.160 --> 0:00:23.320 It's time for another classic episode of tech Stuff. This 0:00:23.360 --> 0:00:28.480 one is titled the Great Google Car Crash of twenty sixteen. 0:00:29.080 --> 0:00:32.360 Scott Benjamin joined the show to talk about this. This 0:00:32.520 --> 0:00:35.040 was one of those news stories that was a pretty 0:00:35.040 --> 0:00:38.800 big deal in twenty sixteen. The episode originally published on 0:00:38.880 --> 0:00:45.199 April twenty seventh, twenty sixteen. And you know autonomous cars, 0:00:46.200 --> 0:00:50.040 well really that we don't have a truly autonomous car yet, 0:00:50.920 --> 0:00:55.800 but it was kind of a new concept, like new 0:00:55.800 --> 0:00:58.320 in the sense that we hadn't really seen too many 0:00:58.400 --> 0:01:02.080 of them out on actual street back in twenty sixteen. 0:01:02.280 --> 0:01:06.360 So Google getting into a little car accident was big news. Now. Obviously, 0:01:06.440 --> 0:01:09.360 in the years since, we have had a lot more 0:01:09.440 --> 0:01:15.839 accidents involving autonomous vehicles or semi autonomous vehicles, so things 0:01:15.840 --> 0:01:22.560 have definitely continued to be complicated and sometimes tragic, but 0:01:22.720 --> 0:01:28.560 this one was more of an interesting, lighthearted look at 0:01:28.840 --> 0:01:32.480 why an autonomous car might get into an accident. So 0:01:32.560 --> 0:01:36.720 sit back and enjoy this episode from twenty sixteen. We're 0:01:36.720 --> 0:01:40.760 talking today about a peculiar event, something that happened in 0:01:40.800 --> 0:01:43.880 February twenty sixteen, and it's just taken me this long 0:01:43.920 --> 0:01:46.560 to finally get around and addressing it. You might have 0:01:46.640 --> 0:01:50.240 heard in previous episodes of Tech Stuff about how I 0:01:50.280 --> 0:01:54.600 would champion the fact that Google with their self driving 0:01:54.640 --> 0:02:01.120 cars had had enormous success, flawless. You might argue, success 0:02:01.120 --> 0:02:03.960 something like one point four to five million miles driven 0:02:04.480 --> 0:02:07.720 without a single accident caused by the autonomous system. That 0:02:07.960 --> 0:02:11.560 there had been about fourteen or so accidents, but all 0:02:11.600 --> 0:02:15.400 of those were either the fault of a person manually 0:02:15.480 --> 0:02:20.160 driving the car in manual mode or another driver colliding 0:02:20.240 --> 0:02:23.840 with the autonomous car, but never the fallow the autonomous 0:02:23.840 --> 0:02:29.639 car itself. It was a perfect system until February twenty sixteen. Yeah, 0:02:29.639 --> 0:02:31.720 and that is the day of what I'll caohol, And 0:02:31.720 --> 0:02:34.320 I really don't mean to over dramaticize this at all, 0:02:34.639 --> 0:02:37.360 So maybe I'm titling this episode. I think maybe sure 0:02:37.880 --> 0:02:41.079 we should call this the Saint Valentine's Day Google self 0:02:41.200 --> 0:02:44.639 driving Car Massacre. Oh, that's an excellent title, and I'm 0:02:44.680 --> 0:02:47.280 all for it. Definitely not overly dramatic in any way. Well, 0:02:47.280 --> 0:02:49.639 I mean it's it's also funny because we're recording this 0:02:50.440 --> 0:02:53.840 the week after I've gotten back from south By Southwest 0:02:53.919 --> 0:02:57.360 and this was a topic that was discussed heavily at 0:02:57.360 --> 0:03:03.240 south By Southwest because till this incident, it was a 0:03:03.480 --> 0:03:07.360 very easy sell to say autonomous cars are the way 0:03:07.400 --> 0:03:10.600 to go. And then this little accident happened, and and 0:03:10.680 --> 0:03:13.480 it it wasn't terrible. We'll get into the details of 0:03:13.520 --> 0:03:16.880 the accident, but this little accident happened and suddenly it 0:03:17.040 --> 0:03:22.160 sounded like Google's autonomous car had caused an enormous pile 0:03:22.280 --> 0:03:25.120 up on the highway. Everyone was much more cautious. So 0:03:25.200 --> 0:03:27.440 you're maybe not buying my alternate title. Then, is that 0:03:27.480 --> 0:03:30.560 what you're saying? What's your alternate title? No, no, that 0:03:30.639 --> 0:03:32.880 