Hit enter to search or ESC to close

Alex Roy sits down with returning guest, Sam Abuelsamid, former auto engineer and current Principal Analyst of E-Mobility Research with Guidehouse Insights. The duo discusses the hottest topics in mobility, AI, and the autonomous vehicle space in this year-in-review episode.

Listen On

Apple
Google
Spotify
iHeart Radio

Share

View Episode Transcript

Episode Transcript

Alex Roy (00:06):

Hey everyone, welcome to No Parking, the podcast that cuts through the hype around self-driving and artificial intelligence. I’m Alex Roy, and it’s the end of the year. That means it’s time for our annual myth-busting episode. We also make a few predictions, so that means we have to bring back old friend of the show, Sam Abuelsamid. Oh my God, is my memory failing me? I’m pretty sure his title is principal analyst at Guidehouse Insights. But I think he gets promoted every time he comes back on the show. It’s like, right before we do the show. Anyway, Sam’s been making amazing predictions for as long as I can remember. And that means we should listen to what he has to say. Let’s jump right into it.

 

Alex Roy (00:50):

Well, it’s awesome to be back with you, Sam Abuelsamid, as usual. I don’t know how the world would function without a voice like yours in the transportation sector, saying true things with intelligent analysis. How have you been?

 

Sam Abuelsamid (01:05):

I’m doing great, Alex. And trust me, the world would roll along just fine without me. And contrary to what Mr. Musk says, the planet will survive billions of years long after all of us humans are extinct. We don’t need to be a multi planetary species.

 

Alex Roy (01:23):

Well, wait a second. That’s a whole other podcast. But I actually am, I’m not saying I’m bullish or bearish on Musk’s interplanetary plans, but I’m absolutely convinced that we do need to colonize other planets. I’m just not convinced that it’s going to happen in the next five years, as he just announced yesterday. We’re going to Mars in five years. I’m not so sure.

 

Sam Abuelsamid (01:44):

I’m not so sure we actually… If we can’t preserve this planet as one where we can survive and thrive, we sure as heck should not be on some other planet doing what we’ve done here.

 

Alex Roy (01:55):

I agree with that. But if we go back 110 years, between 1900 and 1910, was there a pundit saying, You know, these horses are just, it’s not going to last. Cars are the future. Was there a guy?

 

Sam Abuelsamid (02:14):

Probably Henry Ford.

 

Alex Roy (02:17):

Okay. Before Henry Ford, there had to have been someone who looked at perambulators and steam engines and said, This is the future.

 

Sam Abuelsamid (02:25):

I’m sure that there were.

 

Alex Roy (02:27):

All right, that’s another show. Let’s dive right into this. All right. Top 10 myths of 2021. I feel like every year we have to repeat a couple of them, but they always reboot themselves. And another myth of 2021, of course, number 10: driver assistance and autonomous vehicles are the same thing.

 

Sam Abuelsamid (02:45):

They are absolutely not the same thing. But it’s interesting, I do a fair number of speaking gigs every year for various conferences. And one of my recent themes in the talks I’ve done, including last month for the NVIDIA Fall GTC, was… The title of the talk was From Autonomy to ADAS. Because for many years, there’s been this idea of going from ADAS to autonomy, that we would have this progression. And that ADAS would provide the building blocks, and we would just continually step up. And that’s kind of the approach that Tesla has taken. But in fact, what I’ve been seeing over the last couple of years, is kind of the reversal of that. Rather… Because most companies that are developing automated driving systems are, they realized the difficulty and probably near impossibility of trying to evolve driver assistance into full autonomy.

 

Sam Abuelsamid (03:54):

And so they’ve been working to design systems from the ground up that were designed to be autonomous. And obviously, as we know, since there has not been, any kind of large-scale deployment of that technology yet, that it’s clearly not quite ready for the masses. It’s getting there, but in order to take advantage of some of those lessons that have been learned over the last decade, decade and a half, and try to get some return on the massive investments that have been made, what we’re actually seeing is the technologies that have been developed specifically for automated driving systems, are actually starting to filter down into driver assist.

 

Sam Abuelsamid (04:37):

So we’re seeing things like LIDAR that are coming to Level 2 systems. And also imaging radar, and more advanced compute. All these things that, these high power, high performance compute platforms, that are coming down into more basic driver assist systems. And so that’s actually going to be how, at least in the near term, a lot of this investment that’s been done is going to benefit large numbers of people globally. Rather than the near-term prospects for fully automated systems.

 

Alex Roy (05:19):

You know, that brings us to myth number nine, which is-

 

Sam Abuelsamid (05:26):

So we’re going in no particular order today?

Alex Roy (05:28):

We’re going in reverse order. Like Dave Letterman, we’re going to go in reverse order.

 

Sam Abuelsamid (05:32):

Oh okay, so that was number 10. I thought that was number one that we just did.

 

Alex Roy (05:34):

No that was number 10. But number nine is: supply chains are great. For many years you, people were saying how difficult it was to get LIDAR. And if you had paid attention to the business, you knew that it was hard to get LIDARs. They were expensive, they weren’t manufactured in large volumes. But suddenly everyone knows that our supply chains actually weren’t that great. Can you unpack that?

