WEBVTT - Driverless Cars from 2007 to Today 0:00:04.120 --> 0:00:07.160 Get in touch with technology with tech Stuff from how 0:00:07.200 --> 0:00:14.080 stuff Works dot com. Hey there, and welcome to tech Stuff. 0:00:14.120 --> 0:00:17.360 I'm your host, Jonathan Strickland. I'm an executive producer with 0:00:17.360 --> 0:00:19.840 How Stuff Works in iHeart radio and I love all 0:00:20.040 --> 0:00:23.479 things tech. And in our last episode, I talked about 0:00:23.520 --> 0:00:27.720 the two thousand seven Urban Challenge as the autonomous car 0:00:27.880 --> 0:00:31.680 competition that DARPA held, and they were specifically trying to 0:00:31.720 --> 0:00:35.800 spur on innovation in driver less car technology. One of 0:00:35.800 --> 0:00:39.720 the elements I didn't really talk about was how cooperative 0:00:39.920 --> 0:00:43.000 the experience was. I mentioned that there was sort of 0:00:43.000 --> 0:00:46.080 this air of cooperation, and I talked about how Dave 0:00:46.159 --> 0:00:50.559 Hall had created a lidar tool that ended up being 0:00:50.680 --> 0:00:54.920 used by lots of the teams, but it went well 0:00:54.960 --> 0:00:59.000 beyond that. Teams were eager to share their approaches and 0:00:59.040 --> 0:01:02.480 their technologies each other, the algorithms they were using for 0:01:03.240 --> 0:01:07.600 decision making processes. There was a lot of excitement as 0:01:07.640 --> 0:01:11.160 these different, very intelligent groups of people got together and 0:01:11.200 --> 0:01:15.720 began to cross pollinate their ideas. The real goal wasn't 0:01:15.720 --> 0:01:19.840 necessarily to beat out everyone else, although of course everyone 0:01:19.840 --> 0:01:22.600 would have liked to have been on the winning team. 0:01:23.160 --> 0:01:26.679 But really the real goal was to overcome this huge 0:01:26.840 --> 0:01:31.560 engineering challenge of developing technologies that would be necessary for 0:01:31.600 --> 0:01:35.240 a car to maneuver through an urban environment safely under 0:01:35.280 --> 0:01:39.920 its own power, following all applicable traffic laws, and integrating 0:01:39.959 --> 0:01:45.000 with human driven vehicles in that space. Some people like 0:01:45.240 --> 0:01:48.760 Sebastian Throne of Stanford Racing Team, would say that the 0:01:48.920 --> 0:01:52.720 greatest achievement to come out of the challenges wasn't winning 0:01:52.840 --> 0:01:57.480 top prize. It was when teams would share their knowledge 0:01:57.800 --> 0:02:01.000 and their experience with one another, allowing separate lines of 0:02:01.040 --> 0:02:04.360 research and development to start to converge, and it's set 0:02:04.400 --> 0:02:07.960 the ground for what would come next. Now, the challenges 0:02:08.000 --> 0:02:10.560 in two thousand four, two thousand five, and two thousand 0:02:10.680 --> 0:02:15.400 seven proved that autonomous cars could exist on some level, 0:02:15.800 --> 0:02:18.320 and a lot of the technology that went into making 0:02:18.360 --> 0:02:23.679 cars navigate and operate independently would go into other systems 0:02:23.680 --> 0:02:28.480 and other applications, stuff like various driver assist technologies like 0:02:28.600 --> 0:02:32.799 lane detection or automatic breaking. And it also encouraged some 0:02:32.840 --> 0:02:36.000 big companies to continue supporting efforts to making driver lest 0:02:36.040 --> 0:02:40.320 cars a reality. So one thing to acknowledge is, even 0:02:40.360 --> 0:02:42.680 if we were to say we're never going to be 0:02:42.919 --> 0:02:47.520 at a time where truly autonomous cars, ones that could 0:02:47.520 --> 0:02:51.320 take us from any starting location to Indy, any ending 0:02:51.400 --> 0:02:55.959 location that's connected by roads to our starting point. Even 0:02:55.960 --> 0:02:58.000 if we say that's never going to be possible, not 0:02:58.120 --> 0:03:02.919 truly rely Apple, what we can say is we've already 0:03:02.919 --> 0:03:07.840 seen numerous technologies that were spawned by these competitions that 0:03:08.000 --> 0:03:11.120 are we're not only going into cars today, but are 0:03:11.200 --> 0:03:16.880 saving lives in the process. So that's that's already a 0:03:16.960 --> 0:03:22.520 great outcome to this DARPA challenge, even if you, you know, 