I'm totally no. I think it is a maskacer, but 0:03:32.919 --> 0:03:35.040 I think it's a massacer in the sense of the 0:03:35.080 --> 0:03:40.440 public perception of autonomous cars. I see, Okay, yeah, I'm 0:03:40.480 --> 0:03:42.880 thinking of from a pr st, I could I could 0:03:42.920 --> 0:03:46.880 take this opportunity to gloat and say, ah, they're not 0:03:46.920 --> 0:03:50.160 as infallible as you thought they're They're not perfect, but 0:03:50.160 --> 0:03:52.440 but I'm going to take a different stance here in 0:03:52.560 --> 0:03:54.800 this in this podcast, And and I think that as 0:03:54.880 --> 0:03:57.480 we as we talk through this, we're going to realize 0:03:57.520 --> 0:04:00.640 that they've been held to a much higher standard than 0:04:00.680 --> 0:04:04.240 they probably need to be right. And I know that's 0:04:04.280 --> 0:04:06.200 that's tough to to take, you know, when you when 0:04:06.200 --> 0:04:08.200 you just hear it that way, But listen to our 0:04:08.280 --> 0:04:11.400 argument back and forth about this, and and understand that 0:04:12.040 --> 0:04:14.960 they're being held to perfection when they probably shouldn't be. 0:04:15.320 --> 0:04:18.400 When humans, I mean, we're not perfect, of course, right, 0:04:18.560 --> 0:04:24.960 there's a vast m a chasm between what the standards 0:04:25.000 --> 0:04:27.440 that they're held to versus the standard that human test 0:04:27.480 --> 0:04:29.480 drivers are held too. Sure, yeah, if you look at 0:04:29.560 --> 0:04:32.440 if you look at the standard driving test that you 0:04:32.520 --> 0:04:35.799 have to pass before you get a license, I would 0:04:35.880 --> 0:04:40.159 argue most autonomous cars could likely pass such a test 0:04:41.120 --> 0:04:45.560 as close to flawlessly as you can get. But you 0:04:45.600 --> 0:04:48.279 don't have to be flawless when you take a driver's test. 0:04:48.360 --> 0:04:52.760 There's room for you to not completely do something perfectly, 0:04:52.760 --> 0:04:55.840 Like if your parallel parking isn't exactly right, you're going 0:04:55.880 --> 0:04:58.600 to get points deducted from your total, but you may 0:04:58.600 --> 0:05:01.880 still be you know, high enough, score high enough so 0:05:01.880 --> 0:05:04.040 that you could pass the full driver's test. Exactly. You 0:05:04.120 --> 0:05:06.080 knock over cone. It's not really a big deal. But 0:05:06.200 --> 0:05:09.320 an autonomous car knocks over a cone, everybody points at 0:05:09.360 --> 0:05:11.040 it and says, oh, we'll get that thing. It's it's 0:05:11.080 --> 0:05:14.880 pilot junk. Yeah, exactly. It's interesting that you point that 0:05:14.920 --> 0:05:18.160 out too, because that ties into a different discussion. I 0:05:18.200 --> 0:05:21.360 saw it south by Southwest that wasn't specifically about autonomous cars. 0:05:21.360 --> 0:05:23.360 It was about robots. So this is a little bit 0:05:23.400 --> 0:05:26.279 of a tangent, but it goes to illustrate the point 0:05:26.320 --> 0:05:30.159 you just made. I love tangents, So this h pammel 0:05:30.160 --> 0:05:33.360 about robots. There was a woman, Leila Takayama, who used 0:05:33.360 --> 0:05:36.080 to work for Google X, but not in the autonomous 0:05:36.080 --> 0:05:41.480 car division. She talked about how she ran an experiment. 0:05:42.440 --> 0:05:45.479 She got a guy from Pixar to do a series 0:05:45.520 --> 0:05:50.360 of very simple animations to show people the interactions between 0:05:50.400 --> 0:05:53.760 a robot and a person and then to judge which 0:05:54.000 --> 0:05:59.360 robot is considered intelligent versus not intelligent. And the whole 0:05:59.400 --> 0:06:02.359 point of this was to show the differences between succeeding 0:06:02.360 --> 0:06:07.200 and failing, but giving no indication that the robot understands 0:06:07.200 --> 0:06:11.279 it succeeded or failed, or building in expressions for the 0:06:11.360 --> 0:06:16.279 robot to follow a success or failure to indicate it 0:06:16.480 --> 0:06:21.039 quote unquote understands what happened. And it was fascinating because 0:06:21.320 --> 0:06:25.760 they showed a very