 

Sam Abuelsamid (06:01):

Yeah, I think one of the key things that we’ve learned, hopefully we’ve learned it in 2021, is that the trend we’ve had over the last several decades of increasing consolidation, of fewer and fewer companies making products in fewer and fewer places, added a lot of brittleness and fragility to our economy. Looking at the chip sector, for example. There are fewer companies now that are manufacturing semiconductors than there were 20, 30 years ago. And they’re doing it in fewer and fewer places. More and more of it is being done in places like Asia. And this is, I don’t want to sound like too much of a socialist, but this trend of allowing companies to grow bigger and bigger and to optimize for lowest cost.

 

Sam Abuelsamid (06:58):

I mean, I remember my Econ 101 class in college, this concept of competitive advantage. Make the stuff in the places where you have the greatest competitive advantage. The problem with that is, that assumes that everything else is going to be fine. That you’re not going to have any natural disasters or geopolitical tensions or trade wars. And that’s obviously not the case.

 

Sam Abuelsamid (07:27):

And by concentrating too much, too many things in too few places, it’s made us vulnerable to disruptions in those few places. Whether it’s an earthquake or a fire or a trade war, or a global pandemic. And so what we’re actually starting to see now is a shift back where we’re trying to start to diversify that supply chain a little more. Make some of these things, make things like chips, whether it’s the silicon wafer manufacturing or the packaging of the chips, or the manufacturing of LIDAR sensors, or literally anything else. Trying to make them in more places and localize it more, so that you are less vulnerable to logistics disruptions, like not being able to offload container ships.

 

Sam Abuelsamid (08:22):

All these things are piling up. And so I think we’re going to see a very interesting shift in the global economy over the next several years.

 

Alex Roy (08:31):

What’s really ironic about the supply chain problem is that when I moved into my new apartment in Florida and I wanted to furnish it, it was almost impossible to get anything that I liked. I ended up going to Ikea and just simplifying my life by buying everything that was white and in stock in the store. And as a result, I live in the most Miami Vice apartment ever.

 

Sam Abuelsamid (08:53):

Yeah. I’m seeing you on screen right now, and it’s like one of the old Jony Ive Apple videos, where he was always in a white room and there was nothing around him. That’s kind of what your background looks like right now.

 

Alex Roy (09:06):

The other thing which really struck me is that when I talk about supply chain with my friends, people have this assumption that a single truck drives from a factory literally to their house with the products. They’re just not aware that there are one or two or sometimes many more distribution points in between a factory and their home. I just read this book called, was it, not Just in Time, Always On… The Christopher Mims book about supply chain. And what I loved about it is that each chapter was devoted to a single piece, or like a single segment, of a supply chain to move a box from a factory to one’s house.

 

Alex Roy (09:52):

And when people talk about automation, and automation potentially “supply chain problems,” they seem to think that you’re going to just automate all those segments. And it’s just not true. Logically, or at least realistically, we may see one or more segments automated, but you’re still going to need people because people ultimately are the ones who make sure things run. Mistakes get made, people solve them.

 

Sam Abuelsamid (10:18):

Yeah. That’s been one of the challenges, one of the many challenges we’ve seen this year. At this point, we’ve probably all heard about the backlog of container ships sitting off the coast of California. And I’ve been in Southern California a couple of times in recent weeks. And driving along the coastline, seeing all these ships sitting out there. And the reason why they’re sitting out there is because we don’t have enough truck drivers to get the containers out of the port. And so there’s nowhere to allow the ships to come in and offload their containers. The ports are full of containers that they can’t get out because they don’t have truck drivers. So the ships are sitting out there waiting until those containers move so they can come in and offload their own containers, and then move back. So it’s this cascading effect of a whole bunch of different challenges.

 

Alex Roy (11:10):

Yeah. People just, when you say it’s brittle, the other thing that’s very brittle is our understanding of complexity. And, I’m buying my daughter a book, she’s turning three, about how things are made. I think what we really need kids to have are books about like how things get to us. But that’s a whole other podcast.

 

Sam Abuelsamid (11:29):

Yeah, well by the time she’s in kindergarten, she can be reading Chris Mims’ book, it’s called Arriving Today.

 

Alex Roy (11:34):

That’s the one. All right. Number eight. I love this myth because it’s been occurring each year, but this year it got really loud. Going driverless, when an autonomous vehicle company goes driverless, that means they’re the leader. Unpack that myth for us.

 

Sam Abuelsamid (11:54):

As with the supply chain question, it’s not that simple. Where are you going driverless? Under what conditions are you going driverless? Can you go driverless everywhere? Or can you go driverless on a few specific roads in Chandler, Arizona? Or one particular neighborhood of San Francisco? What does that really mean?

 

Sam Abuelsamid (12:18):

That doesn’t necessarily mean you’re in the lead. There’s likelihood that other companies, if they chose to follow the same path, test in the same places, they could probably be doing the same thing or perhaps even more, under those same conditions. It’s all part of the way we define automation, this idea of Operational Design Domains. Where is the system designed to work? If it’s designed to work under very specific conditions, a specific domain, yeah you could be in the lead in that domain. But that doesn’t mean you’re in the lead overall.

 

Alex Roy (12:58):

This is a subset of that myth. When do you think we’re going to see two autonomous vehicle developers operating in the same place and actually competing that way?