0:03:22.720 --> 0:03:27.960 again agree that ultimately this was to make more more 0:03:28.280 --> 0:03:32.760 automated military ground combat vehicles, and even in that case 0:03:33.400 --> 0:03:37.680 that was intended to help save soldiers lives. So that's 0:03:37.720 --> 0:03:41.440 a noble endeavor as well. But let's talk about some 0:03:41.480 --> 0:03:44.960 of the companies and organizations that formed in the wake 0:03:45.400 --> 0:03:48.320 of these grand challenges. You had General Motors and Carnegie 0:03:48.360 --> 0:03:51.800 Mellon University get together. They launched a research and development 0:03:51.800 --> 0:03:56.800 program called the Autonomous Driving Collaborative Research Lab. Then you 0:03:56.840 --> 0:04:01.480 also had Volkswagen establish a similar effort with stand For University. Now, 0:04:01.480 --> 0:04:04.960 those academic research programs didn't receive nearly as much public 0:04:05.000 --> 0:04:09.480 attention as the other really big entity that got involved 0:04:09.480 --> 0:04:13.520 in the field not long after the challenges, and that 0:04:13.560 --> 0:04:17.760 would be Google, later known as Alphabet or way Mo. 0:04:18.279 --> 0:04:21.760 So Alphabets, the parent companies, the holding company under which 0:04:22.320 --> 0:04:25.320 UH the smaller companies and smaller by smaller, I just 0:04:25.400 --> 0:04:29.560 mean hierarchically would spin off. So you have like Google, 0:04:29.800 --> 0:04:33.080 which continues to focus on the core business of the 0:04:33.120 --> 0:04:37.200 overall company, but you have these other subsidiaries that are 0:04:37.240 --> 0:04:40.760 subsidiaries of Alphabet, not of Google anyway. One of those 0:04:40.760 --> 0:04:46.240 would be a division that was specifically in research and 0:04:46.279 --> 0:04:49.640 development in the area of autonomous cars. So in two 0:04:49.680 --> 0:04:53.279 thousand seven, Sebastian Throne took a sabbatical from being a 0:04:53.279 --> 0:04:56.280 professor at Stanford to go work at Google for a 0:04:56.279 --> 0:04:59.480 short while. Turned out to be longer than a short while, 0:04:59.520 --> 0:05:03.159 but he and a team were hired on essentially to 0:05:03.320 --> 0:05:06.800 help develop Google's street view tool. That's the tool that's 0:05:06.839 --> 0:05:11.440 integrated into various maps systems that Google does and allows 0:05:11.440 --> 0:05:15.360 you to take a street level look at different locations. 0:05:15.880 --> 0:05:19.159 It was done by driving special cars outfitted with special 0:05:19.200 --> 0:05:23.240 cameras through all these different streets and all these different locations. 0:05:23.720 --> 0:05:27.080 They had about a year UH scheduled to try and 0:05:27.160 --> 0:05:30.520 map out all the roads they could in various major cities, 0:05:30.560 --> 0:05:32.560 and they ended up doing it in nine months, so 0:05:32.600 --> 0:05:36.840 it's pretty impressive anyway. That project has had plenty of 0:05:36.880 --> 0:05:40.760 news coverage, not just because of its utility, but also 0:05:40.839 --> 0:05:44.680 because of concerns about privacy and security. Not everyone is 0:05:44.720 --> 0:05:48.000 super crazy about the idea of using an online tool 0:05:48.080 --> 0:05:51.440 to virtually coast down streets and take a look at 0:05:52.080 --> 0:05:56.800 different addresses, not to mention possibly spying people that you 0:05:56.920 --> 0:06:00.560 recognize in places they should not be, but that's a 0:06:00.600 --> 0:06:04.400 topic for another time. In two thousand nine, Google would 0:06:04.440 --> 0:06:09.200 go a step further and secretly began testing autonomous vehicles. 