 Okay, welcome. My name is Robert Hampshire. I'm an associate professor here at the Ford School of Public Policy. I also have affiliations with the University of Michigan Transportation Research Institute named Umtree. So welcome today. It's my pleasure to introduce our speaker and our lecture today on the opportunities and challenges of autonomous vehicles, the roles for government. This is sponsored by the Center for Local, State, and Urban Policy and the Ford School's Public Policy Talk Series. Also, I'd like to thank our co-sponsors today, which are the Science and Technology and Public Policy Program and the University of Michigan Transportation Research Institute. So thank you for all the sponsors and co-sponsors. Before I introduce Christopher Hart, let me just say a few words about today's format. We'll take some questions at the end from the audience, and so you'll see people who have index cards. Please write your questions on those index cards, and we'll start to collect those around 4.30. And so towards the end of the talk, we'll then ask those questions. So today we have some Ford School students who are going to help ask the questions. So Avril Prakash and Eric Hans will help ask the questions. Also with support from Kristi Richardson. So thank you. So let me introduce our speaker today, Christopher Hart, who is the former chairman of the National Transportation Safety Board. He has the honor, I think, of being appointed by two U.S. presidents, by different parties, George Bush Sr., and also President Obama, to serve on that board. Chris is an engineer, but also a lawyer who worked at the Federal Aviation Administration for years at NHTSA for years and in private practice. I had the pleasure of serving, being on a panel with him last year at Princeton Reunions. We both had Princeton background, so that was really a great time. So I figured that the Ford School community could really benefit from hearing his sage advice based on his years of experience in his career. So at this point, I'd like to just introduce Christopher Hart and take it away. Thank you. Good afternoon, everyone, and thank you very much, Robert, for that very kind introduction. In case none of you ask any questions, I planted some with him to make it look like I know what I'm talking about. So I appreciate that. So this is quite an honor to be here and have a chance to talk to you guys about something that's going to be very transformative to everything we do, and that is autonomous vehicles. And the specific question is, what is the role of the governments in this equation? And in order to do that, I have to kind of lay the groundwork to some of the benefits and challenges of autonomous vehicles, which I will do. So thank you for, again, the opportunity to be here. It's a privilege and an honor. And if Robert has anything to say about it, hopefully I'm going to say something that you guys enjoy hearing that will be productive to you. So I'm going to talk about the opportunities. I'm going to talk about the aviation mode that's most advanced in automation, and that's aviation and lessons learned from them. There are quite a few other lessons about automation that the car makers are going to have to address because automation on the ground is much, much more challenging and complex than automation in the air. So there's going to be a lot of other challenges for the car community that weren't faced and that won't be faced in aviation. And then I'm going to end it up with the roles of the governments. So the opportunities that automation has had has been a huge success story everywhere that we've seen it at the National Transportation Safety Board. By the way, let me just give a little background about what they do. The NTSB is the Accident Investigating Agency for the federal government. So they investigate accidents in all modes of transportation and come up with findings and recommendations, and the recommendations are intended to help prevent recurrences of those accidents. And let me just emphasize they are just recommendations. The NTSB can't require anybody to do anything, but contrary to what the media would have you say, the media would have you think, well, when I received this letter from the NTSB that says, we recommend such and such, I emphasize where I'd recommend, and so I don't have to do it, File 13. Well, that doesn't happen, and the media would have you think that happens a lot, but more than 80% of the time the recommendations of the NTSB are followed, because A, they're good ideas, and B, if you don't follow that recommendation and then you have an accident because you didn't, then you really have some explaining to do to the jury after that. So that's really the main leverage for the NTSB, and why more than 80% of the time, even though people don't have to, they still do the recommendation or something like it to help make transportation safer. So that's what the NTSB does, and they do it in all modes. What really got me interested in the highway mode, because the NTSB didn't do historically that many highway accidents, and most of them that we did were big vehicles, buses, and trucks with commercial professional drivers, as opposed to Joe Public with his car. And this one, the one that I'm going to talk about, which is the Tesla accident in Williston, Florida in 2016, was the first one that I had seen in my NTSB career that involved Joe Public and his car. And we looked at that because of the automation impacts, because we've been dealing with automation for years and years, especially in aviation, but also in other modes of transportation. So in aviation, automation has shown amazing safety benefits, productivity, and operating efficiency, getting more airplanes through the airspace, and reducing pollution all at the same time. And I think that the car automation scenario has even more opportunities for improvement, especially in safety and many others. So the question is how do we get from here to there, because I'm a gadget geek. From what I say, you may think that I'm anti-automation, but I'm actually very pro-automation. I'm a gadget geek. The question is how do we get there without hurting too many people in the process, because the public is already skeptical of the whole notion of automation, including me, and every time they see an accident, I have to wonder, wow, yet again, do we see the car people not paying attention to the aviation automation mistakes that have occurred over the decades and not learning from those mistakes and making the same mistakes over again, which is a real shame, because then it makes that skeptical public even more skeptical. So how do we get from here to there, because I'm convinced that eventually we will be there. I have to say that the estimations of it were wildly over-optimistic, because it wasn't that many years ago when you saw estimates that by 2020, the streets will be full of driverless cars. Well, folks, it's 2020, and guess what? So that's what I'm going to talk about is, ultimately, what is the role of the various governments, federal, state, and local governments in helping to make this happen? So let's look at some of the automation lessons learned in aviation. First of all, is the importance of human-centric automation. So when they first started automating an aviation, they said, here we have the technology to do it, let's do it, and that led to some bad results a few times. So then that's when they realized, maybe we ought to focus it on human-centric automation instead of trying to automate to get rid of the human. Let's call it human-centric automation, which means developing the automation around the human. So when automation isn't perfect, you need to have, in cars, you need to have some grateful exits. First of all, you need to have a graceful exit if the driver is not paying enough attention. And the first, the Williston-Florida Accent, I'm going to talk about, is one where the driver wasn't paying enough attention. But in addition to that, you've got two big what-ifs in any automation endeavor. The first what-if is what if it fails, and it's not doing that very much anymore. The automation is very reliable, so failure is not that common. But number two is what if it encounters unanticipated circumstances, circumstances that weren't anticipated even by the designers of the automation. Those are, that's when you need graceful exits if that happens. And that's why when people ask me how long before I get on an airliner with no pilot, well, no time soon because until they figure out those what-ifs, what if the automation fails, what if the automation encounters unanticipated circumstances, i.e. sully, what if that airplane had been automated, then until that happens, we're not going to see airplanes without pilots. And then last but not least, another lesson from aviation automation is when you have a system that's very reliable and you ask a human to be a monitor for that system, it doesn't work very well. And I'll show you an accident where that happens. So human-centric, as I said, aviation began with automation because we have, automate because we have the technology to do it. And they realize, oops, that doesn't necessarily produce the desired results. So now they're doing human-centric and the car makers are trying to do human-centric, but the ongoing crashes reveal they need to do better human-centric. Here are some examples of not being human-centric enough. To me, the word autopilot is a huge, huge mistake. Because I'm a pilot, and I know that when the autopilot is on in the airplane, I'm still important. So I'm not going to just go back and fall asleep and say, autopilot, you do your thing. I'm still important. But Joe and Suzy Public don't necessarily know that. When they hear the word autopilot, that's why most of the car makers, to their benefit, use something like driver assist versus autopilot, gives the general impression, okay, autopilot will do this thing. I'm not necessary here. And that creates disengagement by the driver. In addition to that, when the systems are very good, then if your system is really good, you're going to encourage the