 

Sam Abuelsamid (13:13):

I think we will see that probably in the first half of 2022, in San Francisco. Cruise has been indicating that, essentially as soon as they get their permit from the California Public Utilities Commission, which will allow them to operate a robotaxi service, they’re going to turn it on very soon after that.

 

Sam Abuelsamid (13:36):

And Waymo, similarly, they’re on a similar timeline. They got their permit from the DMV in California around the same time that Cruise did, and then subsequently applied for their PUC permit. And they’re operating in the same area, the western side of San Francisco, residential area. And I would expect that within, probably by the middle of next year, we’ll see both of those companies operating commercial services in that area.

 

Sam Abuelsamid (14:06):

Now, are they going to be able to cover the entirety of a particular neighborhood? Or again, like in Chandler, very specific roads? We don’t know yet. But I think that’s probably going to be the first head-to-head commercial competition we’re going to see.

 

Alex Roy (14:25):

There was a fun tweet the other day from my friend, Patrick George at The Drive, I think you know him. He said, Oh, well, you know, electric vehicles are going to mean the end of car reviews. But actually I disagree. I think electric vehicles and especially autonomous vehicles are going to mean a whole new era of reviews of mobility services. And I can’t wait for someone to compare two services, head-to-head, same pickup, same drop, like Top Gear style. And that’s going to be, I think, a lot of fun.

 

Sam Abuelsamid (14:54):

Yeah. I remember growing up reading comparison tests in the buff books, of top sports coupes or top pickup tru-

 

Sam Abuelsamid (15:03):

… top sports coups or top pickup trucks or whatever it might be. You’d have three, four, five, eight, 10 vehicles in the same segment that they were running back-to-back and doing the same tests, and doing this with mobility services and with EVs, doing charging tests with EVs. How quick can I get 200 miles back into this car from these different stations? And not just testing the vehicles themselves, but also the infrastructure that supports them.

 

Alex Roy (15:34):

If I had been around 100 years ago, I absolutely would’ve compared a horse and a car in exactly that way. I wonder what what Top Gear’s equivalent was in like 1915. All right. We’re moving on to myth number seven. Oh my God. Someone’s been really attacking me on Twitter, on LinkedIn about this myth. Here it is. Geofencing means it’s not autonomous.

 

Sam Abuelsamid (16:02):

That’s a nonsense statement. You can be autonomous within specific constraints. On the other side of that, if the human still has to watch and pay attention and be ready to take over, it’s absolutely not autonomous.

 

Alex Roy (16:21):

Right, right.

 

Sam Abuelsamid (16:22):

So you could have something that’s not geofenced but has to be supervised. That’s not autonomous. So I would say that a vehicle that can operate, that you can just tell it where to go within a given geographic area and then it does it without any further intervention, that’s as autonomous as anything.

 

Alex Roy (16:44):

The other aspect to this myth, which I thought was really interesting, is I’m only aware of one company that has claimed they’re ever going to build a go anywhere, anytime vehicle and sell it, and that’s our friend, Tesla. But just about every other company has stated that they’re going to develop and deploy according to principles like an operational design domain, which could be a geofence. It could be weather, whatever. And every time a company says, “We’re going to deploy a service in city X,” somebody comes out of the woodwork and says, “Oh, you’ve down-scoped. You’ve walked back what we thought you were going to do.” Where did the notion come from originally that autonomous vehicles would operate anywhere, anytime? Did it predate Tesla?

 

Sam Abuelsamid (17:38):

I think both of us are known for having issues with the SAE levels of automation definitions, level zero to five. Level five is this idea of a vehicle that can operate anywhere under any conditions. This was a taxonomy that was defined by a bunch of engineers for engineers, to categorize what a system can do. It was never meant as a public-facing thing. It was Elon Musk, that in October of 2016, when they had the conference call where he announced the launch of Autopilot version two and the first sentence out of his mouth, paraphrased slightly, but he used the words “level five,” basically, he said, “Every Tesla vehicle coming out of the factory today has all the hardware it needs for level five autonomy.” He’s the first one, really, that I know of that has said, “We’re building a system that can operate anywhere.” Nobody else has ever really said that, that I’m aware of. So I think I would say October 2016 is kind of where it started.

 

Alex Roy (18:52):

All right. Myth number six, Europe and China are ahead of the United States in autonomy.

 

Sam Abuelsamid (19:02):

I would not say that’s true about Europe. It could be argued that China is. I think we will see more wider-scale deployments in China over the coming decade. I think we’ll see AVs operating in more places, carrying more people and more goods in China this decade than we will anywhere else. In terms of the development, it’s probably pretty comparable in terms of where some of the Chinese companies like Baidu and Pony, and Plus on the trucking side, and AutoX and a few others are compared to some of the premier companies in the US like Argo and Waymo and Cruise and Aurora. I would say they’re roughly on a par. In Europe, there’s less direct activity going on. In fact, probably one of the most advanced programs in Europe is going on in Munich and Hanover with Volkswagen and some little company called Argo AI, I think.

 

Alex Roy (20:20):

You’re too kind.

 

Sam Abuelsamid (20:22):

Technically, that’s an American thing, but it’s going on in Europe. They’re testing versions of the VW ID.Buzz with the Argo self-driving system on there.

