A future filled with autonomous vehicles promises to be a driving utopia. But as today’s guest asserts, on the back of her extensive research the implications of a huge increase in autonomous vehicles on our streets aren’t rosy by default. Sarah walks me through what the various implications are, and how local governments and AI practitioners can partner on policy and technology to create a future that works for everyone.
A future filled with autonomous vehicles promises to be a driving utopia. Maximum efficiency navigation decreasing traffic and congestion, safety features that drastically reduce collisions with other cars, bikes, or pedestrians, and an electric-first approach that lowers greenhouse gas emissions.
But as today’s guest asserts, on the back of her extensive research the implications of a huge increase in autonomous vehicles on our streets aren’t rosy by default. Sarah Barnes works on the micro-mobility team at Lyft, and has published a variety of works that document the expected implications of more autonomous vehicles in major metropolitan areas— implications that are good, bad, and ugly. Sarah argues that without a serious focus on three transport revolutions—making transport shared, electric, AND autonomous, congestion and pollution could be here to stay. Sarah walks me through what the various implications are, and how local governments and AI practitioners can partner on policy and technology to create a future that works for everyone.
[00: 00: 19] RS: Welcome to How AI Happens, a podcast where experts explain their work at the cutting edge of artificial intelligence. You'll hear from AI researchers, data scientists, and machine learning engineers as they get technical about the most exciting developments in their field and the challenges they're facing along the way. I'm your host, Rob Stevenson, and we're about to learn how AI happens.
A future filled with autonomous vehicles promises to be a driving utopia, maximum efficiency navigation that decreases traffic and congestion, safety features that drastically reduce collisions with other cars, bikes and pedestrians, and an electric-first approach that lowers greenhouse gas emissions. But as today's guest asserts on the back of her extensive research into autonomous vehicles, the implications of a huge increase in autonomous vehicles on our streets aren't rosy by default.
Sarah Barnes works on the micro mobility team at Lyft, and has published a variety of works that document the expected implications of more autonomous vehicles in major metropolitan areas, implications that are good, bad and ugly. Sarah argues that without a serious focus on three transport revolutions, making transport shared, electric and autonomous, congestion and pollution could be here to stay. Sarah walks me through what the various implications are, and how local governments and AI practitioners can partner on policy and technology to create a future that works for everyone.
[INTERVIEW]
[00:02:00] SB: Hello, AI world. Today you're talking to a geographer and city designer. So I know that my background is maybe a little bit atypical for this particular podcast. But I've always focused on transportation technology, which has led me to be very interested in the world of autonomous vehicles. Particularly when the media started first really picking up on them probably in around 2015, I started to do some research there, mostly because for the first time, in a long time, we were getting a lot of media attention saying this was going to completely revolutionize transportation, and it was going to solve all of these major problems that we were seeing across the board. From congestion, to traffic fatalities, health and well-being problems, autonomous vehicles, we’re going to be the solution that we could wrap around our cities and solve major problems.
So I ended up writing my master's thesis on autonomous vehicles and the impact that they would have on the City of London and the UK, and the specific context where autonomous vehicles could either help or hinder the urban transportation system based on the different types of urban topology. Following that, I co-authored Siemens’ whitepaper and autonomous vehicles, which was called Cities in the Driving Seat. And that was about how local regulators could really lead the charge in terms of defining how EVs are applied in specific urban context to actually meet their urban agenda needs and trying to get to the outcomes that we say that we're after.
Alongside this, I'm a bit of a research net. So I've researched everything from EVs in China, to disaster risk reduction globally. I've also been published by the UN, through their disaster risk reduction department. I'm a big climate change advocate as well, or rather, climate change activist advocating for not a lot of climate change. And so that was part of their making cities resilient campaign.
Alongside this, I have a newsletter that is all about autonomous vehicles, that now has hundreds of subscribers from every corner of the world who constantly check my spelling, which I appreciate. And the intention of that newsletter was partially based on the experience I had doing research around autonomous vehicles, where it's working with industry experts and hearing their perspectives and realizing that a lot of people in government didn't really feel that they had the technical expertise to understand how to regulate autonomous vehicles and what they needed to do.
And on the other side, industry experts from automobile OEMs and tech companies who had sort of disassociated what the actual problems they were solving from the systems that they were trying to solve for. So you had people who really understood how the mechanics of a car or AI and machine learning worked, but didn't necessarily understand how transportation systems work and network impacts there.