driver to disengage. So example is, I recently got a Volvo, and my Volvo has lane keeping assist. And so the lane keeping assist that doesn't have very high gains is kind of a little bit sloppy, so it'll drift to one side, and hit that line, hit the other side, and then by the time you do that a few times, that's kind of annoying. So you hold the steering wheel just so you don't do the back and forth. Tesla, on the other hand, has very tight gains, and if you don't do the back and forth, you do very tight in the middle of the lane, and that encourages the driver to disengage. And to me, that's not a good idea if you want the driver to stay engaged. And so that, to me, that's another failure to learn from aviation automation. So another one is, what's the graceful exit if the driver is not adequately attentive? The two graceful exits I've seen are, A, turn the automation off, B, stop the car in the road. Well, I'm not satisfied with either one of those because turning the automation off, whether the person has a medical event and you turn the automation off, now you're asking for a crash. And stopping the car on the road, I can't tell you how many crashes we've looked at at the NTSB because stoppage on an interstate, usually because of construction. So there's a construction queue. The one that you all might remember was the one on the Jersey Turnpike where there was a construction queue stopped and the Walmart truck ran into the construction queue and killed Tracy Morgan's buddy and basically entered Tracy Morgan. So that was one where usually when you have construction queue, you have like two miles of signs, warning, be prepared to stop construction ahead, et cetera, et cetera. If the car just stops without the two miles and still you had the truck coming into this queue at highway speed. If you don't have those two signs, what are we going to see for people stopping on the road because most of these systems in the cars today say don't use them except on interstate highways. Well, if you use it on interstate highway in the middle of the interstate highway, that's pretty much asking for an accent. So to me, we're not ready because we don't have a graceful accent for driver inattention. Here are some other human-centric realities. If the automation is not adequately user-friendly and much of it isn't, then it could be used improperly or worse, it could distract the driver. What is that chime that I'm hearing now and you're looking around, looking around for the chime crash, then you run into something because you weren't paying attention to what's in front of you because you're trying to address the chime or what if it just turned off and now I lose the benefits of the automation. So none of those are desirable, desirable outcomes. What about automation being uncertain because the lane markings after the sand and salt on the lane markings, that's going to wear them down, so what happens when the lane markings become not so good or in my car, again with the lane keeping as this, when you come to the exit then the right line now diverges from the traffic and the car sort of starts to follow that until it realizes, oops, I don't want to do that, and so that's just another example of what about uncertainty because of the lane markings, what about uncertainty because of the street are slippery like they are right now, what's the story on will the car affect its stopping distance algorithm because it knows that the streets are slippery, what about unanticipated circumstances like an object in the road, so I'll be talking about one of the crashes, Tempe, Arizona, where that happened, and what if the system is uncertain, does it warn the driver in enough time for the driver to take over and handle the situation. So here's Williston, Florida that I talked about, this was where a tractor trailer, opposite direction tractor trailer was turning left at an intersection on a high speed road like 55 or 60 mile an hour road, but it was not an interstate it had occasional intersection, so here's this truck turning left on this road and the Tesla submarine under the trailer, and of course it decapitated the driver by sharing off the top of the car, 74 miles an hour with no skid marks. So this is one where the driver had been bragging in social media about, oh yeah, I got my autopilot, so while it's driving I'm playing Sudoku or whatever it's doing, so he's bragging about how powerful this autopilot is, well so the owner's manual Tesla said, oh that shouldn't have happened because the owner's manual says only use it on highways and limited access roads, well two problems of that, one is I think this was a highway so to me, the person was following the owner's manual, but more fundamentally number two, who reads the owner's manual? I read the owner's manual twice a year, daylight saving time starts, daylight saving time stops how do I change this clock? Well these days you don't even have to do that because now the clock changes itself, so it's so much for the twice times, two times a year when I look at the owner's manual, but who looks at the owner's manual? So the point is the Tesla was over relied on the driver looking at the owner's manual, and the driver was over relied on Tesla's ability to develop automation that didn't need and so as you can see it barely touched the truck, but meanwhile the driver was killed by shearing off the top of the car. Here's another one and this happened to me, not in an automated car, but just in a regular car where I'm behind a car, then the car suddenly moves out of the lane and I realized it moved out of the lane because there was an obstacle in front of that car and I'm thinking wow, I was just lucky that I could move out of the lane on short notice because if somebody had been beside me, I would have probably run into that object that was in front of it, so that's what happened was the Tesla was behind a car in the left lane, there was a fire truck in the left lane, dealing with an accident, the car moves out of the left lane to get around the fire truck, the Tesla continues on and crashes into the fire truck, so this is one, I think maybe they, I may have to update this slide, I think the NTSB may have completed the investigation of this one I'll have to look that up and see, but the point is this was one where once again the autopilot was not up to the task and you need the driver to do the job. Here's one where when I was looking for a situation where the automation actually failed and like I say, that's very rare because the systems are very reliable these days I had to look to, I couldn't find any in aviation, I had to look to the subway system in my town, DC to find a situation with automation actual failure, so this was one where the trains are registered on the system electronically and occasionally for reasons we didn't know before this accident, the electronic signature disappears and when the electronic signature disappears, we know in the main dispatch board that shows where all the trains are when the electronic signature of one of the trains disappears they get a warning at the dispatch board and then we asked them what do you do with that warning at the dispatch board and they said nothing because we get that warning 500 times a day so that's not very useful, guess who didn't get the warning, who didn't get that warning was the train behind the train that stopped because this train that disappeared had stopped and the train behind it didn't get the warning so the train behind is saying, okay track ahead is empty, we'll accelerate to full speed so that's what the train was doing, when it rounded this curve you can see from this picture that the crash occurred on a curve, that's what limited the sight distance of the operator, soon as she saw this train stopped on the curve, hit the emergency brake, but it was too late and it killed her and a passenger so if the automation fails and doesn't warn the operator in time, you got a huge problem and that's one of the issues for car automation, when it's uncertain or fails, does it warn the driver in time to let the driver be able to salvage the situation, then there's what I say happens much much more often, I could give you 15 or 20 aviation accidents involving unanticipated circumstances but this one is kind of the sort of the really example of how even aviation after automation for decades doesn't have it totally figured out so this was Air France from Rio to Paris and as most flights are from North and South America to Europe it was a red-eye flight at night so they're taking off, they go up to 35,000 feet at cruise altitude at night and they're near thunderstorms and as soon as they reach cruise altitude the chief pilot, who is the most experienced pilot on board, goes back to take a scheduled rest break so question number one is, was that a good time to take a scheduled rest break when you saw on the radar that there were thunderstorms 80 to 100 miles in front of you so but yet the pilot chief pilot took a rest break so here's all this super cool water out there because of all these thunderstorms out there and the airplane has what's called a pitot tube on it which projects from the fuselage and sticks forward and it has a hole in the front and the purpose of the hole in the front is that the air goes into the hole and the airplane knows how fast it's going by how hard the air is coming into that hole so that means you want to make sure that the hole doesn't freeze over because if it freezes over then no air comes into the hole and the airplane has no clue how fast it's going and in order to prevent that from happening you have heaters on the pitot tubes ok well a couple of times this particular type of airplane had the heaters be overwhelmed by the super cool water and it froze over the pitot tubes and the airplane didn't know how fast it was going but the pilots pulled out of this situation successfully so they figured ok number one we probably need better heaters on the pitot tubes so we'll do that number two these other pilots pulled through it ok so it's probably not an emergency so we won't we won't schedule an emergency we'll