 

Alex Roy (20:30):

For a long time, or at least it’s widely believed that United States and the DARPA Challenges, the Carnegie Mellon and Stanford MIT, are the original seat of autonomous vehicle development. What’s the gate that’s holding back the United States from being the leader? It’s not technical. Is it political? Is it legislative?

 

Sam Abuelsamid (20:51):

No. I would say that the US is as much of a technical leader as any region. Like I said, the Chinese, pretty close, maybe slightly behind from a technical standpoint, but not far off. But there’s the technical side of it and there’s the commercial side of it and the legal side of it. We’ve got a bunch of pilots going on in various locations around the United States, where they’re carrying passengers. There’s one here, just a few minutes away from me, here in Ann Arbor that’s carrying passengers on a daily basis. But that commercial and legal side of it is challenging here, because our political leaders have not been able to reach a consensus on how to put some rules around the deployment of this at a national level.

 

Sam Abuelsamid (21:50):

We’ve got a patchwork of state regulations that cover testing more so than commercial deployment, but we haven’t got any federal regulations as to how do we validate these systems? How do we decide that these systems are safe enough to be used on public roads and to carry paying passengers? And until we can come to some… There’s also been political things going on, on the executive side, the National Highway Traffic Safety Administration, as far as whether we should regulate or at all, or how much we should regulate. I think this year, we’ve seen some significant progress towards at least trying to learn more about what’s actually going on and how we should go about the process of regulating, if we do it at all.

 

Alex Roy (22:44):

So that’s the gate, is federal legislation that would, I guess, harmonize, standardize.

 

Sam Abuelsamid (22:52):

Well, that would enable companies to deploy vehicles that… One of the challenges for an automated vehicle, we have a bunch of regulations, Federal Motor Vehicle Safety Standards that are designed to protect vehicle occupants, to guarantee it or to try to guarantee a certain level of safety for new vehicles. Some of those rules are not really applicable to an automated vehicle. Like for example, we have rules that there’s got to be a steering wheel in there and pedals to allow the human driver to control it, and AV doesn’t need those. Or mirrors, outside mirrors, and AV doesn’t need those. We’ve got some mechanisms in place to allow waivers, to allow companies to build a certain number of vehicles, but it’s kind of a haphazard approach, and what we need is a more coherent system of regulating and doing… I think what we need actually is a type approval process. Just like human drivers, before we get a driver’s license, we have to go out and take a written test, do a road test, and demonstrate that we have a certain level of knowledge of how to operate a vehicle and what the rules of the road are and that we know how to do it comparatively safely. It’s a very, very low bar for that, and I’ve recently written a blog post that may be up by the time people hear this about that, but it’s at least something.

 

Sam Abuelsamid (24:22):

We don’t have anything like that, any kind of standard of performance for AVs other than what the companies themselves said internally, and I think we need to go beyond that. There’s a lot of people in this industry that I trust to do the right thing, including your boss, but that doesn’t mean, I think, that Argo or Cruise or Waymo or anybody else should be allowed to deploy these vehicles at scale without some sort of external oversight that the systems are in fact safe enough.

 

Alex Roy (25:00):

Moving on, myth number five. Level three automation counts as autonomous. The reason we’re talking about this a few days ago, it was a LinkedIn post from a Daimler leader Ola Källenius, saying that they’re the first company to get official government approval to deploy level three automation. Could you explain what level three automation is, or what you think it is, and why it is or is not autonomous?

 

Sam Abuelsamid (25:36):

One way that you can look at levels of automation without using the SAE terminology is if you think about the elements of your body that you use for driving or to get around. Traditional vehicles, it’s hands on, feet on, eyes on, brain on, hopefully. As you add driver-assist systems, you can start to turn off some of those things. For example, when you’re using cruise control, you’re still brain on, eyes on, hands on, but you can take your feet off the pedals, so it’s feet off.

 

Sam Abuelsamid (26:16):

You go to something like a so-called level two system like GM’s Super Cruise or Ford’s Blue Cruise, now it’s hands off and feet off because they do lane centering and they’re hands-free steering systems, but you’re still eyes on and brain on. You have to be alert, paying attention, and ready to take the wheel at any time. Then the next step up would be eyes off. Now you don’t have to watch the road 100% of the time, but you still have to be alert. You can’t take a nap. You can’t climb into the backseat. You can’t watch a video. You have to be alert enough that within a, some would say reasonable period of time, I don’t necessarily agree with that, within like eight to 10 seconds to take back control of the vehicle, that eyes off, that’s what we’re calling level three.

 

Sam Abuelsamid (27:13):

That’s what Mercedes’ Drive Pilot system that they’re launching right now on the S-Class and the EQS in Germany is. It’s brain on, eyes off, hands off, feet off, so you don’t have to watch the road. If brain is on, I don’t consider that autonomous. Only when you don’t have to be paying any attention at all and don’t have to be prepared to take control at some point is it truly autonomous. This is one of the areas where the level three… The way the SAE stuff is set up, level zero, one, two are considered driver-assist. Level three, because it’s eyes off, they count that as autonomous or automated. They don’t actually use the term autonomous. They use automated, which I actually prefer that term. I prefer the term automated than autonomous. I personally don’t consider that automated. I consider that a driver-assist system.

