WEBVTT - Self-Driving Bus Accident 0:00:03.600 --> 0:00:06.400 A self driving shuttle bus in Las Vegas operated for 0:00:06.440 --> 0:00:10.320 almost two whole hours before being involved in a minor accident. 0:00:10.720 --> 0:00:15.360 Are self driving vehicles really the safest option? Yes, they are. 0:00:16.160 --> 0:00:23.439 I'm Jonathan Strickland and this is text Updaily. Let's cover 0:00:23.480 --> 0:00:27.120 the news first. A French company called Navia developed an 0:00:27.120 --> 0:00:31.600 autonomous electric vehicle designed to travel along predetermined routes. The 0:00:31.640 --> 0:00:34.000 city of Las Vegas had such a shuttle to carry 0:00:34.000 --> 0:00:37.680 passengers around the Fremont East district. The shuttle itself has 0:00:37.680 --> 0:00:40.200 a very modest route at just over a half mile. 0:00:40.760 --> 0:00:45.600 It's really just a convenience vehicle. In early November, after 0:00:45.640 --> 0:00:49.199 being an operation for about two hours, one such shuttle 0:00:49.240 --> 0:00:52.920 collided with a truck, or rather, it's more appropriate to 0:00:52.960 --> 0:00:57.040 say a human operated truck backed into the shuttle. According 0:00:57.080 --> 0:01:00.080 to a Triple A representative, the accident was due to 0:01:00.200 --> 0:01:03.040 human error, not the shuttle. Triple A, by the way, 0:01:03.160 --> 0:01:06.800 is a sponsor of this driverless shuttle pilot program. A 0:01:06.880 --> 0:01:09.080 representative of the city of Las Vegas gave a bit 0:01:09.160 --> 0:01:11.759 more information on the official Tumblr page for the Las 0:01:11.880 --> 0:01:15.200 Vegas City government. It read the shuttle did what it 0:01:15.240 --> 0:01:18.479 was supposed to do and that its sensors registered the truck, 0:01:18.800 --> 0:01:22.679 and the shuttle stopped to avoid the accident. Unfortunately, the 0:01:22.760 --> 0:01:25.800 delivery truck did not stop and graze the front fender 0:01:25.920 --> 0:01:28.919 of the shuttle. While the headlines may make it sound 0:01:28.959 --> 0:01:32.800 like driverless vehicles are unreliable and unsafe, a closer look 0:01:32.840 --> 0:01:36.920 at stories shows that that's not really the case. Google's 0:01:37.000 --> 0:01:40.319 self driving vehicles have been in operation for a few years, 0:01:40.440 --> 0:01:44.280 running tests quietly in northern California, and while the cars 0:01:44.319 --> 0:01:47.560 have been involved in several incidents, nearly all of those 0:01:47.600 --> 0:01:50.440 were due to human error. In some cases, the Google 0:01:50.520 --> 0:01:53.360 vehicle was in manual operation mode at the time, meaning 0:01:53.360 --> 0:01:55.640 a human driver was at the controls and made a 0:01:55.680 --> 0:01:59.600 mistake resulting in an accident. In other cases, drivers of 0:01:59.640 --> 0:02:02.160 other vehicles failed to follow the rules of the road 0:02:02.280 --> 0:02:05.840 and caused an accident. In fact, before a Google spun 0:02:05.920 --> 0:02:08.920 out its self driving car division into a new company 0:02:08.960 --> 0:02:13.079 called Waymo, the list of incidents only contained one case 0:02:13.160 --> 0:02:15.680 in which the company determined the self driving system was 0:02:15.720 --> 0:02:19.000 at fault. In that incident, a Google car was navigating 0:02:19.040 --> 0:02:22.320 a street in Mountain View, California, and approaching an intersection. 0:02:22.720 --> 0:02:24.919 The car sinced an obstacle in the road. It was 0:02:24.960 --> 0:02:27.959 a pile of sandbags around a storm drain. The car 0:02:28.040 --> 0:02:30.959 moved slightly out of its own lane to pass the obstacle, 0:02:31.080 --> 0:02:33.040 then attempted to get back into the lane it was 0:02:33.080 --> 0:02:35.880 supposed to be in. The only problem was that a 0:02:36.000 --> 0:02:39.400 bus was approaching from behind and started to fill that space. 0:02:39.720 --> 0:02:42.880 The self driving cars AI assumed the bus driver would 0:02:42.880 --> 0:02:46.000 slow or yield to the self driving vehicle, and it 0:02:46.040 --> 0:02:49.200 began to merge back into its lane. Then the self 0:02:49.280 --> 0:02:52.840 driving car learned the valuable lesson that two objects cannot 0:02:52.880 --> 0:02:55.960 coexist in the same physical space at the same time. 0:02:56.240 --> 0:02:59.040 There was a low speed collision. Google admitted the car 0:02:59.160 --> 0:03:01.400 was at fault for making a baseless assumption that the 0:03:01.480 --> 0:03:05.120 human driver would yield, And yes, the apology did sound 0:03:05.160 --> 0:03:07.360 a bit like Google was throwing a little shade at 0:03:07.360 --> 0:03:10.799 the bus driver, but only a little. If you look 0:03:10.800 --> 0:03:15.200 at statistics, self driving cars are extremely safe, particularly if 0:03:15.240 --> 0:03:17.720 you're looking only for the times they are at fault 0:03:17.800 --> 0:03:20.960 in an accident. Google's cars have driven more than two 0:03:20.960 --> 0:03:24.239 million miles on streets in the United States with only 0:03:24.360 --> 0:03:29.079 one at fault accident. That's significantly lower than for human drivers. 