simple experiment with a robot trying 0:06:25.760 --> 0:06:27.760 to open the door. And again, this is an animation, 0:06:28.640 --> 0:06:32.440 so there was different scenarios. There's one where the robot 0:06:32.800 --> 0:06:35.680 opens the door and the door opens and then the 0:06:35.760 --> 0:06:38.080 robot just sits there and it's done because it's done 0:06:38.080 --> 0:06:40.560 what it was supposed to do. There's one where the 0:06:40.640 --> 0:06:45.480 robot opens the door and then kind of perks up, like, oh, 0:06:45.040 --> 0:06:47.960 I did what I wanted to do. There was one 0:06:47.960 --> 0:06:50.600 where the robot fails to open the door and then 0:06:50.720 --> 0:06:53.480 does nothing, and then there was one where the robot 0:06:53.560 --> 0:06:55.760 fails to open the door and then slumps down a 0:06:55.760 --> 0:06:57.760 little bit as if to say, oh, I'm disappointed I 0:06:57.800 --> 0:07:01.279 didn't succeed. They then asked able to judge which robots 0:07:01.320 --> 0:07:03.839 they thought were the most intelligent, and everyone said the 0:07:03.960 --> 0:07:08.240 robot that failed but showed disappointment was more intelligent than 0:07:08.279 --> 0:07:11.280 the robot that succeeded but didn't show any expression at all. 0:07:11.520 --> 0:07:14.120 No kidding, And when you think about that again. It's 0:07:14.160 --> 0:07:18.360 holding robots to a standard that doesn't necessarily apply to 0:07:18.400 --> 0:07:22.120 them because of the human element this human robot interaction. 0:07:22.920 --> 0:07:27.920 We're holding autonomous cars to a similar standard that perhaps 0:07:28.160 --> 0:07:31.360 is not fair. We're holding robots to a standard that's 0:07:31.400 --> 0:07:34.080 not fair. But that means that people who are designing 0:07:34.080 --> 0:07:37.200 autonomous cars and people who are designing robots have to 0:07:37.240 --> 0:07:41.280 take that into consideration because that's the way humans are. Well, 0:07:41.320 --> 0:07:45.800 this is interesting because you're you're mentioning specifically imitating human behavior. Yes, 0:07:45.840 --> 0:07:48.840 and this comes up in an article that I read 0:07:48.920 --> 0:07:51.800 in let's see, it was in the Verge. Yes, and 0:07:51.960 --> 0:07:53.640 the Verge article you might have read the same thing. 0:07:53.680 --> 0:07:56.400 An excellent article. Yeah, it really is. And a person 0:07:56.440 --> 0:07:59.160 by the name of Jennifer Harun. She's the head of 0:07:59.200 --> 0:08:02.160 business operations for Google self driving project. And by the way, 0:08:02.360 --> 0:08:04.000 let's come back to the details of the accident and 0:08:04.040 --> 0:08:06.360 just a moment. Sure, we'll describe what happened, but she 0:08:06.400 --> 0:08:08.680 says that, well, you know what, maybe I need to 0:08:08.720 --> 0:08:10.480 back this up just a second here. How about this, 0:08:10.920 --> 0:08:12.880 Let's describe the accident and then we'll talk about what 0:08:12.920 --> 0:08:16.400 she said, because it plays perfectly into it. So in 0:08:16.560 --> 0:08:19.160 helping understand what happened. So here's here's how you can 0:08:19.200 --> 0:08:23.600 imagine it. All right, You've got an intersection in Mountain View, California, 0:08:23.600 --> 0:08:26.880 which is where Google's headquarters is located, and you've got 0:08:26.920 --> 0:08:29.600 the Google self driving car. Correct me if I get 0:08:29.600 --> 0:08:31.440 any of this wrong, Scott, I'm going from memory. That's 0:08:31.440 --> 0:08:33.760 all right, I'm doing mostly the same. It's so the 0:08:33.760 --> 0:08:35.719 Google self driving cars in the right lane and at 0:08:35.720 --> 0:08:39.280 once it's planning on making a right turn at this intersection. Yes, Now, 0:08:39.520 --> 0:08:42.280 at the corner of the intersection, there were some sandbags 0:08:42.360 --> 0:08:45.400 that were a partial obstruction of the lane. Yeah, I 0:08:45.400 --> 0:08:48.920 think you're blocking a sewer entry a great maybe or something. Right, 0:08:49.000 --> 0:08:53.640 So the Google car detected that there was an obstruction, 0:08:54.080 --> 0:08:56.480 and so it had to plan an alternate way to 0:08:56.559 --> 0:08:58.840 make its right turn. It still wanted to follow the 0:08:58.920 --> 0:09:02.800 route that it had planned, so the change would have 0:09:02.840 --> 0:09:05.559 been for it to kind of edge into the next lane, 0:09:05.600 --> 0:09:07.680 over the next lane to the left before making a 0:09:07.760 --> 0:09:13.800 right turn. Behind the Google car, approaching at a blistering 0:09:13.840 --> 0:09:17.319 speed of fifteen miles per hour was a bus, and 0:09:17.400 --> 0:09:20.800 so the Google car recognized there was a bus coming. 