just do it the next time this airplane has regularly scheduled maintenance which this airplane was scheduled to do in the next two or three weeks so now when the ice blocks the pitot tubes and the airplane doesn't know how fast it's going there are lots of systems that don't operate if they don't know how fast the airplane is going for example the automatic pilot won't operate the automatic throttle that maintains your speed it won't operate the protection against the aerodynamic stall it won't operate without knowing how fast you're going so there are lots of systems so what these pilots would have gotten would be all of a sudden a bunch of error messages that you've lost your airspeed information but they would also have six or eight error messages of these six or eight systems that quit because they don't have the airspeed information and they won't work without it and so the pilots responded inappropriately and ended up crashing so in that one we looked at the totality of the circumstances as we always do when I say we not the NTSB because we weren't in the charge of this accident the French variation of the NTSB was in charge of it but did the pilots ever have training about what happens if you lose airspeed information and cruise no that would be something you could easily simulate in the simulator but they never had training about that did the pilots ever have any training about flying manually at cruise altitude most of the time they don't fly manually at cruise altitude most of the time it's illegal to fly manually at cruise altitude because the way that worked was back in the old days they used to have 2,000 feet of vertical separation between opposite direction traffic then when they needed more highways in the sky so to speak they reduced that to 1,000 feet of vertical separation and the designer of that system said hmm 1,000 feet that's not really very far away there's too much of an opportunity of a mid-air collision because 1,000 feet just I don't trust humans to do that successfully so therefore I'm going to require that you be on automation at cruise altitude so in the mainland United States anything above 29,000 feet you have to be on automation so that's why these guys had never had training about how to fly manually at cruise altitude and believe me it's a very different airplane in the very thin air of 35,000 feet than it is when you're taking off and landing but yet these guys had never done it before had they ever recovered from unusual attitudes from a stall at cruise altitude definitely not they didn't even know how to fly manually at cruise altitude so they hadn't had that and plus they weren't talking to each other so because they weren't talking to each other as soon as the airplane lost the airspeed information the right side pilot yanks back his side stick all the way to the aft stop which causes the airplane to start to climb but he didn't talk to the left side pilot to tell the left side pilot what he was doing and another link in this chain is if this had been a Boeing airplane and I'm not slamming airbus I'm just saying this is a difference between them if this had been a Boeing airplane where you have a big control yoke between your legs control yoke back to his belly then it would be in my belly too because the control yokes are connected but when he moved the side stick to the left side pilot the side stick didn't move so the left side pilot had no clue that that's what the right side pilot had done was move the side stick all the way back to the aft stop and they weren't talking that's crew resource management they weren't talking to each other to make sure they're on the same page bottom line they responded inappropriately and crashed so this was just a poster child example of the the whole system not having their hands fully around automation so those are some of the challenges that the car industry is going to have to pay attention to that aviation has faced there's a whole bunch of others in the interest of time to save time for answering questions I won't go over these in great detail but I am going to go over each one of them just a few words artificial intelligence learns with experience drivers not trained street testing essential I'm going to go over each one of these in turn to talk about how these are additional problems because like I said automation on the ground is much much more challenging than automation in the air and after I go over all these challenges then I'll talk about what roles are the government going to play in all this situation so automation that learns doesn't happen in aviation because whenever the automation changes significantly in aviation you have to retrain the pilots to the new automation and that's one of the questions that's come up in this 37 737 max situation which I'm also working on so in aviation if the automation changes enough they have to retrain the pilots and so once the automation is in place it's locked down and that's what they train the pilots to and the pilots do that until the automation is changed again and then it's locked down so there's no artificial intelligence there's no learning in process while the automation is working so that's going to to have that automation that changes so that when you come out to your car the next morning and then you wonder and then you get the word that there was a software upgrade then you wonder gee I wonder if I'm going to be able to handle this car the way it behaves now versus the way it behaved yesterday when I left it and how much training is there of drivers zero so there's not going to be any training of drivers so that's going to be a huge challenge and that means that when the automakers change the automation they have to assume two things one is that there won't be any training because that is never going to happen number two is that the owner won't read the owner's manual and of course this change from last night wouldn't be in your owner's manual anyway but the point is you have to assume the worst case that how driver friendly is this for the driver to respond to this new change in automation so that's an example of an issue they're going to face on the ground that they don't face in the air so like I said training airline pilots train and retrain and retrain regularly recurrently doesn't happen in cars so that means designers of automation have to assume worst case drivers won't have any training and drivers won't look at the owner's manual so the possible outcome from untrained drivers is a you turn the automation off and you lose the protections and be like I said you can become distracted and because you come to become distracted and what's that chime and what do I do about it then you have a crash just because you're distracted so the question for the auto dealers and I gave this presentation to the auto dealers once and they there were a lot of eyebrows raised and I said today the dealer just hand you the key is this going to change that liability scenario so they not have to not only give you the key but also train you to some extent about how to use the car so I'm not just talking about point you to a website I'm talking about actually train you to some extent about how the car works so stay tuned on whether that's going to change in terms of the role of the auto dealers in this equation how many people in the auto showroom do you know of that are substantively familiar enough with the car to train you so that would be a whole different experience going to an auto dealer and have someone who's trained enough to tell you about how it works this is one that you'll be interested in because you got the M-City which is your test track but what we're finding big time is street testing actual street testing is essential no matter how much test track testing you do and no matter how much lab testing you do actual driving on the streets is going to be essential to make a car street ready so that's going to be a big challenge and the reason that's a big challenge is because it raises the essential conundrum that assuming that you're a responsible enough manufacturer that you do a whole lot of test track training a whole lot of lab training before you ever put your car on the streets you got a very reliable car and now the conundrum you run into is what I said about aviation learn that humans are not good monitors of reliable systems and that's what happened in Timtee so improved system design meaning a warning about the uncertainty the automation if it's not sure what's going on it needs to warn you in a timely manner and also it needs better monitor training to tell the monitors guess what this is a scenario that humans don't do very well namely monitoring reliable systems and here's how we're going to train you to respond to that reality that what you're doing is not something that humans normally do very well so Timtee Arizona was a driverless street test with a monitor and you've seen the YouTube videos where the monitor was looking down at that moment so the woman was walking across the street with her bike at night not in a crosswalk and the NTSB this may take some more updating too because I think the NTSB investigation of this one might be completed as well but this was the first fatality of a pedestrian from a quote driverless vehicle so this was one where again the driver wasn't paying adequate attention but look at the way this works it doesn't matter what the driver was looking at the driver could be distracted by a whole bunch of things many of which are legitimate distractions like I could be looking at the street side or I could be looking at the rear view mirror to change lanes in fact I did have an accident once where I was looking in the rear view mirror to change lanes and the guy in front of me stopped just at that moment crash so there's no way that even with a driver that's not distracted by things they shouldn't be distracted by like their ipad or their iphone or whatever there's still no guarantee that a driver is going to be 100% of the time looking out and that's why this accident was pretty much inevitable with all the street testing that's going on that