And so the newsletter is sort of a bridge building exercise to pull in new research, and articles, and policy perspectives and industry news so that each of these groups can better understand what's happening in the space.
I love thinking about autonomous vehicles, but in terms of the work that I feel most attached to actually seeing through. A couple years ago, I got pulled into the world of bike sharing while living in the UK. And so after a number of years working in bike sharing in the UK, I moved to the States, joined Lyft, joined their bike sharing team. And I now work on government partnerships here to help bring cycling to the masses and sort of lower barriers to accessing cycling as a mode of urban transportation. Now might be a good time to also say that I do not work on Lyft’s autonomous vehicle team. I don't work on their policy team. Disclaimer, disclaimer, disclaimer. These opinions are my own. They're good, they're bad, they're ugly, and they're my own.
[00:05:59] RS: It seems like an oversight on Lyft’s part. I mean, do they know about this background you had? Do they know about this newsletter? It feels like you could contribute pretty easily over there.
[00:06:07] SB: I mean, I think there's definitely folks know about it. But I think, for me, it's just how I choose to put my time into the world. And I think, for me, I love AV policy. I think about it a lot in my spare time. The newsletter gets published weekly. So I'm constantly reading through policy updates and industry updates. But I think it's a really complex industry that still has a long way to go before we really get to some solutions-oriented implementation. And so, right now, still bikes for me.
[00:06:37] RS: Because you've done all this wonderful research, could you maybe rattle off a handful of examples of ways you anticipate the increase in autonomous vehicles on our streets in our cities making an impact? What are some of the good, the Bad, and the ugly things that we can expect?
[00:06:53] SB: Yeah. One of the things on the positive side that we tend to speak a lot about is the potential for autonomous vehicles to bring transportation mobility options to people who have historically perhaps been disadvantaged from current systems. So one particular group that has been at the forefront from the very beginning from the very first Waymo media public press release around autonomous vehicles was people with disabilities, and particularly people who are blind and would never be able to access something like a driver's license, which in America is seen as a huge symbol of freedom and autonomy.
And yet, for the most part, we've yet to see solutions that would actually serve people who have physical disabilities or who currently can't drive. The current version of autonomous vehicles that are out on the road today and that people are talking about are vehicles that you would likely need assistance getting in and out of and setting route destinations and understanding when you've arrived at your destination. And should you get out on the left or the right? Is there a curb on the other side? When you get out of the vehicle, in what direction should you be walking to actually get there? So if you can walk as well. Like there's a ton of information that we think about, and I think there's potential for autonomous vehicles to really serve a large population of people who've been disadvantaged by the way the current system is designed. But that's only going to happen if the devices and the vehicles themselves are designed with that population. And it's not with that population in mind. It's like actually bringing people in to the design process and saying what does the solution for you specifically look like?
And right now, what we're seeing from a lot of autonomous vehicle providers is that they're trying to design solutions that will be able to grow at scale, which means they're trying to market products that are for mass consumption, and mass consumption, similarly to what we have today, does not necessarily place people with disabilities at the heart and center of it.
As with everything, the good can come, but it has to be intentional. And so I think, for me, the positives of having electric, shared shuttles are huge. There's an entire research center out of UC Davis called the Three Revolutions, and they define the three revolutions as shared, electric and autonomous. And their research overwhelmingly shows that the electrification of vehicles is going to have immense cost savings in terms of greenhouse gas emissions and air pollution.
But they weigh that off with, if you make it too easy to drive, we're just going to end up increasing the miles that we go. And so if we're not actively committed to the shared and the electric portion of these three revolutions, then we see the exact opposite trend happen, which is that we only increase greenhouse gas emissions. We only increase air pollution even with just driver assistance technology. Not even complete automation. Drivers drove on average an extra 5000 miles than their non-assisted counterparts. And it was because they were willing to spend more time in traffic. And it was because the burden of being in traffic was less stressful for that driver personally. It really depends on how we apply the solution and if the intention is to commit to all three revolutions, or if we're just committing to the one.
[00:10:34] RS: Does this mean that until we see level five automation, which is fully autonomous, right? Like, no driver assistance. You get in your car, and you can play tiddlywinks instead of paying attention to the road? Will we need to see that level of automation before we expect to see any sort of impact, adverse or otherwise?
[00:10:56] SB: No. I mean, I think we've already seen a lot of these impacts already. So again, like, basic Driver Assistance, which is probably a level one, level two, depending on what type of vehicle you're in, is already showing that it's changing driver behavior, for drivers who they're more comfortable in traffic, and thus, the cars are traveling further and more.