0:06:09.240 --> 0:06:11.560 So they had been in development of that for a 0:06:11.600 --> 0:06:13.760 couple of years as well, and Sebastian Thron had worked 0:06:13.760 --> 0:06:19.280 on that. They were retrofitting Toyota Prius cars at first, 0:06:19.720 --> 0:06:23.279 and the goal was to conduct ten one hundred mile 0:06:23.560 --> 0:06:27.760 trips in those cars without interruption, So uninterrupted one mile 0:06:27.839 --> 0:06:32.360 trips with Toyota previous is ten times. Over the course 0:06:32.400 --> 0:06:36.320 of two thousand nine, Google would rack up more autonomously 0:06:36.440 --> 0:06:40.600 driven miles than had been accumulated over all previous years 0:06:40.600 --> 0:06:46.200 of experimentation among all autonomous car programs, so in one 0:06:46.279 --> 0:06:50.680 year they eclipsed everything that had come before it. In 0:06:51.760 --> 0:06:55.279 Sebastian Throne would officially join Google as a Google Fellow 0:06:55.520 --> 0:06:58.960 and found a secret research and development division within the 0:06:58.960 --> 0:07:04.120 company called Goal X. Now, among the many projects that 0:07:04.160 --> 0:07:07.240 division would focus on, the autonomous cars was just one 0:07:08.320 --> 0:07:11.960 would be the self driving cars, and Thron, having participated 0:07:12.080 --> 0:07:15.240 in all three of the DARPA challenges, had really deep 0:07:15.360 --> 0:07:18.120 contacts in that field, and he could call upon them 0:07:18.440 --> 0:07:21.880 to help solve difficult problems that remained in pursuit of 0:07:21.920 --> 0:07:25.640 the goal of a truly autonomous car that could interoperate 0:07:25.720 --> 0:07:30.320 with human society. So Thron ended up hiring people from 0:07:30.360 --> 0:07:34.040 the various competing teams of the DARPA challenges to join 0:07:34.240 --> 0:07:38.280 him at Google in developing further autonomous car technologies. Technically, 0:07:38.480 --> 0:07:41.840 the company had been doing this since two thousand seven. Really, 0:07:42.840 --> 0:07:45.440 Dave Hall, the guy who suited up lidar to make 0:07:45.480 --> 0:07:49.200 it an indispensable tool, became a billionaire through his work 0:07:49.200 --> 0:07:52.080 at Velo Dine. In the meantime, and read Whittaker over 0:07:52.120 --> 0:07:56.760 at Carnegie, Melon would continue teaching at that university, effectively 0:07:56.800 --> 0:08:00.920 training the next generation of roboticists and compute of scientists 0:08:00.920 --> 0:08:05.680 throughout the entire industry were continuing to develop machine learning 0:08:05.760 --> 0:08:10.560 strategies that would become useful in multiple applications, including teaching 0:08:10.560 --> 0:08:14.520 a car how to drive itself. There were numerous groups 0:08:14.560 --> 0:08:17.720 that thought, this is the real secret to developing a 0:08:17.760 --> 0:08:22.120 truly autonomous car is through machine learning. Not programming a 0:08:22.120 --> 0:08:25.800 car on what to do in any given situation, but 0:08:26.000 --> 0:08:29.600 teaching a car, sort of akin to how you would 0:08:29.640 --> 0:08:33.760 teach a young driver about how to conduct him or 0:08:33.800 --> 0:08:38.640 herself in a car. By this stage, Google's tests had 0:08:38.720 --> 0:08:42.760 drawn attention from the press. The New York Times published 0:08:42.760 --> 0:08:47.000 a piece in October twenty ten titled Google Cars drive 0:08:47.080 --> 0:08:51.040 themselves in Traffic. The article revealed that Google had been 0:08:51.040 --> 0:08:55.079 conducting tests in plain site for several months, but at 0:08:55.120 --> 0:08:57.959 that point had not commented on what those tests were 0:08:58.000 --> 0:09:01.880 all about, and the company was content with people just 0:09:01.960 --> 0:09:05.600 assuming that it was another Google street view car. There 0:09:05.600 --> 0:09:09.439 were fifteen engineers working on the project, and they had 0:09:09.520 --> 0:09:13.360 hired a dozen or so drivers whose job it was 0:09:13.400 --> 0:09:16.280 to monitor the performance of the vehicles and to take 0:09:16.320 --> 0:09:20.800 over if necessary. The engineers testing the vehicles had three 0:09:20.880 --> 0:09:25.240 main ways that they could take control back from the car. 0:09:26.160 --> 0:09:30.240 In each test vehicle, engineers had installed a button, a 0:09:30.360 --> 0:09:33.560 nice candy like red button that was on the right 0:09:33.559 --> 0:09:36.679 hand side of the driver so they could easily hit 0:09:36.720 --> 0:09:39.200 the button and that would switch controls to manual. But 0:09:39.400 --> 0:09:40.840 you can also do it if you tapped on the 0:09:40.880 --> 0:09:44.000 brakes or if you turned the steering wheel, the car 0:09:44.080 --> 0:09:48.640 would hand over a control to the driver. Google had 0:09:48.760 --> 0:09:52.880 a little hiccup in two thousand eleven, a public hiccup 0:09:53.480 --> 0:09:55.560 