0:09:21.640 --> 0:09:23.319 He was moving at a very slow speed at two 0:09:23.360 --> 0:09:27.840 miles per hour. The Google Car said, well, based upon 0:09:27.920 --> 0:09:30.760 my programming, what I should expect happen is that the 0:09:30.800 --> 0:09:34.120 bus will slow down, allow me to move through. I'll 0:09:34.120 --> 0:09:37.720 clear the intersection, the bus will continue. What actually happened 0:09:37.840 --> 0:09:40.600 was the Google Car made the move into the lane, 0:09:40.960 --> 0:09:44.000 the bus continued, and there was a low speed collision 0:09:44.200 --> 0:09:46.560 and there were no injuries. No one was hurt. There was, 0:09:46.679 --> 0:09:49.360 in fact a driver behind the wheel of the Google Car. 0:09:49.400 --> 0:09:52.760 It's just the driver wasn't in control. The autonomous system 0:09:52.800 --> 0:09:55.800 was in control. And some people might say, well, you know, 0:09:55.840 --> 0:09:58.400 the bus driver just didn't let the car in. But 0:09:58.520 --> 0:10:01.760 Google actually said this is important. Google came out and said, 0:10:02.480 --> 0:10:06.720 we accept responsibility for this. This is something that it's 0:10:06.800 --> 0:10:09.520 it's valuable that this information has come to light because 0:10:09.520 --> 0:10:12.480 it means that we need to revisit this particular part 0:10:12.800 --> 0:10:16.920 of the autonomous car programming. Now, I thought that was 0:10:17.000 --> 0:10:19.520 really interesting. First of all, I have never heard of 0:10:19.520 --> 0:10:23.880 a company accepting responsibility for something so fast in my life. Well, 0:10:24.320 --> 0:10:25.920 they did, and they didn't. I mean, there's a couple 0:10:25.960 --> 0:10:28.200 of versions of this. Now you got the details of 0:10:28.240 --> 0:10:30.280 the accident correct, Although I did hear and this this 0:10:30.360 --> 0:10:32.960 is a bit confusing. I didn't hear that the lanes 0:10:32.960 --> 0:10:35.800 in this particular part of town are extremely wide, and 0:10:35.840 --> 0:10:39.120 so what happened was the car kind of edged itself 0:10:39.120 --> 0:10:43.800 over toward the curb. So it was I guess, mimicking 0:10:43.920 --> 0:10:45.720 human behavior again. And I'll get to that in just 0:10:45.760 --> 0:10:48.400 a minute. But um, you know, once this accident happened 0:10:48.520 --> 0:10:50.920 and they said, you know, we do need to investigate this, 0:10:51.280 --> 0:10:53.480 they did that to the tune of about thirty five 0:10:53.679 --> 0:10:57.240 hundred new tests that they've now implemented since this accident 0:10:57.440 --> 0:10:59.320 that said, we're going to watch for this. You know, 0:10:59.520 --> 0:11:03.040 we need to understand a little more deeply that some 0:11:03.080 --> 0:11:06.000 of these larger vehicles may have a different, more difficult 0:11:06.000 --> 0:11:09.360 times stopping in traffic than a smaller vehicle will. And 0:11:09.360 --> 0:11:11.600 and that's the reason why some of these bigger vehicles 0:11:11.640 --> 0:11:14.080 like to continue on their path and think, well, maybe 0:11:14.080 --> 0:11:17.040 someone behind me will let them in. Google did say 0:11:17.400 --> 0:11:20.840 we were relying on an element of human kindness to 0:11:20.920 --> 0:11:23.120 let us into that lane, and that's normally what happens. 