We've already had our first death with Elaine Herzberg, who was hit by an Uber-operated autonomous vehicle in Arizona in 2018. And so if you were to ask her family, I'm sure they would say the impact of the technology is here now. There're tons of debate currently happening right now about how the federal US government regulates autonomous vehicles and who they give power to. Is it a state level decision to regulate autonomous vehicles? Is it a local city level decision to regulate autonomous vehicles? Can the federal government do it?
And it's an incredibly tricky problem, because you have a lot of companies who are trying to test this innovation. And from their perspective, the only way that the technology is going to be safe enough to be in public and to exist in the way that cars currently exist on our roads today and in our lives is if we can test autonomous vehicles in that same environment, if we can put them up to that same test and see how they perform.
One of the issues is that while the companies want to be doing using cities as its testbed for innovation, there's not a lot of transparency about what is happening with the vehicles. So Waymo is currently suing the California DMV for wanting to publish information about any type of hiccup, really, that the vehicle experiences. And the DMV is very much so in favor of saying, “Look, you're using the city as a testbed. And we should let the public know precisely how good your technology currently is so that they can also make informed decisions about if they want to be driving beside you, or if they want across the street when they want to cross the street.” And it's a really interesting time in the autonomous vehicle industry, because the impact on policy is already here, right? Like, we are already having these discussions, and we're trying to decide what is more important, innovation or a safe public realm? Testing, or keeping people informed about the risks that they might experience while being out in the city?
And everyone experiences risk when they walk, when they cycle, when they're in a car. We have infrastructure solutions today that make our cities not only safer for pedestrians, cyclists and public transit riders. We have infrastructure that also makes it safer to drive. And one of the things that I think is poorly managed in today's conversations is the fact that I do a lot of work advocating for bike lanes as a cyclist. And no one tells drivers that, actually, bike lanes are also proven to make driving safer, because you are less likely to collide with a cyclist, or worse, another pedestrian. And it helps keep each road user in their own safe lane with other people who are behaving in a way that they understand and comprehend and can follow.
And so one of the impacts I see most clearly today is how conversations around infrastructure get sidelined by conversations about technology. So we're spending more time thinking about like, “Oh, can we bring this innovative technology and not necessarily – Do we already have a solution for this that we could apply?” And the answer is yes. And the answer requires actual funding and political willpower to bring it through, and things like that. But oftentimes the solutions that we want and need for our cities to flourish and to really address any suite of problems that that we think we have are already out here. They're just not nearly as sexy as an autonomous vehicle.
[00:14:56] RS: The virtue of AV is just sort of assumed, right? It's like, “Oh, it's going to – It's a panacea for all these things,” like you rattled off. And maybe people aren't as thoughtful about what it's actually going to mean when there're entire fleets of these things deployed. What's interesting, though, is there's already fleets of cars deployed, right? Like we already have these homuncelon metal things flying around constantly everywhere you look. So I'm curious, what do you think is the difference? When you think about the implications of there being more autonomous vehicles in cities, how is that different considering that we already have cars everywhere?
[00:15:38] SB: One of the things that I really like to do is also note that our cities were actually not purpose built for cars to begin with. The Cumberland Road, which was built in the 1840s in America, was the first purpose-built road that we had, and it was for wagons, right? Like, for the most part, people can't really imagine North America without car culture. It feels so deeply embedded. And we've sort of forgotten that, for most of human history, cars were not on our roads, and our cities and our roads were not actually designed to move cars.
India started paving gravel roads in the 1830s, as well. Like, these are completely before when we think of Ford Model T being developed in 1908. Like, decades before, we were trying to move people, move commercial goods, the rest of it. And the first paved road was in Detroit, of course, in 1909. So that was Woodward Avenue.
So much of the danger in this conversation comes from thinking through how we want to prioritize our roads and our street based in our urban space for this new version of autonomous vehicles. I can remember, when I first was doing research organizations, like Comcast had come out and said, “We think we need to have fast lanes for autonomous vehicles that are dedicated, so that we can segregate them from non-AV. So that while we're in this in between phase, we've got the safety of having AVs just rolling down their own lanes.” And it becomes this whole issue of, “Okay, well, who has access to those AVs? How do you get access to them? Does everyone else just sit in non-AV congestion life because we've pulled a lane over here? And are we going to continue to design and make decisions that prioritize car movement versus the concept of people movement?”