0:03:29.480 --> 0:03:32.560 New drivers are forty times more likely to get into 0:03:32.600 --> 0:03:36.720 an accident than a self driving vehicle is. However, self 0:03:36.800 --> 0:03:39.400 driving cars do tend to get into accidents caused by 0:03:39.480 --> 0:03:41.960 human drivers at a rate higher than you'd see in 0:03:42.040 --> 0:03:45.560 an experienced human driver. Self driving cars might be a 0:03:45.560 --> 0:03:49.400 bit too polite and safe, avoiding aggressive tactics that can 0:03:49.480 --> 0:03:53.480 sometimes prevent an accident. Maybe self driving cars place more 0:03:53.520 --> 0:03:55.680 faith in humans to do the right thing when they 0:03:55.800 --> 0:04:00.440 drive foolish robots that will be their downfall to be 0:04:00.520 --> 0:04:03.320 serious for a moment. Self driving cars have the potential 0:04:03.400 --> 0:04:07.200 to make an enormous positive impact. For the year sixteen 0:04:07.240 --> 0:04:10.160 in the United States, the National Safety Council estimated that 0:04:10.280 --> 0:04:14.120 forty thousand people died in motor vehicle crashes. From a 0:04:14.160 --> 0:04:17.760 statistical standpoint, self driving cars have the potential to prevent 0:04:17.880 --> 0:04:21.560 thousands of deaths every year. On a lesser note, they 0:04:21.560 --> 0:04:25.000 could also reduce property damage incidents and reduce strain placed 0:04:25.000 --> 0:04:28.400 on emergency rooms and first responders. They could also help 0:04:28.440 --> 0:04:32.040 alleviate traffic snarls and reduce commute times in dense cities. 0:04:32.560 --> 0:04:36.200 While several companies, including Waymo, are aggressive in getting self 0:04:36.279 --> 0:04:39.440 driving cars out on the streets and more real world scenarios, 0:04:39.720 --> 0:04:42.679 we also have to face some facts. The National Highway 0:04:42.720 --> 0:04:47.599 Traffic Safety Administration recognizes six levels of autonomy, from zero 0:04:47.680 --> 0:04:50.200 to five. At zero, you have a car that is 0:04:50.240 --> 0:04:54.960 completely human operated. Levels one and two cover driver assist systems, 0:04:55.000 --> 0:04:57.919 in which an automated feature might kick into gear when needed, 0:04:58.120 --> 0:05:00.279 but the car is still largely under the control role 0:05:00.320 --> 0:05:03.520 of a human driver. A Level three system is largely 0:05:03.560 --> 0:05:06.480 automated but can allow for a human to take control 0:05:06.520 --> 0:05:10.359 in a safety critical situation. This one is tricky because 0:05:10.440 --> 0:05:13.080 handing off the control of a vehicle isn't easy to 0:05:13.160 --> 0:05:15.960 do in a seamless way. It's sort of like having 0:05:16.000 --> 0:05:18.440 a passenger reach over and grab the steering wheel to 0:05:18.520 --> 0:05:20.880 yanket to the side in order to avoid a crash. 0:05:21.279 --> 0:05:24.160 The passenger and driver might find themselves struggling against each 0:05:24.160 --> 0:05:27.400 other for a moment. Level four is a fully autonomous 0:05:27.480 --> 0:05:31.360 vehicle that can operate in an operational design domain or 0:05:31.480 --> 0:05:34.719 o d D. That means the vehicle has restrictions on 0:05:34.760 --> 0:05:38.040 where it goes and under what conditions it operates. This 0:05:38.200 --> 0:05:40.159 is the level the shuttle bus I mentioned at the 0:05:40.160 --> 0:05:43.080 top of the show is supposed to inhabit. The vehicle 0:05:43.160 --> 0:05:46.560 should be capable of handling all situations that fall inside 0:05:46.600 --> 0:05:49.599 that O d D. You could argue the shuttle failed 0:05:49.640 --> 0:05:51.960 to do this and that it didn't find some means 0:05:51.960 --> 0:05:55.160 of preventing the fender bender, though without all the details 0:05:55.200 --> 0:05:58.520 that's hard to say. Level five is a fully autonomous 0:05:58.560 --> 0:06:01.360 system that can work under any dry driving scenario. That 0:06:01.400 --> 0:06:03.480 would be a car that could adapt to conditions the 0:06:03.520 --> 0:06:07.520 way human drivers can, no matter whether traffic or road conditions. 0:06:07.960 --> 0:06:10.159 Most experts in the fields say we are still a 0:06:10.240 --> 0:06:13.240 far away away from this, as teaching machines how to 0:06:13.279 --> 0:06:17.120 recognize a threat versus something that is ultimately harmless is tricky. 0:06:17.560 --> 0:06:21.600 Humans can extrapolate based upon experience. Machines have trouble with that. 0:06:22.080 --> 0:06:24.440 All that being said, I'd still hop into a self 0:06:24.520 --> 0:06:27.599 driving shuttle without hesitation, assuming I was confident in the 0:06:27.640 --> 0:06:30.919 abilities of the company that made the sensors and software. I, 0:06:31.080 --> 0:06:34.560 for one, welcome our robot chauffeur Overlords. To learn more 0:06:34.560 --> 0:06:38.520 about autonomous vehicles and all other things tech related, subscribe 0:06:38.520 --> 0:06:42.240 to the tech Stuff podcast It's robot Approved. See you 0:06:42.279 --> 0:06:42.960 again soon.