accident was pretty much inevitable and just as a matter of full disclosure I was engaged by the law firm that was asked after this crash Uber stopped voluntarily street testing and they engaged a law firm to get them back on the streets and the law firm asked me to help them so I was engaged in this one and they have started back on the streets I don't know if they've gone to any other cities yet but they did start back on the streets but the point was that this is an example where it was inevitable with all this testing going on inevitable there was just going to be that moment in time when the person was not looking out just at the moment when they needed to be and by the way there was an infrastructure issue on this one because the sidewalk that the pedestrian was on didn't have it stopped at the street and it didn't have a crosswalk so this woman was walking her bike on the sidewalk and then crossed the street on something but that's neither here nor there because the point is they still shouldn't have hit her and so there was a lot of criticism about well jeez how come Uber turned off the Volvo driver alert system to keep collisions from happening and the reason they did that was because they wanted to invent a system that would work on any car not just a system that would work on top of the Volvo system so that's why they disabled intentionally disabled the Volvo system to detect objects and stop in time so this is just another example of where as we try to figure out how to do this better there's going to be some challenges and people are this is not with wrong intent this is people trying to do what they think is the right thing to do in that moment but this is going to take again better warning of uncertainty and better training of the monitors to say we're asking you to do what we know humans don't do very well which is monitor reliable very reliable systems graceful exits as I mentioned you got to have graceful exit if the automation fails or if it encounters unanticipated circumstances and that's important when you have drivers in the car but if you take the driver out of the car then it's obviously crucial it's not just important it's essential graceful exit this is what I can say when people ask me how long before I get on an airliner with no pilot the joke about the airliner of future is there'll be you look up front there'll be a pilot and a dog and the pilot's job is to feed the dog the dog's job is to bite the pilot if he touches anything so that's the airplane of the future well here's one where both engines ingested birds pilots were unable to reach the airport so they landed in the river and what would automation have done in that situation in fact this is when I talk to aeronautical engineering crowds I tell them there was actually automation kind of undermined this situation because this was one where Sully tried to do as soft a landing as possible by flaring just before he hit the water to minimize the vertical impact speed of the airplane into the water the automation that was on at the time which is called a Fugoi damper those of you who are aeronautical engineers know what a Fugoi is Fugoi damper kept him from getting from pulling the nose up as much as he wanted to stopped him from pulling the last three and a half degrees of nose up which caused him to hit the water much harder than he had anticipated which then breached the bottom of the fuselage which is where the water came into the fuselage and also forced the vertical member up through the floor into the leg of the flight attendant who was sitting on top of it so so question was if that automation had not stopped him from doing that by the way this is automation he had no clue of and he was very upset when he found out from our report that yeah there was automation that kept you from getting from flaring as softly as you wanted to he was very upset as any professional would be because of any professional wants to know what does the airplane do what what are the capabilities of this automation what does it do and what does it not do he wasn't told about this Fugoi damper in his airplane so he hit the water much harder than than he intended to so that was one where automation actually undermined the outcome that he wanted mixing driverless with humans again very challenging because of the lack of variability of the humans out there and so even if you design a completely driverless car you still have to consider the other humans in the system like the pedestrians and the bicyclists and and the motorcyclists and other people who you won't take this steering wheel out of my cold dead hand who love driving so there's going to be a lot of other humans out there and that's it's much more challenging to mix automation with humans than it is just to have straight automation software updates that's what I said about are you going to be able to handle the difference of behavior in your car when you come out the next morning and not only that but look at how thoroughly complex software is tested for unintended circumstances sometimes I wonder with the frequency of the software updates how thorough is that testing to see if there are any potential unintended consequences from adding this new software to the existing system I have a feeling there's not very adequate testing of that before they bring these new systems into place aviation system designers figure out when they change the software they figured out normally by bringing pilots into fly it they don't just rely on their human factors experts they bring pilots into fly it in the engineering simulator before it actually goes out into service that that's not so easy in cars because the simulators aren't nearly as good and there's a much greater variability of drivers that you would have to subject this test to so that's going to be much more challenging in cars than it is with airplanes cyber protection that one really worries me because I just don't see much about it with manufacturers I see a lot of academia interested in it like here but wow that's kind of scary because not only do you have to be protected against today's cyber invasion protocols but today's cyber invasion protocols are ever advancing so it's got to be a continuous thing it's not a one time thing because then you have to get better against protocols as well and I'm just very concerned that I don't see how well that's happening and not only that but every time you add additional protections to your software to protect it from cyber attacks then you got to again just with any other software change you got to look at the unintended consequences of the software change so I'm not satisfied I'm very concerned that lack that inadequate attention is being paid to the whole area of cyber protection this is an interesting one competition regarding safety airlines don't compete on safety and that's they have a cultural aspect in that industry that not very many industries have and that is anybody's crash is everybody's crash so when public sees an airplane crash they don't say oh that airplane was airline X and I'm not worried because I'm going on airline Y quite the contrary they say wait a minute that was a German airplane that crashed in the French Alps that was a suicide by the way but a German airplane that crashed in the French Alps and I'm worried because I'm flying to Pittsburgh so that's the way it is in aviation is anybody's crash is everybody's crash and the only other industry I know that really is that way big time is the nuclear power industry where you see Fukushima Daiichi and then the German government says no more nuclear reactors so they are very much that way too that anybody's accident is everybody's accident so the airlines don't compete on safety and that's why you have never seen an airline ad that says we are the safest airline out there because they do not compete on safety in fact they scratch each other's back regarding safety because not only do I not want to crash myself I don't want Robert's airplanes to crash either so nobody anybody's crash is everybody's crash on the other hand the automakers compete vigorously on safety and to me that's a good thing because to the extent you buy the car with the 5 stars instead of the 4 stars that helps safety innovations penetrate the fleet faster so I'm not sure it would be a good idea to try to transfer that model from the aviation system to the car system and say you guys shouldn't complete on safety I think there is room for competition on safety the trick is how do you take advantage of the fact that it penetrates the fleet faster when it's got 5 stars and still have opportunities for the safety messages to be transmitted to everybody so everybody knows how that works so that's going to be a challenge I think there's a way to take advantage of that safety benefit without reducing that system that now helps safety implementations penetrate the fleet faster we don't see any ethical issues in any automation we look at in any other mode before today so the example that I've often used is you're going along in your lane fat dumb and happy and for reasons that don't matter there's an 80,000 pound truck coming at you in your lane 15 pedestrians on the sidewalk so what is your automation going to do is it going to take you in the truck to protect the pedestrians or are going to take out the pedestrians to protect you and you know that's the kind of issue that I don't think should be decided we need to have some serious discussion on that but again I don't see any talk about ethics except possibly in academia but to me the sooner you put automation ethics issues into the system the better it's just like safety if you put safety on after the fact it's not going to be as efficient and as effective I think the same is true with ethical protections and so how are we going to address the ethical issues that are faced by this automation mode that we haven't seen in any other automation mode I don't have a good answer to that except I think we need to get seriously involved with this one and I don't see any federal engagement on this issue so let's look at some of the roles