So I think when you when about transportation planning, when you think about city design, one of the major outcomes from the last 30 years is that people have really tried to shift their perspective from thinking that a successful city is one that moves a lot of cars, to one that moves a lot of people.
[00:17:40] RS: And it's also important to remember that when you say autonomous vehicles, everyone just thinks Teslas and thinks cars, but we need to be imagining public transport. What is then the balance with policy? Is that where you think cities ought to be positioning their focus to encourage autonomous when it comes to maximum efficiency type vehicles like trains and buses?
[00:18:00] SB: Yeah. So there's already autonomous trains in the world. The SkyTrain in Vancouver is one example. And that's been up and running since the 2010 Olympics. And so it's also important to recognize that we can apply technology to forms of transportation that really serve the masses, and not necessarily individuals per se.
Each city ultimately is going to have very different contexts in order to have mass transit that pushes mode shift, reduces car usage, and thus, reduces things like air pollution, and greenhouse gases, and all of the other positive externalities that come with reduced vehicle miles traveled. There tends to have to be good land use policies. That type of transit needs to be supported by high-density housing, and jobs, and cultural amenities. You've got to have cities that are interconnected that people want to move between and, thus, we can start to get these network systems built up that people will use, because it's the most efficient way of getting there.
If you don't have the population density, and the housing density, and the job density, some of the more rural communities, some of our more suburban city centers, things like that, AVs can take a lot of different shapes and forms and can actually start to play part of the solution. There's a lot of interesting companies that are really dedicated towards the idea of creating electric shared and autonomous shuttles that can serve as first and last mile connections to regional transport. And I think, ultimately, one of the most important things that we should all be thinking about collectively is that autonomous vehicles aren't going to be the only solution. They're going to be a part of a wider network of solutions.
[00:19:46] RS: As you pointed out earlier, cities were not originally designed for cars. And a useful measure of effectiveness of a city is how many people they move versus how many cars they move. And the reason that cities are have conceded so much real estate to vehicles – I'm sure there're other podcasts people can listen to about lobbying campaigns and how the large auto manufacturers have managed to eke out so much space. Keeping all that in mind, it strikes me that what's at stake is just round two with the autonomous vehicles of like conceding more and more space to cars in cities, and maybe making it less efficient for a pedestrian or for anyone who wants to get around. Is that kind of how you envision this right now, is that with all of these new cars coming to market, that we're going to get round two of cars for cities?
[00:20:33] SB: Yeah. So cars for cities is definitely one of my favorite topics. There, as you alluded to, is a long history of, in the 50s, like, automobile OEMs, really purchasing up public transport companies under the guise of being another public transport company, and then shifting all forms of public transit to petroleum-based vehicles. It's great. And these companies were actually convicted for doing that, because it was monopolizing the transit system. So there's quite a great history there in terms of what that version of it is.
And I think, today, when we think about this new round of city versus kind of auto OEM, or technology, or however we want to look at it, people really debate it in two sides, the first being, we're going to end up dedicating a lot more space and policy work to autonomous vehicles, and it will result in an increase in infrastructure that is car-oriented.
But one of the things that's quite clever that a lot of companies are doing is talking about the fact that, “Oh, if we bring autonomous vehicles here, it's going to be so much more efficient to move people. We're not going to need nearly as many parking lots. We're not going to need the street parking.” Like, these are the main things that people focus on. And they say, “Oh, we're not going to need that anymore, because not everyone is necessarily going to own their own AV.” That argument very much so relies on we're going to free up urban space for other pieces of infrastructure. So you can build housing, and you can build schools, and hospitals and other pieces of really important social infrastructure. And autonomous vehicles are part of this revolution that is going to really benefit urban areas from the amount that they will reduce that type of infrastructure.
That is, you can kind of see how both arguments could potentially come forward. And I think what's really important, and kind of back to some of my previous points, is that, actually, if we really wanted to, we could be removing space for car infrastructure now if we really want to and converting that to those other uses. And that, in turn, would make driving really annoying, quite frankly.
And one of the best ways to convince drivers to get out of their cars and to use other modes of public transport is to actually agitate them. I'm sorry to every driver who's listening to this and realizing that a lot of climate policy is actually based on you being agitated. But it's very much so true. Driving is always convenient. We will always get more driving.
[00:23:11] RS: Yeah. Yeah, the opposite is true, too. There's that strange – I mean, it's not a strange phenomenon, that when you add a lane to a highway, you don't make it faster for anyone. You just get more cars.