 

Alex Roy (28:13):

Well, I’m so glad to hear you say that, because a lot of people, including my mother who lives in Germany said to me last week, “Oh, can I get one of these Mercedes-Benzes that’ll drive me around?” But I’m like, “Actually, no one can, because it’s not what they do.”

 

Sam Abuelsamid (28:27):

Well, another thing about those systems, they only function at speeds up to 37 miles an hour. It’s designed to be a traffic jam pilot. If you’re stuck in a traffic jam on a highway, not even in a city traffic jam but on a divided highway because it has to be geofenced, up to 60 kilometers per hour, 37 miles per hour, it will take over the steering control. You don’t have to be paying attention, but as you approach that speed threshold or want to leave that particular road and go onto city streets or something else happens, you have to be ready to take control at any time.

 

Alex Roy (29:05):

The thing I don’t understand, and this has been plaguing me, is whenever people talk about this Mercedes system or any level three system, no one ever says, “Oh, don’t worry. You’ve got no liability.” If you get into a Waymo or an Argo autonomous vehicle, you get in the back, it’s like getting in a human-driven taxi. I don’t have any liability when I get in the back of a taxi. It’s the driver. It’s the service. It’s whoever provides a taxi. Has Mercedes said anything about who’s got liability if you use their traffic jam pilot and you’re in the car?

 

Sam Abuelsamid (29:40):

Yeah. They have said while the system is active and functioning, Mercedes takes responsibility for it. “We have liability.” So if a crash happens while the system is functioning, you are not responsible for that. They are responsible.

 

Alex Roy (29:56):

Now, are they the first company to offer such a system and make that claim?

 

Sam Abuelsamid (30:00):

They are not the first company to offer such a system.

PART 2 OF 4 ENDS [00:30:04]

Sam Abuelsamid (30:03):

They are not the first company to offer such a system. Honda, earlier this year, launched a similar system in Japan. It’s only a hundred vehicles.

 

Alex Roy (30:11):

You can’t buy it.

 

Sam Abuelsamid (30:11):

It’s very limited. You can buy it. You could buy one of a hundred examples of the Honda Legend, the former RLX. They discontinued here. I’m not sure what they have said about liability. Previously, Volvo has said when they launch systems that are at least level three, when the systems are functioning, they will be liable. They haven’t yet launched such a system, but they have said that they will be liable. I think they were actually the first automaker to do that.

 

Alex Roy (30:44):

Hm. It’s a very Volvo thing to do.

 

Sam Abuelsamid (30:46):

Yeah.

 

Alex Roy (30:47):

All right. Myth number four. This one makes me crazy. Pedestrians and cyclists will need beacons.

 

Sam Abuelsamid (30:56):

No, absolutely not. One of the key factors for an automated system is that, just like a human driver, it needs to be able to see around its environment, look at its environment, and understand what is around it, and then respond accordingly. So it has to have a semantic understanding to be able to distinguish between a car and a cyclist, and a pedestrian, and a dog, and a plastic bag that a gust of wind is blowing across the street. It needs to be able to distinguish all of those things and respond accordingly, just like we do.

 

Sam Abuelsamid (31:32):

Now, that said, that doesn’t necessarily mean that we shouldn’t have beacons, particularly on bicycles. That can help, just as I think that it’s also a good idea to have V to X communications to allow vehicles to communicate with each other and signal their intent to each other. All of these things are extra layers of safety that don’t detract from the automation. I mean, again, as humans, you and I come to a four-way stop, and we arrive at the intersection at the same time. We use nonverbal communications. We give hand signals.

 

Sam Abuelsamid (32:20):

Or you’re trying to pull out of a parking lot into a crowded street, and I stop just ahead of the driveway. And I wave my hand and say, “Come on out. Go ahead.” We use those kinds of communication signals to signal our intent to each other and what we’re going to do. I’m telling you that I’m going to wait here and let you pull out of that driveway so we can keep things moving. Because if I pull ahead another 10 feet, I’m not gaining anything, so I might as well let you go.

 

Sam Abuelsamid (32:53):

Doing the same kind of thing with automated vehicles, I think, is a great thing. But if a vehicle is operating in an environment where there’s no other vehicles around, or may not be any connectivity, it has to be able to understand its environment and still operate safely, even if the connectivity is not there. So we’re talking about extra layers of capability to provide more situational awareness. If bikes had beacons on them that broadcast their position, their trajectory, that would be helpful, especially because most of the sensors that we have on vehicles are limited to line of sight. It doesn’t matter how good Argo LIDAR is, it still can’t see through that truck that’s in front of it or that car that’s parked on the curb.

 

Sam Abuelsamid (33:44):

And so, if a person’s smartphone in their pocket has a beacon that’s broadcasting their position and their direction, that is a signal to an AV before it can see that person, before the sensors can actually see them that, “Hey, there’s some there. I need to watch out for that.” So you don’t have to have beacons. The systems have to be able to operate safely without them, but adding them, I think, is a good thing.

 

Alex Roy (34:13):

Yeah, I agree. I don’t know where this beacon thing came from, this straw man. I hope we don’t have to wear beacons. I wouldn’t want to rely on a system that required people outside the vehicle to have beacons because if I have a cell phone, that’s my beacon. And my battery goes dead. I wouldn’t want to ride down the street knowing the vehicle relied on my cell phone because my phone often dies. I use it too much. All right. Myth number three, it’s along the same lines, autonomous vehicles need road lines painted.