of the governments I'm going to start with the federal government because I think it's very important for NHTSA to stand up to this challenge which they're not doing now because they're in a regulatory regime that says if you want to put in one new regulation you got to get rid of two old regulations so that is a pretty arbitrary guideline and it puts a constraint on what they can do but I think it's important for two reasons to have the feds get in charge of this one is because you don't want different requirements in each state with a patchwork quilt of requirements from state to state the other one is the fact that there are lots of other countries that are involved in this issue that it's not a US issue this is a worldwide issue so Japan and Germany and Sweden and Korea and all those countries that make cars need to be engaged in this as well and who's that engagement going to be with it's not going to be with the state of Michigan it's going to be with the federal government and they need to be engaged in that in order to effectively do that so I think there's a huge opportunity a huge requirement need for the federal government to become more engaged in this endeavor and I don't see it with today's federal government so let's move to the state and local government issues so if the feds don't to the extent the feds don't establish requirements the states are going to have to do it so what about things like licensing requirements which the states historically have done what kind of license are you going to need for your driverless car I don't know how that's going to work do you have to have hands on the wheel some states say yes some states don't what about the requirements regarding even having a steering wheel or brakes all of that so what about the infrastructure how much is that going to be what what are the states and cities going to have to do regarding the infrastructure when this is fully automated you may are we still going to have traffic signal are we going to need lines on the streets what about street signage in general what about street parking will there be street parking will there be big parking garages because the predictions are there's going to be a whole lot fewer cars because we're going to move to a system where people won't own cars anymore but they'll beckon cars and that means that instead of being used one or two hours a day like your car probably is these cars are being used much more which means a you're not going to have you're going to have a lot fewer cars and b you're probably not going to own your own car because it's not worth it for you to put all the fixed cost into buying this asset that's going to be parked for 22 hours a day so so there's a whole lot of changes and then so that means without parking garages what about the revenue that the cities and states get from the parking garages what about all the dedicated should there be dedicated lanes for AVs how about bicycles, motorcycles and pedestrians should they be segregated somehow so there's lots of issues that the state and local governments are going to have to address what about the revenue from auto registration if there's fewer autos what about the fuel taxes because if cars are going electric which I think is I've got an electric car and I think the car the future is going to be electric what's that going to do to fuel taxes what's that going to do to all the revenues that the cities get now from parking and moving violations what about parking lot taxes I mean there are so many issues that nobody really has the answer to now because this is going to be so transformative that until we get some idea what direction this is taking it's going to be very difficult for the states and cities to respond in a way that helps to complement these changes because we don't know yet what direction these changes are going to take what about resources well they need as many traffic police what about emergency response well they need as many ambulance services what about infrastructure maintenance because there won't be as much infrastructure to maintain so all these issues what about slippery streets again the automation today is not able to handle slippery streets what about here's one that's not a state or local government issue but what about organ donors if we are not killing a hundred people every day like we are today what's that going to do to the organ donor situation stay tuned so automation offers lots and lots of potential benefits and I'm a big fan of automation and it's going to not only the live save but so many other benefits but there are lots and lots of challenges so I think first the auto world needs to pay attention to the aviation world which has been automating for decades and there's a lot of lessons learned and it's a shame every time we see one of those lessons learned that's not followed by the auto industry and they make a mistake which makes the even more skeptical and not only do they have those issues but as I said there are a whole bunch of other issues because automation on the ground is much more challenging and complex than in the air there are a bunch of issues that the auto world is going to have to deal with that the aviation world didn't so anticipate big changes for everybody this will truly be transformative if you try to predict ten years ago all of the things your iPhone will do now you would probably be way off if we try to predict today what our iPhones will do in ten years will probably be way off nobody I've seen so many predictions on how this is going to look that nobody is really ready none of the states and local governments are really ready to put serious infrastructure changes in place because nobody knows yet which way this is going to go so thanks again for the opportunity to come and talk about these issues this has been fascinating and I look forward to the questions that you might have thank you very much thanks everyone now we have questions and so our illustrious for school panel students will ask those questions and also we can take questions via twitter the hashtag policy talks great well thank you so much for your time really appreciate this my name is Avril Prakash I'm a second year MBA MPP focused on emerging tech regulation interested in the disruption happening in the auto industry so really excited to hear about you speak today the first question that we have here is are the auto companies aware of the lessons you've described well that's a real good question I have to wonder if they are because I don't see them following it but there's no secrets these are all out there in the public domain so it's not like they have to find somebody who's going to shoot them if they tell them there's these secrets there are no secrets this is all out there in the public if they're aware of it I'm not seeing it because I'm seeing to the extent they're making the same mistakes over again to me that tells me they are not either they're not aware of it or they're just not paying attention to it because they think it's that's aviation this isn't so it's not applicable and I think and then there's a follow up to that is are they invested in learning from these lessons what do you expect in the future as you talk about these changes that we should expect I'm hoping that eventually if I give this presentation in enough places I'll capture some attention to make them realize there's there's lots of history for automation development aviation and they need to pay more attention to that history thank you and thank you for your talk I'm Eric Hans I'm a second year MPP at the Ford School and the co-chair of Mobility Policy Lab a student organization focused on growing discussion of mobility and transportation policy here at Ford and across campus we have several questions about states and I'm going to try to condense them into one sure so what challenges do you see with in lieu of unified federal regulation about competition between states in developing smart policies where are the potential fault lines where are the incentives where are the drivers that could produce error and other challenges thank you that's a great question and the best I can say on that since I am not that invested in this industry because most of my work was aviation based so I'm just hoping that the states are talking to each other I don't know if there's like a trade association that they can work through that they're talking to each other so that will create some level of harmony between the states as opposed to having the patchwork quilt of 50 different ways to do it I don't know how much that's happening I just hope it is because to me 50 different answers is not going to be a good solution and a follow-up question to that do you think that the states currently have the appropriate resources to be able to handle such complex systems in a way that both ensures safety in consistency but also meets the policy outcomes that states are interested in that's a good question and it's hard to know because it's hard without knowing what kinds of changes are going to need to make and therefore we don't know whether some of these changes are going to reduce expenses here in return for expenses there and just what the nature of those changes are going to be that's very difficult to know at this early stage it's going to be I predict an iterative situation where what the states and local governments do is going to depend on what the car makers do and vice versa I think it's going to be very iterative so it's very difficult to predict at this early stage where that's going to go I'm sorry not to have the answer but this is so transformative that I think it's going to be much bigger than just not having parking on the streets it's going to be a much bigger change than that so switching gears a little okay so the Chrysler electric cars won't have gears by the way so that's probably why nobody laughed so the Chrysler 300's 2020 model has no printed owners manual and it's actually all on the app console so how could a car built that way adapt to being an autonomous vehicle well the absence of an owners manual doesn't surprise me at all