that was when one of its driver less vehicles was 0:09:55.600 --> 0:09:58.240 involved in a low speed car crash, and at the 0:09:58.280 --> 0:10:03.240 time everyone assumed this was the first uh in real 0:10:03.320 --> 0:10:07.840 world car crash with an autonomous car. However, this particular 0:10:07.880 --> 0:10:11.080 incident was only a very tiny hiccup because it soon 0:10:11.120 --> 0:10:15.120 became public that the car, the autonomous car, was actually 0:10:15.200 --> 0:10:18.440 in manual control mode at the time. The human driver 0:10:18.520 --> 0:10:23.200 behind the wheel was responsible for the accident, not the 0:10:23.280 --> 0:10:27.720 driver lest car technology. Google's driver lest cars would maintain 0:10:27.840 --> 0:10:36.040 a perfect safety record in autonomous mode until publicly. That's 0:10:36.040 --> 0:10:39.240 assuming if you say perfect safety record in the sense 0:10:39.320 --> 0:10:43.400 that the autonomous vehicles were not found to be at fault. 0:10:43.640 --> 0:10:47.760 There were incidents where autonomous cars were involved in car accidents, 0:10:47.800 --> 0:10:52.480 but in every case until two thousand sixteen, every public 0:10:52.720 --> 0:10:57.959 case it was determined that the other driver was at fault. Again, 0:10:58.559 --> 0:11:01.360 stress on public. I'll back to what I mean in 0:11:01.400 --> 0:11:05.240 just a few minutes. In two thousand twelve, Google began 0:11:05.320 --> 0:11:08.440 to expand its fleet of driverless cars. It added a 0:11:08.559 --> 0:11:12.280 Lexus r x f H to the mix, so it 0:11:12.320 --> 0:11:15.160 wasn't just Toyota Prius is Uh. They had a couple 0:11:15.160 --> 0:11:17.840 of others as well, and the company began to develop 0:11:17.920 --> 0:11:22.080 its own sensors and began to replace the off the 0:11:22.120 --> 0:11:24.600 shelf kind of stuff it was buying. I mean off 0:11:24.640 --> 0:11:26.679 the shelf if you know which shelves to look at, 0:11:26.840 --> 0:11:31.640 and still pretty exclusive materials, but now the company was 0:11:31.720 --> 0:11:35.959 developing stuff in house purpose built for their own cars. 0:11:36.520 --> 0:11:39.160 Two thousand twelve was also when a few Google employees 0:11:39.160 --> 0:11:42.600 were allowed to start testing this technology on highways around 0:11:42.600 --> 0:11:46.640 Google's campus, so it was outside of just the direct 0:11:46.800 --> 0:11:50.960 team working on the project. Now other Google employees could 0:11:51.440 --> 0:11:54.480 end up driving an autonomous car and allowing it to 0:11:54.559 --> 0:11:59.080 operate an autonomous mode in specific regions and under specific 0:11:59.160 --> 0:12:02.040 sets of circumstance answers. There are a lot of different 0:12:02.120 --> 0:12:05.040 rules in place to participate in this, so you couldn't 0:12:05.080 --> 0:12:08.360 just turn it on autonomous and sit back all the time. 0:12:08.840 --> 0:12:11.280 And two thousand twelve was when the state of Nevada 0:12:11.360 --> 0:12:14.920 made history by becoming the first state to license autonomous 0:12:14.920 --> 0:12:18.760 cars for use on state roads. California would follow suit 0:12:19.000 --> 0:12:22.200 that same year, but the bill that Governor Jerry Brown 0:12:22.480 --> 0:12:26.320 had signed would only take effect starting in two thousand fifteen. 0:12:26.840 --> 0:12:30.920 In January two thousand thirteen, Audi and Toyota both showed 0:12:30.920 --> 0:12:34.400 off autonomous vehicle concept designs at c S, so it 0:12:34.480 --> 0:12:38.120 showed that lots of UH entities were still very much 0:12:38.160 --> 0:12:41.880 interested in driverless cars, not just Google. In two thousand 0:12:41.920 --> 0:12:45.520 and fourteen, Google unveiled a prototype electric car that had 0:12:45.559 --> 0:12:48.400 a top speed of twenty five miles per hour. There 0:12:48.480 --> 0:12:53.480 was no steering wheel, no break, no accelerator, no controls 0:12:53.920 --> 0:12:56.839 to allow a human driver to actually take manual control 0:12:56.880 --> 0:12:59.440 of the car. Inside the car, the car did have 0:12:59.480 --> 0:13:01.959 some control, but there were buttons that would tell it 0:13:02.040 --> 0:13:05.160 when it could go and when it should stop. So 0:13:05.440 --> 0:13:08.760 these vehicles were not intended for commercial use. They were 0:13:08.800 --> 0:13:11.400 part of Google's R and D to to test out 0:13:11.480 --> 0:13:15.960 the possibility of a vehicle that doesn't even have manual controls. 