 

Sam Abuelsamid (34:48):

We can operate without road lines. We do it all the time. But again, like the beacon thing, the more information you have, the smarter decisions you can make, hopefully. And the reality is that as much progress has been made with AI, in most respects, it’s still not as good as human cognition. So the systems have to be able … if you’re driving down a gravel road, for example, or in a residential neighborhood, most of the time, residential neighborhoods, I know mine doesn’t have any painted lane markings on the road. But you need to be able to detect where is the edge of the road so I’m not driving across somebody’s lawn. I want to stay on the pavement. But I also need to know, okay, where’s the opposite edge of the road, and where is oncoming traffic coming?

Sam Abuelsamid (35:40):

So you don’t have to have painted lane markings, but what you do have to have is knowledge of what is this road that I’m on? Is it a one-way road, a two-way road? Is it a residential road? Is it an urban center street? So that you can understand what are the things that you can do? What are the behaviors that you should follow in that particular environment? So if I’m driving through a residential neighborhood to drop somebody off at home or pick somebody up, or drop off a package, I want to stay on the right half of this road as much as possible, except to move over to pass a car parked up by the curb and be able to safely navigate this environment. And maps, having built up maps of an area, are a great assistance to that.

 

Sam Abuelsamid (36:40):

There may be times when you may have to operate in a place where the map is either non-existent or out of date, and you still need to have the basic understanding of the rules of the road. I mean, we drive into neighborhoods all the time where we may not be intimately familiar with it, that we don’t drive on all the time. But we know what to do in a particular environment. And but, again, like with the beacons, having that upfront knowledge helps the system make smarter decisions along the way. So it doesn’t have to guess as much. That’s the thing. The maps are more important than the lane markings.

 

Alex Roy (37:22):

Myth number two: camera-only autonomous vehicles are the ultimate goal.


Sam Abuelsamid (37:30):

I think maybe if you’re talking about a purely economic perspective, trying to make the cheapest possible system, that is one potential goal that you want to get to because cameras are the least expensive sensors that we have. They are certainly, I would say, probably an essential part of any AV sensor suite because those are what we use to do most of the classification of what’s around us to get that semantic understanding, understand and distinguish different types of objects.

 

Sam Abuelsamid (38:10):

That said, I am extremely skeptical that, at least in the foreseeable future, we will get to a point where a camera-only solution is going to be sufficiently safe and robust to operate, certainly not as a level five vehicle. I think level four, where one of the restrictions is perhaps daylight hours or some other good weather conditions, is probably the limit of where we’ll get in the foreseeable future with a camera-only system. Because one of the challenges with cameras is cameras are passive sensors.

 

Sam Abuelsamid (38:54):

We got two kinds of sensors, passive and active sensors. Passive sensors are just taking in the information that’s coming to them without any reference point of where it’s originating from. And that makes it very hard to do accurate distance measurement to get depth perception. There are some ways around that. There are companies like Light that are doing multi-vision systems that spread the cameras further apart so you have more parallax. So you get more of that stereo vision capability that humans have, or most animals have. And they can do fairly accurate distance measurement with that. But again, there are limitations on what you can do with a camera.

 

Sam Abuelsamid (39:35):

You can also do things like take off the IR cut sensor so you have more low light visibility than you have with just a standard camera that is generating the kind of images that we want to see. Humans don’t actually need to see what the cameras are seeing, but you want to be able to detect things in the dark. But active sensors that are generating some signal that is sent out and reflects back so that we have a reference point of where it started from like radar, like LIDAR, those, I think for the foreseeable future, are going to be an essential part of the sensor suite so that you can get that accurate distance measurement. And not just distance, but also orientation of the objects, because you need to know, not just how far an object is away from you, but what angle is it at relative to you?

 

Sam Abuelsamid (40:33):

From the conversations I’ve had with Bryan and from many others as well is one of the biggest challenges in developing an automated driving system is not just the perception, but the prediction. Because once you figure out what’s out there, then you have to predict what they’re going to do in the next few seconds so you can plan a safe path through that environment. And you can’t do accurate predictions if you don’t have an accurate picture of the orientation of an object. If a car in the lane adjacent to you and ahead of you is parallel to you, then you can reasonably assume that at least in the next second or two, it’s probably not going to change lanes. But if it’s at a slight angle, then there’s a much higher probability that it’s going to pull into your lane, and you have to respond accordingly. So that’s where the active sensors, high resolution, active sensors, can be hugely valuable to you.

 

Alex Roy (41:35):

I was trying to answer the same question for my own mom, and my answer was, “Well, do you want something that’s just good enough or the best we can do?” And the history of all progress and technology and innovation has always been you want to see what the best is we can do. And the best always moves the bar forward. So if my daughter … I’m putting Coco in the car. I’m not settling for good enough because I don’t know what that is.

 

Sam Abuelsamid (42:02):

Totally agree.

 

Alex Roy (42:03):

Yeah, not good enough. All right. Myth number one. Actually, we could do 20 more of these, but this one’s really dear to my heart. The average vehicle owner is good enough to be a self-driving technology tester.