0:11:23.160 --> 0:11:24.880 It really does. Usually there's a back and forth or 0:11:24.920 --> 0:11:27.079 you know, maybe there's always going to be that off 0:11:27.120 --> 0:11:29.280 handed time where you know, someone does cut through and 0:11:29.600 --> 0:11:31.360 they're just like they're like, no, I need to get 0:11:31.360 --> 0:11:33.560 through that intersection in this light cycle, and no one 0:11:33.679 --> 0:11:35.160 is going to stop me. Yeah, I mean I'm gonna 0:11:35.160 --> 0:11:37.200 be ten feet ahead of you when all this happens, right, 0:11:37.320 --> 0:11:39.480 That's where I you know, that's that's my goal. But 0:11:39.679 --> 0:11:42.800 usually what happens is it's an alternating pattern and they 0:11:42.800 --> 0:11:45.120 expected that to happen, and it didn't happen in this case. 0:11:45.120 --> 0:11:48.319 And this is where it plays right into what Jennifer 0:11:48.360 --> 0:11:50.640 had mentioned. Now, Jennifer Harun, who is the head of 0:11:50.679 --> 0:11:55.000 business operations for Google self driving project, explained at and 0:11:55.000 --> 0:11:57.000 I think she was at the south By Southwest conference 0:11:57.080 --> 0:12:00.240 as well, and she said that the Lexus, it's a 0:12:00.320 --> 0:12:02.560 Lexus vehicle that was outfitted with this gear so that 0:12:02.640 --> 0:12:05.760 it struck the bus in part because it was imitating 0:12:05.840 --> 0:12:08.800 human behavior. And that's I found that interesting that she 0:12:09.040 --> 0:12:14.000 kind of is deferring the fault here and to say, well, 0:12:14.040 --> 0:12:16.040 we were just imitating what we see on the streets 0:12:16.040 --> 0:12:18.280 and that's that's kind of what happened. So well, it's 0:12:18.280 --> 0:12:21.000 a it's a double deferral in a way, right, because 0:12:21.040 --> 0:12:22.960 first they say we were counting on an element of 0:12:23.000 --> 0:12:25.959 human kindness, which is already kind of a deferral in itself. Sure, yeah, 0:12:26.040 --> 0:12:29.280 you're you're essentially saying, well, we were we were thinking 0:12:29.320 --> 0:12:31.360 that the bus driver would be a decent human being 0:12:31.360 --> 0:12:35.320 and not a Lexus hating scumbag. That's I'm paraphrasing what 0:12:35.360 --> 0:12:38.040 they said. Obviously I'm taking a little liberty. And then 0:12:38.040 --> 0:12:41.199 in this one, you're saying, she's also saying, well, we 0:12:41.320 --> 0:12:44.160 designed the car to behave the way we see actual 0:12:44.240 --> 0:12:47.679 cars behaving on the road. So in both cases, you're 0:12:47.720 --> 0:12:50.560 almost it's a little bit of backing away from taking 0:12:50.600 --> 0:12:53.960 full responsibility exactly. And this imitating human behavior that she's 0:12:54.000 --> 0:12:57.079 talking about was that they had recently taught the vehicles 0:12:57.080 --> 0:12:59.600 to hug the right hand side of that lane when 0:12:59.600 --> 0:13:02.280 they're making that right hand turn, and that that's when 0:13:02.320 --> 0:13:06.000 it's encountered the sandbags that were unexpected. And so this 0:13:06.080 --> 0:13:07.800 is what I find interesting is if it's a wide 0:13:07.880 --> 0:13:10.320 lane and she's saying that it was hugging the right 0:13:10.360 --> 0:13:12.839 hand side of that lane trying to make the turn 0:13:12.880 --> 0:13:15.640 as most humans do, if it was in the center 0:13:15.679 --> 0:13:18.120 of the lane, she's saying, if it had just behaved 0:13:18.160 --> 0:13:19.839 as they normally would do it. You know that it 0:13:19.840 --> 0:13:21.679 would be in the exact dead center of that lane. 0:13:22.000 --> 0:13:24.280 The bus wouldn't have had the gap, I guess, and 0:13:24.320 --> 0:13:26.319 try to try to make it through that gap, so 0:13:26.640 --> 0:13:28.880 it would have just been behind the car. Never would 0:13:28.880 --> 0:13:31.680 have happened. So she's saying, in effect, because we're trying 0:13:31.679 --> 0:13:33.800 to make it mimic human behavior and we were hugging 0:13:33.840 --> 0:13:37.040 that right side, that's why this accident happened. Maybe we 0:13:37.040 --> 0:13:39.240 shouldn't have done that. But then again they come back 0:13:39.240 --> 0:13:43.120 and say that that's absolutely necessary for them to mimic 0:13:43.200 --> 0:13:45.720 human behavior, because if they don't