0:13:16.840 --> 0:13:20.520 It requires a lot of trust be put into the system. Now. 0:13:20.520 --> 0:13:24.560 To be fair, these were very limited in their scope. 0:13:24.920 --> 0:13:27.600 Top speed of twenty five hour suggests that you would 0:13:27.720 --> 0:13:30.800 use them on like residential streets and stuff. You wouldn't 0:13:31.440 --> 0:13:33.000 use them on a highway. They wouldn't be able to 0:13:33.040 --> 0:13:35.800 get up to speed. In two thousand and fourteen, the 0:13:35.920 --> 0:13:41.160 state of California past legislation requiring any company operating autonomous 0:13:41.200 --> 0:13:45.000 vehicles in the state to submit reports on any accident 0:13:45.120 --> 0:13:49.160 involving a vehicle operating an autonomous mode that would result 0:13:49.200 --> 0:13:53.280 in quote damage of property or in bodily injury or 0:13:53.440 --> 0:13:59.000 death end quote. After that point, Google would report several accidents, 0:13:59.080 --> 0:14:03.079 like more than three dozen. Most of those appeared to 0:14:03.080 --> 0:14:06.480 be the fault of human drivers, not the autonomous systems. 0:14:06.840 --> 0:14:10.120 But according to some Google executives, there were some accidents 0:14:10.160 --> 0:14:13.600 that happened between two thousand eleven and when this piece 0:14:13.640 --> 0:14:17.520 of legislation was signed in two thousand fourteen that Google 0:14:17.640 --> 0:14:22.800 chose to keep quiet, and the logic that the company 0:14:22.880 --> 0:14:25.800 used was that, well, those accidents happened before that law 0:14:25.920 --> 0:14:28.040 was passed, so it doesn't really apply to those. We 0:14:28.040 --> 0:14:32.240 don't need to talk about accidents that have already happened. 0:14:32.680 --> 0:14:35.480 Just from this point forward, we'll talk about it. One 0:14:35.480 --> 0:14:39.960 of those accidents involved someone who had participated in two 0:14:40.080 --> 0:14:43.520 of the DARPA Grand Challenges, someone who was working for 0:14:43.600 --> 0:14:46.600 Google and who had become something of a thorn in 0:14:46.640 --> 0:14:52.200 the company's side, and that person was Anthony Lewandowski. I'll 0:14:52.240 --> 0:14:54.480 explain more in a second, but first let's take a 0:14:54.560 --> 0:15:06.040 quick break to thank our sponsor Lewandowski. If you listen 0:15:06.120 --> 0:15:08.800 to the previous episodes, he was the guy who competed 0:15:09.160 --> 0:15:13.080 using a self balancing motorcycle in the two thousand four 0:15:13.160 --> 0:15:16.200 and two thousand five Grand Challenge competitions. The name of 0:15:16.200 --> 0:15:19.480 that motorcycle, by the way, no big surprise, was ghost Writer. 0:15:20.280 --> 0:15:22.960 He had joined Google in two thousand seven to work 0:15:22.960 --> 0:15:26.160 with Sebastian Thrun on the Google street View project, and 0:15:26.240 --> 0:15:30.440 Lewandowski started a few companies related to autonomous cars. He 0:15:30.840 --> 0:15:33.600 had developed these technologies for ghost Writer, and then he 0:15:33.640 --> 0:15:38.080 started some startups that were focused on specific elements that 0:15:38.160 --> 0:15:41.360 he had created for a ghost Writer, including one that 0:15:41.600 --> 0:15:44.840 used spinning light AR as a sensing technology. He didn't 0:15:45.480 --> 0:15:48.440 invent that, but he did develop our start up a 0:15:48.480 --> 0:15:53.080 company that specialized in that then he pushed Google to 0:15:53.280 --> 0:15:57.080 buy the tech his other companies happened to be making, 0:15:57.680 --> 0:16:01.000 which you might think is a little questionable. Lewandowski was 0:16:01.000 --> 0:16:05.040 in a position to market his own company's products for 0:16:05.120 --> 0:16:07.800 a project he was working on with Google, so he 0:16:07.840 --> 0:16:10.920 was in charge of hardware for autonomous cars over at Google, 0:16:11.520 --> 0:16:13.720 and then said, well, for us to outfit these cars, 0:16:13.800 --> 0:16:16.640 let's buy this tech from this these two companies. Oh, 0:16:16.680 --> 0:16:19.280 I happen