 

Sam Abuelsamid (42:18):

Absolutely not. Definitely not. I spent half of my career as a product development engineer in the auto industry. And I’ve spent a lot of time behind the wheel of many different kinds of vehicles, testing, ABS, traction control, stability control, electro-hydraulic braking systems, other stuff. And it’s trying to drive and capture data and look for the subtle anomalies that are indicative of problems all at the same time. It’s really, really challenging. And to be able to capture the information and to translate what that means and provide feedback to others that you’re working with of what just happened and what was the situation, I think … and at the same time, do this safely.

 

Sam Abuelsamid (43:25):

Most of the testing work that I ever did was on closed test tracks. We did a lot of work on the test track before we ever went out onto public roads to try out the same thing. We had to have a very high degree of confidence before we tested it on public roads. And doing that either on the track or on the street, and especially on the street, because on the street now, you’re testing stuff that … you’re running an experiment where you’re surrounded by other people that are using the road, other road users, pedestrian, cyclists, drivers that did not consent to being part of that experiment. They’re not even aware that they’re part of an experiment.

 

Sam Abuelsamid (44:13):

And when you’re talking about a system where, when it goes wrong, the consequences are potentially life and death. That is irresponsible to put in the hands of people that you have not done a significant amount of training with, to be able to look for the anomalies and intervene before bad things happen so you’re not putting others at risk unnecessarily.

 

Alex Roy (44:46):

That was a beautiful and epic explanation of why a lot of people just shouldn’t be driving cars at all, I think.

 

Sam Abuelsamid (44:54):

That’s true.

 

Alex Roy (44:56):

I mean, if you go to certain countries, the licensing requirements in some European countries are extraordinarily serious. And in the United States, there’s …

 

Sam (45:03):

… in some European countries are extraordinarily serious. In the United States, the gating is not that, it’s not that tough.

 

Alex Roy (45:08):

But what’s interesting, and I think a lot of people overlook, is despite the fact that it’s actually not that hard to get a driver’s license in the United States or North America, Canada’s much the same way, despite the fact that it’s pretty easy to get a driver’s license and that this is actually a pretty challenging task that we do often on a daily basis, all things considered, humans are actually pretty good at it. I mean, we drive 3.2 trillion miles a year, have roughly six and a half million crashes, which amounts to about one crash every half a million miles. For the average driver, that works out to about once every 30 years or so, and that’s not fatal crashes, that’s all crashes.

 

Alex Roy

Most people go through their entire lives without ever being in a crash. Despite the fact that we have fairly limited training, the human brain, for all of our flaws, and we have many of them, is actually remarkably good at doing this very complex task in extremely varied environments. This is something that in the kind of variability of environments that we operate in, that computers still have not achieved. We’ve gotten it to the point where in certain limited environments they can approach or surpass what humans can do, but those are often the environments that are also the easiest ones for humans to deal with. The bar that humans actually set is surprisingly high. Almost 40,000 people a year die on the roads in the United States alone and that is a tragedy.

 

Sam (46:56):

That has gotten worse this year.

 

Alex Roy (46:57):

I mean, it was around 37,000-38,000. It’s gone up to about 40 this, we’re on pace for about 40 this year. That is an enormous tragedy. That’s way too many people. We need to do better. But considering how much we do this task of driving is surprisingly actually surprisingly low. If you look at the statistics going back to the 1960s, since the 1960s, the rate of fatalities on the roads has dropped by almost 80% in terms of fatalities per hundred million miles. It’s dropped by almost 80%. We have made enormous progress.

 

Alex Roy (47:37):

But just as with all things in engineering, the first 90% takes 10% of the effort, is kind of the traditional rule of thumb and the last 10% takes 90% of the effort. It’s actually, in something like automated driving, it’s probably closer to 99.1. It’s those edge cases that out of all the driving we do, those edge cases are the ones that tend to happen least frequently and are the toughest to deal with. That’s the challenge, that’s why we need people that know what they’re doing evaluating these systems and developing these systems to do it safely.

 

Sam (48:25):

All right. Would you like to make any predictions for 2022 because I’ve got one, but you should go first.

 

Alex Roy (48:37):

I think the easiest one, I think we will see more pilot deployments of AVs in more locations in 2022. I think that’s a pretty easy one to make. I think Argo has already has previously said they’re going to do some robo taxi and some delivery stuff before the end of the year so you’ve got a couple of weeks left to get that done. I’m sure we’ll see more of that in 2022.

 

Alex Roy (49:06):

I think we’ll also see more driverless commercial operations in 2022. But again, as I said earlier, still very limited locations, limited scenarios, and I think one of the things that is going to be interesting to watch is I think we will see a lot more, we’ll continue to see more emphasis on deliveries. I think goods delivery is still a key area where both from a technical and safety and economics standpoint, a lot of this technology makes sense and is going to be easiest to deploy in the near term.