that causes trouble as well. 0:13:45.760 --> 0:13:48.880 There's brother issues that are right. If a vehicle, let's 0:13:48.880 --> 0:13:52.839 say that it's an autonomous car heading toward an intersection, 0:13:53.720 --> 0:13:58.880 the light goes from green to amber, and there's technically 0:13:59.040 --> 0:14:03.680 enough space for the car to break safely and come 0:14:03.679 --> 0:14:07.280 to a complete stop as the light turns red. Knowing 0:14:07.320 --> 0:14:11.760 that most humans would just gun it, or at least 0:14:11.800 --> 0:14:13.640 just continue at the same speed to go through the 0:14:13.679 --> 0:14:17.960 intersection while it's still amber, you might want to think 0:14:18.000 --> 0:14:20.160 about that when you're designing your autonomous car so that 0:14:20.200 --> 0:14:23.000 you don't cause a pile up behind you, like you don't. 0:14:24.000 --> 0:14:27.440 If the person directly behind the autonomous car expects the 0:14:27.440 --> 0:14:29.800 car in front of them to continue through the intersection, 0:14:30.160 --> 0:14:32.800 you could potentially get rear ended. Yeah, that happens a 0:14:32.880 --> 0:14:34.720 thousand times around. I mean more than that. But it 0:14:34.720 --> 0:14:38.360 happens all over the world, really, especially in Atlanta, where 0:14:38.360 --> 0:14:40.760 the rule is if the light turns red, three cars 0:14:40.760 --> 0:14:43.720 get to go through. It is so true, isn't it. Yeah, 0:14:43.760 --> 0:14:46.640 it seems like once one goes through, two more following. Yeah, 0:14:46.640 --> 0:14:49.800 it's crazy. Yeah, I've seen it happen in multiple places 0:14:49.800 --> 0:14:52.920 around the city. There's some neighborhoods where it's worse. I'm 0:14:52.920 --> 0:14:55.920 not going to name any names, buckhead, but I'm just saying, yeah, 0:14:56.080 --> 0:14:59.440 never accelerate immediately on a green light anywhere in this area. 0:14:59.680 --> 0:15:02.240 You can expect there's going to be that that odd 0:15:02.240 --> 0:15:04.040 ball car that comes through every right, and there's gonna 0:15:04.040 --> 0:15:06.200 be like a car that's been waiting to turn left 0:15:06.240 --> 0:15:08.040 the whole time, and they say, I'm not waiting another 0:15:08.080 --> 0:15:10.320 light cycle. I'm just going now, like even if they're 0:15:10.320 --> 0:15:12.600 behind the stop line. Yeah, I'm sure, I'm sure it 0:15:12.600 --> 0:15:16.760 happens everywhere's it's particularly bad in these congested areas. Yes, 0:15:17.120 --> 0:15:20.920 So to your point, it is important to take those 0:15:20.960 --> 0:15:23.680 things into consideration when designing the autonomous car. You don't 0:15:23.720 --> 0:15:26.480 want an autonomous car to drive like an inconsiderate jerk 0:15:26.520 --> 0:15:28.720 of a driver. But at the same time, you can't 0:15:28.720 --> 0:15:32.840 have it be so clinically precise that it is standing 0:15:32.920 --> 0:15:35.560 out from all the other drivers. The only way that 0:15:35.600 --> 0:15:37.640 works is if you get to a point where you're 0:15:37.680 --> 0:15:40.560 at a saturation point with autonomous cars on the road, 0:15:41.000 --> 0:15:43.760 where then you can affect the behavior on a mass 0:15:43.800 --> 0:15:47.560 scale across a fleet of cars and not have that 0:15:47.760 --> 0:15:53.040 issue of human drivers having awful interactions with robotic drivers exactly. 0:15:53.080 --> 0:15:56.200 Here's here's the way they stated these these spokesman stated. 0:15:56.400 --> 0:15:58.600 They say, it's vital for us to develop advanced skills 0:15:58.640 --> 0:16:01.080 that respect not just the letter of the traffic code, 0:16:01.440 --> 0:16:03.960 but the spirit of the road. I think that's a 0:16:03.960 --> 0:16:05.400 good way to put it. The spirit of the road. 0:16:05.400 --> 0:16:07.920 I understand that. I completely get that when I read 0:16:08.000 --> 0:16:11.000 it is that, yeah, there's little rules here and there 0:16:11.080 --> 0:16:14.280 that we'd bend, but everybody bends them, and you expect 0:16:14.360 --> 0:16:16.600 you you understand how other drivers are going to behave 0:16:16.600 --> 0:16:19.640 in the same situation, and you expect that to happen, right, 0:16:19.840 --> 0:16:21.920 and you behave in that way and it all works. 