to own those two companies, So I'm using 0:16:19.480 --> 0:16:22.600 the money from the company I work for to funnel 0:16:22.640 --> 0:16:26.520 into companies I own. People began to have questions about that, 0:16:27.920 --> 0:16:31.200 but it totally worked. As people at Google learned about 0:16:31.240 --> 0:16:33.240 what had gone on, they started to get a little 0:16:33.280 --> 0:16:37.440 concerned about this and about Lewandowski. Uh. It got worse 0:16:37.520 --> 0:16:41.560 when it became known that Lewandowski was talking to some competitors, 0:16:41.680 --> 0:16:44.800 some other companies outside of Google about selling them the 0:16:44.880 --> 0:16:49.000 same technology. This was complicated because he was operating those 0:16:49.040 --> 0:16:52.640 businesses outside of Google. He was a business owner and 0:16:52.640 --> 0:16:54.840 and those technically didn't belong to Google. But at the 0:16:54.840 --> 0:16:58.720 same time he was working for Google, so it seemed 0:16:58.760 --> 0:17:03.480 like he might be undercutting Google or helping out their competitors, 0:17:04.200 --> 0:17:09.040 which was complicated. Larry Page, the head of Google, essentially 0:17:09.160 --> 0:17:14.200 directed the company to acquire Lewandowski's businesses rather than risk 0:17:14.320 --> 0:17:18.880 him leaving the company to oversee the businesses himself. Lewandowski 0:17:18.920 --> 0:17:23.240 had indicated that he might leave Google in pursuit of 0:17:23.720 --> 0:17:26.200 leading up these these two companies that he had founded, 0:17:26.920 --> 0:17:33.400 although his commitment to that course of action is something 0:17:33.520 --> 0:17:36.560 that people have questioned that perhaps he just said this 0:17:37.000 --> 0:17:41.000 as a way to encourage Larry Page to shell out 0:17:41.240 --> 0:17:45.960 millions and millions of dollars to acquire these companies. According 0:17:46.000 --> 0:17:49.000 to a report in The New Yorker, several people in 0:17:49.160 --> 0:17:53.720 Project Chauffeur, which was the name the code name for 0:17:53.800 --> 0:17:56.840 the driver less vehicle program at Google, felt that it 0:17:56.960 --> 0:18:00.560 was a mistake to get so tightly connected to levindal Ski. 0:18:00.800 --> 0:18:03.520 Several of the team members questioned his commitment to Google, 0:18:03.760 --> 0:18:06.920 as well as his ethical sensibilities, and this would get 0:18:06.920 --> 0:18:11.679 pushed further after an alleged incident in two thousand eleven. 0:18:11.760 --> 0:18:15.240 So here's how the story goes. There was a Google 0:18:15.280 --> 0:18:18.920 executive named Isaac Taylor who was working on Project Chauffeur, 0:18:19.280 --> 0:18:22.440 and he took a leave of absence from the project 0:18:22.880 --> 0:18:25.280 it was a paternity leave when he became a new 0:18:25.320 --> 0:18:28.880 father in two thousand eleven. When he got back to Google, 0:18:28.920 --> 0:18:32.440 he found out that Lewandowski had made some unauthorized changes 0:18:32.800 --> 0:18:36.600 to the software that guided the driverless cars. So up 0:18:36.600 --> 0:18:39.680 to that point, Google had really placed some pretty tight 0:18:39.720 --> 0:18:43.000 restrictions on the routes that these driverless cars would be 0:18:43.040 --> 0:18:46.080 allowed to take an autonomous mode. This is a type 0:18:46.119 --> 0:18:50.240 of geo fencing, where you restrict the operating parameters for 0:18:50.359 --> 0:18:53.560 a vehicle. The goal was to gather data through many 0:18:53.680 --> 0:18:57.160 miles of travel, but to control that process by limiting 0:18:57.240 --> 0:19:00.520 where the cars could actually drive under a top aymous mode. 0:19:00.960 --> 0:19:04.520 Lewandowski apparently felt this was not satisfactory, so he made 0:19:04.600 --> 0:19:06.960 changes to the code to let the autonomous cars drive 0:19:07.080 --> 0:19:12.639 on routes that previously had been forbidden. Taylor and Lewandowski 0:19:12.800 --> 0:19:17.359 had a rather spirited argument by all accounts, over at Google, 0:19:18.000 --> 0:19:21.600 and at that point Lewandowski demanded that Taylor accompany him 0:19:21.840 --> 0:19:24.240 on a ride in an autonomous car to show that 0:19:24.320 --> 0:19:27.800 this was a good idea. So, according to the story, 0:19:27.840 --> 0:19:30.760 the two people left in a retrofitted Prius and then 0:19:30.880 --> 0:19:34.199 hit the California Rhodes now I'm going to quote the 0:19:34.240 --> 0:19:37.320 New Yorker piece directly here. This is from an article 0:19:37.400 --> 0:19:41.600 titled did Uber steal Google's intellectual Property? And it was 0:19:41.640 --> 0:19:45.280 written by Charles Duig, and it was published October twenty two, 0:19:45.560 --> 0:19:48.439 two thousand eighteen, so you can find this. It's a 0:19:48.560 --> 0:19:53.760 very recent article. And here's Charles Dwig's description of this incident. 