and the cars that do have owners manual nobody reads them anyway so that doesn't challenge me at all but I wonder so ask me again the question because I'm not sure if I understood the so it's basically the crux of the question is how does a car that's built that way that's basically with no printed owners manual so technologically advanced how does it adapt to being an autonomous vehicle I'm assuming I think the crux of the question is whether it's an iterative process within the software or how that might change in the future the only thing I can say to that is the way the owners manual is is probably irrelevant because nobody looks at it anyway so it's certainly if it's electronic owners manual it can adapt as the car is adapting and that's a good sign but to the extent nobody reads it we have a slew of questions about insurance could you could you discuss briefly what role insurance companies will play in an AV future and also you know how does the existing insurance system as it's currently construed form a type a barrier to the adoption of emerging autonomous vehicles well the lawyer that's a good question the lawyers are going to have a field day with this because they're going to try to figure out okay who's responsible for this crash is that the driver or is that the machine or is it the infrastructure or just who's really responsible for this crash and the for a while there there's the the attorney's going to have a field day with that and I am an attorney so I'm familiar with that but how this is how the dust is going to ultimately settle on that I don't know I have to think the manufacturers are going to play a much bigger role in the in the liability because they're the ones who are designing the software so the drivers may still be engaged but to the extent you take the drivers out of it that inherently reduces the potential liability of the drivers and increases the potential liability on both the automakers and the infrastructure designer so that's going to be tough on to predict that I'm going to give you that same answer as before stay tuned because that's so transformative it's hard to know today just what direction that will take sorry not to have any more you know concrete answers but I think that's inherent in the in the aspect of this being such a transformative technology so our next question is how do you feel about an incremental approach to implementation so basically starting with smaller tasks and then moving on up so do you think that's a good approach in this new era well that's that's an interesting question because I know there are huge debates there's kind of two not so parallel channels some say let's do baby steps at a time until we get there some say let's jump all the way to the ultimate conclusion I don't think anybody's ready to jump to the ultimate conclusion because I don't think anybody's ready for a driverless car for prime time and so so the question is as you go baby step at a time what the aviation world is finding out is the closer you get to full automation taking the pilot out the more challenging the human factors issues become that's kind of an irony of it but the closer you get to complete driverless situation the more challenging the the human factors issues are and one of them is what I mentioned that humans are not good monitors of reliable system so the more reliable the systems get the more challenging it is to keep the pilots engaged and you're hearing today about pilots who aren't engaged enough so they lose their basic skills so these pilots in Air France 447 they didn't have the basic skill of how to fly straight and level at cruise altitude because they relied on the automation to do it so there are pros and cons to both approaches I don't think anybody's ready for prime time just to leap to full automation but on the other hand as they take the baby steps toward it the later the closer the baby steps are to removing the human the more challenging the human factors issues become so I personally prefer the baby step approach because I just don't think anybody is yet ready for prime time and I mean like I said the theory is very simple most of this is human error and it says 96% of these accidents are human error so all I gotta do is remove the human and you remove the human error so that would remove 96% of the problem well that sounds good but that's oversimplistic it's not gonna be that simple and you know we're gonna have to do it in measured steps and I think the feds need to play a role in every one of those steps as opposed to just letting it happen I don't mean in terms of creating constraints around the improvements but I think the feds need to be on top of it so that they can play the active role that they need to play not only in the US but in the worldwide community could you provide any new lessons learned that would come from the investigations into the downing of the two Boeing 737 MAX planes that would as they relate to AV adoption I'm not sure how much these are related to AV but I can just tell you the big picture view of the certification process and that's what I worked on was the certification process for the 737 MAX flight control system the big picture view of that is the system for approving airplanes has worked very well for decades and that's reflected by the exemplary safety record that aviation has of going for almost ten years and nine or ten billion passengers without a single passenger fatality that's a very amazing accomplishment so a foundation of that accomplishment is a very good, a very safe airplane and the foundation for a very safe airplane is a very safe system for certifying the airplanes but so what we found was the system has worked well historically but what it needs to do is because airplanes are getting ever more complex as the airplanes get more complex the system needs to be revisited because these two tragic crashes show that the system is not still letting things fall between the cracks they need to upgrade this good system and make it better to respond to this new reality of more complex airplanes and by the way this is not an FAA Boeing issue this is a worldwide issue for all governments and that's why I give kudos to the FAA for studying this with a worldwide group of certification experts from basically every certifying authority in the world except Russia to come together them plus NASA and I was asked to lead that effort it's called the joint authorities technical review I was asked to lead that effort and so we came up with recommendations to the FAA that we gave them last October about how to bring the system up to date so it can respond more effectively to today's realities of more complex airplanes so I'm not sure that there are lessons learned in that one for the aviation world except to the extent you can say that the more complex the vehicles become the more sophisticated the regulation mechanism has to be in order to respond to that new complexity so the next question we have is asking who do you think should participate in decisions of ethics as related to automated vehicles to what extent should citizen expertise be valued by our like technologists or regulators well that's a good question I think it needs to be led by the feds but I think it needs strong input from academia and I think it needs input from the driver community as well so I think all of those need to be in this picture somehow I'm not sure exactly how that would work whether there's any mechanism today to do that but I think it needs to start with the feds and I think the academia like you guys because there is I've seen ethics work going on in many of the universities and I think they need to be involved and not by the way not just us either but it needs to be a worldwide situation so I don't know of any mechanism today to do that and not only get input from academia but also input from the driver public because that's one of the things about input from the driver public is the quality of drivers is very that's a very big bell curve and so trying to sort of get a representative sample of that is going to be a challenge and in speaking with your experience with the NTSB can you talk about the stakeholders that you involved in the decision making process or your recommendation process I think that would be helpful for us to get some more context sure thank you the way the NTSB works is what they call the party system so when they're investigating say a big airline crash then we have what's called the party system and the party system means that everybody who's got a dog in the fight is involved in the investigation so the airline would be in it the pilot union would be in it the air traffic controllers would be in it the airports if they were involved would be in it the engine designers everybody who has a dog in the fight would be parties and the reason for the parties is because they have technical knowledge that the NTSB wouldn't have so if we're looking for what's this piece of metal out in the middle of the field NTSB would have no idea what that is but the person from Boeing would say oh I know exactly what that piece is and that piece is this that and the other or if we talk about how do the pilots usually behave we could say the NTSB wouldn't have any notion of how the pilots would behave but we could ask the pilot union for advice on how that usually works so the NTSB does it with a very collaborative approach that includes everybody who's got a dog in the fight for the purposes of gathering the facts because the NTSB has no way now once the factual part is finished then the NTSB moves to a non-collaborative approach of developing the analysis by itself so the NTSB puts all the facts on the website so that everybody can see them and we invite all the parties to submit their analysis against those facts and we invite the public like you against the facts but the ultimate analysis is done solely by the NTSB and the purpose of that is so that we don't get in accused of saying well Boeing unduly influence that one or American Airlines or the pilot union or whoever unduly influence that conclusion and recommendation and so the