0:16:22.280 --> 0:16:25.040 But when something comes in, a spoiler comes in and 0:16:25.200 --> 0:16:27.320 it follows exactly to the letter of the log the 0:16:27.360 --> 0:16:30.520 way that's supposed to happen, that person maybe you know 0:16:30.600 --> 0:16:34.240 the standout right. Another great example of that in Atlanta 0:16:34.360 --> 0:16:37.240 would be we have a couple of different highways that 0:16:37.400 --> 0:16:39.560 run through the city and one that runs around the 0:16:39.640 --> 0:16:42.920 city two eighty five, and two eighty five is often 0:16:43.000 --> 0:16:45.200 thought of as the type of highway that if you 0:16:45.280 --> 0:16:47.480 get on it, you have to speed. You cannot go 0:16:47.640 --> 0:16:49.480 the speed limit on two eighty five. It's just too 0:16:49.560 --> 0:16:52.160 dangerous because everyone else is going above the speed limit. 0:16:52.280 --> 0:16:56.040 Massive truck traffic, Yeah, yeah, and there are enormous, enormous 0:16:56.080 --> 0:16:59.400 semis rushing down there, and you don't want to get 0:16:59.600 --> 0:17:02.440 you to be poking along when they come up behind you. 0:17:02.840 --> 0:17:06.119 So again, an autonomous car would need to have that 0:17:06.280 --> 0:17:09.399 information and take that into account, unless you got to 0:17:09.480 --> 0:17:11.720 a point where you had so many autonomous vehicles on 0:17:11.760 --> 0:17:14.960 the road that it was no longer a concern. Yeah, 0:17:14.960 --> 0:17:17.160 And this is where we discussed this yesterday, because we're 0:17:17.200 --> 0:17:18.840 talking off air about this just a little bit to 0:17:19.280 --> 0:17:22.520 prep for today, and the idea would be that's kind 0:17:22.560 --> 0:17:25.720 of like schooling, almost like fish schooling, and that the 0:17:25.800 --> 0:17:28.080 cars know where the other one is at all times 0:17:28.119 --> 0:17:30.280 and they can communicate between them. The problem is when 0:17:30.320 --> 0:17:32.920 you throw in the human driver element into that mix, 0:17:33.520 --> 0:17:36.239 or you know, if you have just one autonomous car 0:17:36.400 --> 0:17:39.040 among all humans, that's the other problem. It's the other issue, 0:17:39.240 --> 0:17:41.719 and right now, that's the battle that they're fighting right right. 0:17:41.800 --> 0:17:43.480 So once we get to a point where there's that 0:17:43.880 --> 0:17:47.840 tipping point one way or the other, then things will 0:17:47.880 --> 0:17:51.520 be very different. But there's going to be some growing paints. 0:17:51.560 --> 0:17:53.840 And this also leads into something that I talked about 0:17:54.160 --> 0:17:58.040 earlier in twenty sixteen when I went to CEES Toyota 0:17:58.119 --> 0:18:03.880 had their big AI discussion. You know, they're investing millions 0:18:03.920 --> 0:18:07.439 of dollars in AI research for autonomous cars and beyond, 0:18:08.640 --> 0:18:10.600 and one of the things they've talked about was how 0:18:11.480 --> 0:18:15.440 autonomous cars in general are really really good at handling 0:18:15.520 --> 0:18:19.359 all the mundane stuff that you would typically encounter on 0:18:19.480 --> 0:18:22.080 a normal day driving from point A to point B. 0:18:22.920 --> 0:18:25.720 What they are not good at is dealing with stuff 0:18:25.880 --> 0:18:29.440 that's outside of that norm and the sandbags that we 0:18:29.560 --> 0:18:32.399 talked about earlier. Would be a great example of that. 0:18:32.960 --> 0:18:36.760 It's some form of obstruction that's partially blocking off part 0:18:36.760 --> 0:18:42.840 of the road and that ends up causing a different scenario, 0:18:43.400 --> 0:18:48.200 and sometimes the card behaves in a way that works 0:18:48.240 --> 0:18:51.240 out for everybody. In this case, it didn't. And it's 0:18:52.240 --> 0:18:54.720 not that the car couldn't handle the situation. It's just 0:18:54.840 --> 0:18:57.440 that the method that the car used turned out to 0:18:57.640 --> 0:19:01.760 not