0:19:54.280 --> 0:19:57.639 The car went onto a freeway where it traveled past 0:19:57.800 --> 0:20:01.080 an on ramp. According to people with knowledge of events 0:20:01.119 --> 0:20:05.679 that day, the Prius accidentally boxed in another vehicle, a camera. 0:20:06.359 --> 0:20:09.800 A human driver could easily have handled the situation by 0:20:09.800 --> 0:20:13.760 slowing down and laying the camera merge into traffic, but 0:20:13.840 --> 0:20:18.159 Google software wasn't prepared for this scenario. The cars continued 0:20:18.200 --> 0:20:21.880 speeding down the freeway side by side. The cameras driver 0:20:22.080 --> 0:20:26.119 jerked his car onto the right shoulder, then, apparently trying 0:20:26.119 --> 0:20:29.439 to avoid a guardrail, he veered to the left. The 0:20:29.560 --> 0:20:34.720 camera pinwheeled across the freeway and into the median. Lewandowski, 0:20:34.800 --> 0:20:38.560 who acted as the safety driver, swerved hard to avoid 0:20:38.600 --> 0:20:42.359 colliding with the camera, causing Taylor to injure his spine 0:20:42.640 --> 0:20:47.360 so severely that he eventually required multiple surgeries. The Prius 0:20:47.400 --> 0:20:50.879 regained control and turned a corner on the freeway, leaving 0:20:50.920 --> 0:20:55.000 the camera behind. Lewandowski and Taylor didn't know how badly 0:20:55.080 --> 0:20:58.440 damaged the camera was. They didn't go back to check 0:20:58.560 --> 0:21:00.879 on the other driver or to see if anyone else 0:21:00.920 --> 0:21:05.200 had been hurt. Neither they nor other Google executives made 0:21:05.280 --> 0:21:09.439 inquiries with the authorities. The police were not informed that 0:21:09.520 --> 0:21:13.720 a self driving algorithm had contributed to the accident. That's 0:21:13.760 --> 0:21:17.680 not a great story. I mean it's written very well. 0:21:17.840 --> 0:21:20.080 No offense to the New Yorker or anything like that. 0:21:20.119 --> 0:21:23.280 I mean, it's not a great thing to have in 0:21:23.359 --> 0:21:27.640 your historical record, no matter who you are, and it's 0:21:27.680 --> 0:21:31.040 particularly bad for a company that used to have the 0:21:31.080 --> 0:21:36.639 motto don't be evil. Now Lewandowski would actually double down 0:21:37.280 --> 0:21:41.000 on the results of this incident. Rather than saying, whoops, 0:21:41.280 --> 0:21:47.600 my bad, I done did something terrible that caused damage, 0:21:47.600 --> 0:21:51.800 impossibly injury, certainly injury to his coworker, he ended up 0:21:51.840 --> 0:21:55.199 saying that the incident actually provided proof that there was 0:21:55.240 --> 0:21:57.760 work needed on the algorithms, and that Google could learn 0:21:57.840 --> 0:22:00.119 from the mistake, and that this was kind of that 0:22:00.240 --> 0:22:04.440 Silicon Valley philosophy of failure is good because you learn 0:22:04.720 --> 0:22:08.760 from failure, okay well. The article from The New Yorker 0:22:08.880 --> 0:22:11.840 also states that there was at least one accident that 0:22:11.920 --> 0:22:15.520 happened while a driverless car was in autonomous mode that 0:22:15.640 --> 0:22:20.119 did not get reported to to the police or to 0:22:20.160 --> 0:22:23.880 the press. According to the article, a car nicknamed Kit 0:22:24.640 --> 0:22:29.160 after good Old Night Writer, was rear ended at an 0:22:29.160 --> 0:22:33.560 intersection when the autonomous car breaked suddenly after being unable 0:22:33.640 --> 0:22:36.080 to differentiate a yellow light from a red light, so 0:22:36.119 --> 0:22:40.200 they stopped short or stopped too suddenly, and the car 0:22:40.240 --> 0:22:43.320 behind them was unable to stop and ran into them. 0:22:43.400 --> 0:22:46.360 As for Lewandowski, he would stay at Google until two 0:22:46.359 --> 0:22:49.439 thousand and sixteen, when he'd leave in order to go 0:22:49.560 --> 0:22:52.480 found a new company called Auto O t t O. 0:22:53.520 --> 0:22:57.320 Auto would consult with Uber for their driverless car program. 