analysis is done solely by the NTSB but for the factual development it includes all the parties because they have that technical capability that the NTSB doesn't have A similar question kind of building on the complexity idea so regulators are currently faced with highly complex systems in passenger vehicles recently several unintended acceleration cases have come before NTSA and were essentially ruled to be unknowable due to the complexity of software what ramifications does this have for even more complex AV systems how will regulators respond in the cases of crashes that involve proprietary systems or systems that are unknowable in their complexity well thank you for that question the good news on that is that the cars are becoming more like airplanes and that they're having more like recorders on the cars that tell us what happened so if we wanted to see from the recorder whether anybody was like today most of the cars if you start them you have to have your foot on the brake in order for them to even start to have recorders on them that would show what was the position of the brake what was the position of the accelerator all those kinds of questions and might even have voice recording to hear what's going on the noises in the car so the good news is that the cars are more and more having this kind of recording capability and it's actually not necessarily for accident investigation purposes but it's actually so the manufacturers can use that feedback because like Tesla has continuous feedback from your car to the manufacturer and they use that information to their credit to help improve the quality of the car so the good news is that as the cars become more technologically complex they're going to have more recorders that help people like the investigators figure out what really happened here what was the totality of circumstances so that we can have a better so we're not guessing as to what the cause was one thing another question I had was we've talked about obviously these the government's trying to be prepared for the future of this technology and how oftentimes regulation or government policy can be reactive in this space I'm curious to get your perspective on how government regulation or perhaps any future recommendations could help address the equity issues surrounding in safety and within transportation and how you see that to be changing too the equity issues are a big hole that I've seen that is not adequately being addressed because I'm looking at to the extent people are going to be beckoning cars is that something that only rich people can do or will there be a scenario where anybody can do that and that depends on so many circumstances so for example I'm seeing one of the models I'm seeing is the first mile, last mile model where you would beckon the car to take you from where you are to the mass transit and then you would get mass transit and go the rest of the way on mass transit and if that doesn't take you where you're going that's the last mile thing so that's going to be one scenario if it's a car all the way to your destination which I have trouble seeing because when they talk about autonomous cars will reduce auto congestion I don't see how it's going to reduce congestion if everybody is taking an autonomous vehicle to their destination versus taking you so what I have not seen in any of these models is who's addressing how it's going to affect people of different means and I don't see that happening yet I have no clue where that's going to go I think it needs to be addressed but people with less means don't have as much political clout to make sure that they're covered by this we'll have to see where that goes I'm hoping that people will want to address the totality of circumstances realizing that if these people of less means can get to work more easily we all win so it's not a subsidy thing we're all in it together because it can be beneficial to all of us so I don't have a good answer to that one and again I'm sorry not to have a complete answer but that's what happens when it's very transformative technology so this is another tough question so these vehicles will be collecting a lot of location data and will be reliant in some instances on that location data to be located in space to be properly regulated and also we will have there are certain models for AV that require vehicle to infrastructure connections which provides even more reams of location data could you speak to the privacy concerns of location data and how this issue is currently being thought about in the regulatory community and how it ties into other federal action on personally identifiable information data privacy and location data please that's on the list of issues that haven't been dealt with in aviation because aviation doesn't really have a privacy issue in that sense because the things from the black boxes in aviation belong to the airline and so nobody can have access to them unless the airline wants it so there really isn't that privacy issue so that's an example of an issue that's going to be faced with autonomous vehicles that is not currently being faced in aviation and it will take some privacy protections because I know people are even today people are worried that with your on-star system it will tell where you are so what if I find out that my wife is at her boyfriend's house because it's on-star system so there are lots of privacy issues associated with that which haven't been I haven't seen any serious issues to adequately address those but I'm assuming that the political oomph behind that will be enough that privacy protections will be maintained it may be that they're behind the technology a little bit but eventually I'll be surprised if there aren't some huge privacy protections to prevent misuse of that information so we have a question coming in from Twitter so tangentially connected to automated vehicles how do you do you have thoughts about urban air mobility there seems to be a lot of buzz about projects like Uber Elevate and lots of technical and regulatory challenges around them so I would love to hear your thoughts on this new developing area so that's going to be a great question that's going to be even more interesting than what we're seeing on the ground because not only do they have all the automation issues that we're now seeing in cars but they have two additional issues first of which is propulsion because the ones that I've seen the geometry of them is typically that the vehicle is surrounded by vertical lift fans so let's take a scenario where there's a fan on each corner so if you lose the fan on one corner then you are no longer controllable unless that fan is activated by the other engines and that's the only single example that I know of is the Osprey which has it's a two engine airplane where the wings are like this and it takes off vertically then the wings go forward and it flies horizontally both of those engines power both rotors so if one engine is lost you still got both rotors you still got propulsion for both rotors because if you didn't that would be an instant crash that would be catastrophic so absent some way to maintain that control and I just used the four corner example because I know all you need is three to keep control but the three need to be more or less symmetrically spaced around the middle and if there's three of four on a four corner approach that's not symmetrical so I don't see that happening so that's number one is what are they going to do in the case of a propulsion failure because unlike airplanes which usually crash over uninhabited ground these vehicles every time they crash because they spend 100% of their life over the city they're going to crash into the city somewhere that means they're going to damage buildings that means they're going to hurt people and all it's going to take is one of those if you remember that they used to have helicopter service from the New York airports to the Pan Am building one of those crashed in fact it didn't even crash it just lost a rotor blade on top of the Pan Am building and that cascaded down and hurt a bunch of people then that destroyed the whole service if you have one crash of an urban air mobility vehicle anywhere in the world that's going to seriously impact what you do is now you got low altitude traffic of a bunch of vehicles buzzing around and the FAA has no way to control low altitude traffic like that so that's going to need a whole new infrastructure to control the traffic in addition to having collision avoidance systems in the vehicles but it's going to take a whole new structuring so the bottom line of that is just like I thought 10 years ago the predictions of how soon we're going to have driverless vehicles on the street of urban air mobility vehicles in not so long a period of time are unrealistically optimistic so they're going to have huge challenges to face and the one that I said in Uber had a big program in DC and I said in this taxi which seats eight people so that means it's a pretty big device that means if this thing crashes somebody's going to get hurt so that's going to be even bigger challenge and I think that the predictions of that are even more unrealistically optimistic than they are for cars we have another question from twitter since the US is the largest automark in the world why isn't US DOT taking broader leadership and are there risks that are involved with abdication on the federal level to not set the playing field for AV is there a possibility for other countries to take the torch and run with it there's a big possibility that and that's why I'm disappointed to see that NHTSA is not taking a more pronounced leadership role in addressing this because then that leaves a vacuum for other countries to take over and I think we need to be we need to be leaders in that rather than followers so yes I'm very concerned about not seeing NHTSA take the lead in this I used to be I was the acting administrator of NHTSA myself 20 years ago so I don't want to speak too negatively to them but I'm just not seeing them take the come up to the plate to take these issues in automation