be reliable. Yeah, this was an extremely slow speed crash, 0:19:01.800 --> 0:19:04.640 as we've said. Yes, the bus was traveling fifteen miles 0:19:04.680 --> 0:19:06.239 per hour in the other lane trying to get through 0:19:06.280 --> 0:19:08.879 that gap. But the Google car was traveling I think 0:19:08.920 --> 0:19:11.920 they said to two miles per hour. Yeah, very very slow, 0:19:12.359 --> 0:19:15.520 very slow speed. So the thing is like with the 0:19:15.640 --> 0:19:18.840 with the compensation for this, you know, the thirty five 0:19:18.960 --> 0:19:22.280 hundred tests that they're now going to run additional tests, uh, 0:19:22.359 --> 0:19:25.800 to determine or to find a way around that situation. 0:19:25.920 --> 0:19:28.200 So it's never going to happen again. We're gonna do 0:19:28.240 --> 0:19:31.199 everything we can, but to to think about it that way, 0:19:31.280 --> 0:19:33.760 to say, the thirty five hundred tests that are going 0:19:33.840 --> 0:19:37.080 to allow this vehicle to to think about that exact 0:19:37.160 --> 0:19:39.639 situation and never let it happen again where where it 0:19:39.680 --> 0:19:42.040 just kind of noses out into into a lane that 0:19:42.119 --> 0:19:45.240 appears open. Yeah, that's remarkable. I mean it just lets 0:19:45.280 --> 0:19:48.119 you know that there there are tens of thousands of 0:19:48.359 --> 0:19:52.480 of um programs or are thoughts, I don't know how 0:19:52.560 --> 0:19:54.399 to say it, right, that are they're going through this 0:19:54.520 --> 0:19:58.520 thing at all times, you know, um as calculations and 0:19:58.600 --> 0:20:01.480 parameters and just you know, if this then that you know, 0:20:01.560 --> 0:20:04.400 those scenarios are being run all the time. It's just incredible, 0:20:04.480 --> 0:20:06.560 mind boggling, it really is. And I was looking into 0:20:06.720 --> 0:20:09.480 you said one one point four or five million miles 0:20:09.560 --> 0:20:12.840 have been driven, uh flawlessly, right, I mean they hadn't 0:20:12.880 --> 0:20:15.440 have any problems, you know, at at fault I guess 0:20:15.760 --> 0:20:19.359 the autonomous vehicle. Do you know how much they test 0:20:19.920 --> 0:20:23.760 on a daily basis and on actually on a daily basis? No? No, okay, 0:20:23.760 --> 0:20:25.240 well let's see, I've got I got a note here. 0:20:25.240 --> 0:20:27.879 I should have looked for that reading that. Okay, here 0:20:27.920 --> 0:20:29.760 we go. All right, so actually this is a per 0:20:29.840 --> 0:20:32.080 week and then a per day thing. All right, They 0:20:32.200 --> 0:20:35.480 drive ten thousand miles per week and that's like you know, 0:20:35.680 --> 0:20:38.200 somebody in a vehicle on the road ten thousand miles 0:20:38.280 --> 0:20:41.640 per week, per day though this number is incredible. Per day, 0:20:42.359 --> 0:20:46.879 they are driving three million miles of computer simulation miles. 0:20:47.240 --> 0:20:50.160 Oh so three million the equivalent of three million miles. 0:20:50.240 --> 0:20:52.280 That's because they can quickly just go through that and 0:20:52.359 --> 0:20:56.359 have multiple systems running these things, you know, So the 0:20:56.680 --> 0:20:58.280 amount of testing that they do in a year is 0:20:58.320 --> 0:21:00.879 just unbelievable. I don't have yearly stats or anything, but 0:21:00.920 --> 0:21:04.320 you can extrapolate those numbers to that well. And the 0:21:04.440 --> 0:21:07.680 other point in the Toyota press conference that was interesting 0:21:07.760 --> 0:21:09.520 to me, and this goes back to what you were 0:21:09.520 --> 0:21:11.760 saying at the beginning of the show about holding autonomous 0:21:11.840 --> 0:21:14.720 cars at a different standard than we hold human drivers. 0:21:16.080 --> 0:21:20.520 They talked about how you offer a lot of the 0:21:20.600 --> 0:21:26.440 autonomous car industry talks about the one hundred million mile benchmark, 0:21:26.560 --> 0:21:29.400 saying that you want you want one hundred million miles 0:21:29.480 --> 0:21:33.960 traveled of proven safety. And they said, you know that's 0:21:34.000 --> 0:21:36.399