0:22:57.680 --> 0:23:01.879 Google executives would allege that Leven Owlski took proprietary information 0:23:01.880 --> 0:23:05.000 and trade secrets with him in this move. They had 0:23:05.040 --> 0:23:09.240 proof that he had downloaded and transferred an enormous number 0:23:09.280 --> 0:23:12.400 of files, though there were questions about how valuable those 0:23:12.440 --> 0:23:17.800 files actually were, and Uber would ultimately fire Lewandowski as 0:23:17.880 --> 0:23:21.760 this case developed. The actual trial happened in early two 0:23:21.760 --> 0:23:24.199 thousand eighteen and it was marked with a lot of messy, 0:23:24.359 --> 0:23:28.560 complicated legal maneuvers on all sides. One day I may 0:23:28.600 --> 0:23:32.240 have to do a full episode just on that lawsuit 0:23:32.400 --> 0:23:34.359 and what came out of it. But we can skip 0:23:34.400 --> 0:23:38.240 to the end. Google's Weymo, because that's what the that's 0:23:38.280 --> 0:23:42.120 what Project Chauffeur evolved into, was a subsidiary company called 0:23:42.160 --> 0:23:45.880 Weymo would have ultimately settled out of court with Uber. 0:23:46.440 --> 0:23:50.280 All right back to driverless cars in the same year 0:23:50.320 --> 0:23:53.640 that Lewandowski would leave Google. We get to that publicly 0:23:53.680 --> 0:23:58.560 acknowledged accident that was the fault of Google's autonomous technology. 0:23:58.880 --> 0:24:02.000 So this was the first time that the public heard 0:24:02.240 --> 0:24:07.000 about a traffic accident that was quote unquote caused by 0:24:07.280 --> 0:24:11.520 Google's driverless car tech. It happened at another intersection, No 0:24:11.640 --> 0:24:15.040 big surprise there. Here's the description from the incident report 0:24:15.240 --> 0:24:18.600 with the California Department of Motor Vehicles and it goes 0:24:18.600 --> 0:24:23.159 like this. A Google Lexus Model autonomous vehicle Google a 0:24:23.320 --> 0:24:27.360 V was traveling in autonomous mode eastbound on El Camino 0:24:27.520 --> 0:24:31.359 Real in a real I guess in mountain view in 0:24:31.440 --> 0:24:35.920 the far right hand lane, approaching the Castro Street intersection. 0:24:36.600 --> 0:24:39.880 As the Google A V approached the intersection, it signaled 0:24:39.920 --> 0:24:42.440 its intent to make a right turn on the Red 0:24:42.720 --> 0:24:45.399 onto Castro Street. The Google a V then moved to 0:24:45.440 --> 0:24:48.120 the right hand side of the lane to pass traffic 0:24:48.200 --> 0:24:51.000 in the same lane that was stopped at the intersection 0:24:51.240 --> 0:24:54.760 and proceeding straight. However, the Google a V had to 0:24:54.840 --> 0:24:58.080 come to a stop and go around sandbags positioned around 0:24:58.080 --> 0:25:01.640 a storm drain that was blocking path. When the light 0:25:01.720 --> 0:25:04.959 turned green, traffic in the lane continued past the Google 0:25:05.040 --> 0:25:07.959 a V. After a few cars had passed, the Google 0:25:08.000 --> 0:25:11.000 a V began to proceed back into the center of 0:25:11.040 --> 0:25:14.840 the lane to pass the sandbags. A public transit bus 0:25:15.080 --> 0:25:18.440 was approaching from behind the Google a V. Test drivers 0:25:18.560 --> 0:25:21.760 saw the bus approaching in the left side mirror, but 0:25:21.880 --> 0:25:24.960 believed the bus would stop or slow to allow the 0:25:24.960 --> 0:25:29.240 Google a V to continue. Approximately three seconds later, as 0:25:29.240 --> 0:25:31.639 the Google a V was re entering the center of 0:25:31.720 --> 0:25:34.920 the lane, it made contact with the side of the bus. 0:25:35.280 --> 0:25:38.159 The Google a V was operating in autonomous mode and 0:25:38.240 --> 0:25:40.679 traveling at less than two miles per hour, and the 0:25:40.720 --> 0:25:43.199 bus was traveling at about fifteen miles per hour at 0:25:43.240 --> 0:25:46.440 the time of contact. The Google a V sustained body 0:25:46.520 --> 0:25:49.399