 

Sam (49:49):

All right. My prediction is that 2022 is going to look like 1889-1890 for the elevator business. Otis, he died in 18, six, oh, what year it was ’61 and it was his sons who rolled up the sector and bought a bunch of other companies and made Otis into the company it is today. I think we’re seeing that in autonomous vehicles, but a lot more compressed. I think you’re going to see a lot more M and A, and we’re going to go down to I think it was, there were hundreds of AV companies a couple years ago. I think we’re going to go less than a dozen and maybe less than five or six the next year or two. I think that things are going to rapidly accelerate as people understand that it’s not one winner take all, it’s a winner’s take most. I think we’re going to see the contours of that take shape. I’m sure you’ll be talking about that with us in 12 months.

 

Alex Roy (50:43):

I hope so.

 

Alex Roy (50:44):

Bonus myths I didn’t have on the top 10 list, but going to ask you anyway. The real product is private ownership of AVs. I mean, what’s that about?

 

Sam (50:57):

I don’t think it is. I think we will have increasing degrees of automation in privately owned vehicles. But I think when it comes to true AVs that can go, that are restricted but less restricted, not just limited to highway driving, things like that, I think that’s for a lot of reasons, that’s likely to stay the domain of services model.

 

Alex Roy (51:28):

This notion that everyone’s going to have an autonomous vehicle and it solves all problems, I don’t buy it because as a New Yorker, the last thing I want is everybody to own a car which takes up so much space as their old car, but it drives itself. I think that could actually be worse.

 

Sam (51:44):

I absolutely agree. If we just make all the vehicles autonomous, it will make the congestion problems worse because now all of a sudden you’ll have more people that want to use them to go places instead of using alternatives like micro mobility or public transit, just walking.

 

Alex Roy (52:01):

Just walking.

 

Sam (52:03):

Then you’ve got these vehicles deadheading around trying to find parking spaces. I think it will make it worse. We actually need fewer vehicles used more effectively and more efficiently.

 

Alex Roy (52:14):

The other bonus myth that almost made it to the time top 10 list was automated driving is the panacea that will end crashes and congestion. But I think that the myth actually something that recurs every 10 to 20 years, which is that technology X will solve everything. This is a thing that always comes back. Why?

 

Sam (52:37):

Because I think humans are inherently optimistic for the most part, not always, but they’re generally optimistic creatures. We’ve seen the progress that has been made particularly over the last couple of centuries since the industrial revolution. We think that as we develop new stuff we can make things better. We hope that we can make things better.

 

Sam (53:09):

The reality is when you look at all of the progress we’ve made, yes, we’ve got a lot of things that have made life better for a lot of people, but there are inevitably unintended consequences. We tend to a large degree, blind ourselves to the unintended consequences up front, not pay enough attention to those.

 

Sam (53:38):

This is a big part of what I do is I try to think about what are the … I try to be … This goes back a long way in my career in many of the jobs I’ve had where I’ve sometimes been described as the prince of darkness. I look at what can go wrong with this, because I’ve seen the things that can go wrong. While as an engineer, I do tend towards optimism, it actually I think for a lot of engineers, there’s this strange dichotomy.

 

Sam (54:08):

I wrote an essay about this many years ago, the dichotomy of being an engineer. You want to solve problems, use technology to solve problems, but at the same time, you’re aware of the things that can go wrong. As a designer, engineer, whatever, if you don’t think about those things that can go wrong, you’re being irresponsible and you need to factor that into what you’re doing, look at how the technology can be misused, regardless of whether it’s automated driving or a social network. You have to think about what bad thing can happen with this and try to mitigate the bad as you roll out the good.

 

Alex Roy (54:57):

That reminds me of something someone said to me recently about being a parent. [inaudible 00:55:00] I just being a parent. We are all the parents of the future. In other words, we’re trying to raise the future to be as good as it can be, and that’s a good thing. That’s just human nature.

 

Sam (55:11):

Absolutely. You want things to be better for those that come after you.

 

Alex Roy (55:15):

Sam, you’re the lead transportation analyst with Guidehouse Insights.

 

Sam (55:20):

I am a principal analyst for E-mobility Ecosystems and Automated Vehicles with Guidehouse Insights.

 

Alex Roy (55:27):

Your title gets more awesome every year. Of course, the founder co-host of the will

 

Sam (55:33):

Wheel Bearings podcast, wheelbearings.media, and also actually my colleagues and I on the transportation team that guide us have a new podcast, a roughly biweekly show that we’re doing, we’re called … You can find it at ghtransportcast.com. For a while now, as part of our research, we track the stories in the areas that we cover from electrification, aviation, automated driving, micro mobility and every couple of weeks we gather to discuss the top items on that news tractor that we have internally. We decided to start recording that and putting it out as a podcast.

 

Sam (56:14):

My colleagues, Scott Sheppard, Ryan Citran, Christian Albertson and Saji Ebinada and I, we do that every two weeks. It should be another one out later this week.

 

Alex Roy (56:26):

Sam, thanks so much for coming on. I’m a fan and subscriber for many years. Thanks so much.

Sam (56:32):

It’s always a pleasure to talk with you, Alex.

 

Alex Roy 

Well, always fun to have Sam on the show. If you enjoyed today’s episode, please connect with us on social. We’re on Twitter at No Parking Pod and I’m everywhere on social @AlexRoy 144. Please share No Parking with a friend. Like us, subscribe, give us a five star review wherever you listen to your podcasts. This show is managed by the Civic Entertainment Group. Until next time, I’m Alex Roy, and this is the No Parking podcast.