I'm also curious to learn about like within the NTSB what new technologies you've been using in terms of monitoring whether there been you're a proponent of creating innovation within the recommendations you develop so that would be also interesting to hear about sure we've been I told people that when I took over chairmanship of the NTSB I told people that the NTSB has to be innovative because the transportation world is innovative and if we don't innovate and continue improving we're going to fall behind I can't tell you how pleased I was when I visited the 3M facility in Minneapolis and saw the management philosophy of I think his name was Wright WRIGHT in the 1930s where he said this company has to be innovative and if we're not innovative we're going to fall behind and by the way innovative means thinking out of the box which also means you're going to make some mistakes so that means I'm not going to be hard on you if you make mistakes quite the contrary I'm going to tell you that you're making mistakes tell me you're trying to think out of the box and I'm going to praise you for doing that and see what you can learn from those mistakes so to me that's essential if the NTSB so one of the things that we started doing in my tenure was we started using drones to help us investigate accidents so now the drones can go into buildings where like we had a collision where the train crashed into the terminus at the station to make the building unsuitable for entry but yet we could take drones in there and look and see what was happening or we can map the debris field with a drone and with GPS on the drone we can map the debris field so we know where all the debris fell on this crash and then we can have a digital map so that we can use that later to help us figure out the dynamics of the crash by where the debris is so we've done a lot of we're also doing innovation on how to use the information better and from the cockpit voice recorder so there's lots of innovation going on at the NTSB to help them stay up with this rapidly advancing field and transportation we're getting a lot of action on Twitter today so keep tweeting I have another question about a couple of the crashes that you highlighted the Tesla crash but maybe more appropriately the Tempe Arizona Uber crash involved private vehicles operating on the public right of way that were essentially untested or technologies that didn't have third party oversight operating in the public right of way what is the current safety net to ensure that the public the traveling public can seek restitution or that local governments aren't held liable for public public use of public operation of these untested technologies and this is where one of the ways where I say the feds need to be more engaged in this because without uniform national guidance on that I think we got a problem and it's not only for testing on the streets but it's also for all these upgrades that keep going into your cars I have to as I said I wonder how adequately was that upgrade tested against the rest of the software that it's entering does that generate some unanticipated consequences with adding that piece to the software that's already there seems to me the feds need to be on that and on that as well because to make sure that the software that's added doesn't make the car more dangerous than it otherwise was and to me that's not a state function that's a federal function and so we're going to have to figure out a way that before this gets broadcast to your car it's already been tested in a way that's adequate to the feds so that they will be comfortable with yes this is safe enough to be out on the streets but I don't see that happening today maybe it will happen as this evolves but I'm not aware of it happening today and to the point of these the data on these crashes that we've had these recent issues I'm curious to see like what your responses to the pushback from these companies from the Ubers and the Tesla's of the world who believe that this is an innovation or the future of transportation so how you kind of communicate to the pushback that they're offering in this area well the pushback comes after an accident that's one thing if a pushback is just generic that's a different scenario altogether there's always going to be pushback where the industry thinks oh yeah we got this figured out but they don't necessarily so there's always going to be that and we 737 max is an indication of that so I think that's going to happen but that's where if the feds don't get into this and play a role that's not happening in a way that it needs to so I think that needs again another reason why the feds need to get actively involved in this arena because they're the other ones you don't want the states doing this individually you don't want the manufacturers deciding ethical issues you need the feds in there for some guidance and leadership so we have another question from twitter could you thanks twitter could you address the carbon just kidding I use any nice 17 year old daughter to help me with this stuff but I'm just kidding I do know what twitter is will automation lead to reductions in carbon emissions or what are the various scenarios for carbon emissions and other types of vehicle based pollutants well there are two possible answers to that one is the reduced number of cars is going to reduce the pollution but I think the bigger answer is I think that automation is going to mean electric cars for the most part because I don't because gas power cars don't respond to the automation as well as electric cars do and I've got an electric car so I've been there done that and just to take an example the propulsive efficiency of gasoline power cars is like less than 30% propulsive efficiency of electric cars is like more than 90% so right there there's a three times disparity in the propulsive efficiency so in terms of how many miles you go for this bit of energy you do three times more in an electric car so the reason the gas cars are so far behind on that is because for example they generate so much heat so that's kind of a double whammy because not only does the energy for the heat come from the gasoline so that's not propulsive energy but also you need energy from the gasoline to cool to take care of all that heat and get rid of it with the fan and the pump, the water pump that moves the water through the radiator and then the fan that blows the heat out of the radiator so it's kind of a double whammy it generates more heat and then it's got to get rid of that heat that it generates so that's a big reason why it is so much more the propulsive efficiency is so much worse for internal combustion engine so internal combustion engines make a lot of noise and where does the energy for the noise come from? comes from the fuel so when I tell a Corvette owner friend of mine that my little electric car can beat your Corvette across the intersection because as you know electric motors have maximum torque at zero so that's why the most powerful Tesla will go from zero to 60 in less than four seconds so I tell my friend at the Corvette my car can beat your car across the intersection and he says yeah but your car doesn't make noise so there are lots of reasons why electric cars are far more efficient than internal combustion engines and I'm wondering why am I not hearing a Bruhaha protest from the petroleum industry why am I not hearing a Bruhaha protest from the car industry when people are predicting 60 to 70 percent fewer cars out there I don't know why I'm not hearing it but maybe something is happening behind the scenes I haven't heard of it yet but to me there's a huge difference in pollution from what I just told you was a repulsive efficiency if you really want to look at the question you got to look at beginning to end and see what the totality of the energy use is from the time you get whatever the fuel is out of the ground and either carry it as gas to the gas station and put it in your car or carry it as electricity to your car in the garage so the real answer to that would have to depend on that totality of circumstances but I suspect that even if you look at the big picture the electric cars are going to come way ahead on that which means less pollution and more user friendly and more sustainable could you address the differences in potential emissions impacts of the variety of AV deployments that you've discussed you discussed both a fleet based deployment with on demand service and also the potential for a private model again I think the electric car is going to be a way ahead on emissions of the fleet and to the extent the car is being used continuously instead of being parked so much of the time a lot of the inefficiency of gasoline cars is while they're warming up and so if you have less of that you don't have that concept anymore so I think the electric cars are going to be the way of the future for autonomous vehicles and I think they're going to be far better on emissions than internal combustion engines in the total picture so this will be our last question AV we can stay after and chat if you're okay with that AV represents a complex solution to a seemingly simple set of problems are there examples from the aviation industry that have stripped complexity rather than added more complexity that we can take as lessons learned moving into the future I'm not sure I would describe it as a simple set of problems but certainly aviation as it has become more automated has also become more complex which has increased the challenges and that's where the JATR went with this recommendations of the FAA is airplanes are becoming more complex and the system that approves these airplanes needs to be revised and updated to respond to that increasing complexity Well let's thank Chris one more time for coming Thank you Thank you so much for joining us here at the Ford School we've been taking so many questions that you took a lot of questions Thank you, it's my pleasure There's a reception out front now please join us and you can ask them even more questions if you would like So thank you everyone for attending Thank you