[00:23:21] SB: Exactly. So that concept, the concept of induced demand, where it's essentially if you bring pizza to a party, people are going to eat the pizza. If you add an extra lane of highway to a highway, people are going to say, “Oh, great. There's an extra lane of highway. I'm going to go take the highway because it's going to be less congestion now.” You actually end up drawing more people to that piece of infrastructure by car, and thus create more congestion. And this has been seen – This is a global phenomenon, too, right? Like this is not just North America. This has been studied the world over, and widening highways, which is done to reduce congestion results in more congestion and more traffic.
[00:24:06] RS: So then the idea that more autonomous vehicles would free up city space for other activities, or the increase in efficiency would mean we don't have to have four lanes in a major road in the city. We could have two. And then the other space could be given back to, as you know, storefronts, or sidewalk dining, or trees, parks, whatever. That assumes that we nail the shared part you were talking about earlier, right? Like we have to encourage people to not want their own autonomous vehicle. Is that reasonable? Is that likely?
[00:24:39] SB: I think it's the preferred outcome from a social perspective. People who are transport planners and kind of trying to hit on some of these goals would argue that sharing is definitely a very important piece of this puzzle. The sad news is that, for the most part, each of these companies has now in invested billions of dollars in this technology, and they need a return on investment from that. And the way they get that is from having more people pay for travel. And so that could look like purchasing a car outright and paying a license fee similar to how Tesla does today.
Even if it's a case where someone doesn't necessarily own the vehicle, it's maybe a ride share type environment where people are spending money as if it was a ride hailing service. Companies are going to try and orient people to these cars. They are going to want people to be in these cars as much as possible to get that return on investment. Inherently, unless it is shared, and unless it is electric, it's probably going to backfire on us on a societal level.
And in order for shared to really pick up, it's, again, going to have to circle back to land use and density and cultures that are willing to share. And for the most part in North America, single occupant vehicles are everywhere. I am a nerd. So I like sit at intersections when I have like too much spare time and count how many vehicles go by with only one person inside of them. And it's a completely inefficient use of space. People are not really used to sharing their transportation mode, unless they take public transport.
And so the idea that we're just all of a sudden going to shift an entire culture away from being a single person in a car towards being one of four people in a shuttle, one of six people in a shuttle, one of 50 people on a bus, that is going to require an entire cultural shift. And, again, the people who have sold us this idea that vehicles, personal vehicles are freedom, and that this is a huge milestone. Our sweet sixteens are designed around getting cars, right? Like the people who really sold us that message or the car companies themselves. So the idea that they're going to all of a sudden change their tune now and tell us, “It's okay. You should share this one. We'll just take less money after we funneled billions of dollars into this,” just does not seem very realistic.
[00:27:20] RS: What can cities do then? What can one do as perhaps a voter in a city? Or what sort of policies are there that cities can enact to make sure that autonomous vehicles work for them in the future instead of the other way around?
[00:27:33] SB: On the one hand, one thing that city governments can do is that they can go to their state regulators, and they can work with federal government to say, “You know what? I think I want to be in control of this.”
Cities manage all of their transport systems for the most part, but cars tend to be regulated at a state level, right? Like, each state has their own department of motor vehicles and things like that. So cities really need to start advocating for the fact that a state might regulate and say, “Okay, you get the stamp of approval to operate on public roads.” But cities can also come in and say, “Okay, if you actually want to start providing services in the city, I'm going to have a permit application process, and you're going to have to jump through these hoops to meet specific safety standards. And if you can't meet these safety standards, then I'm not necessarily going to have you operate in a public environment in the city. And so there's certain things that they can control and regulate for.
But the other thing, too, that cities can focus on is creating really comprehensive innovation plans that don't just see innovation as an autonomous vehicle, but look at it from a light rail perspective, and look at it from a bus perspective, and look at it from a walking, cycling, and hard infrastructure perspective. Thinking about other infrastructure that is going to also meet their needs.
One of the things that I constantly say is that if you were to ask a person with a transportation planning background, how do we decrease congestion in cities? There'd be a ton of answers around how to shift, mode shift, away from vehicles, things around land use density, access to public transport, funding for active travel modes. If you ricochet down the list far enough, you'd maybe get an autonomous vehicle.
If you ask trained planners as well, how do we reduce road deaths, as an example? You would get a ton of answers that would vary from providing protected infrastructure, changing the curb radii of street corners, raised crosswalks, street trees, lower speed limits. We have a ton of tools in our arsenal that we can apply. These tools will ultimately also make autonomous vehicles better at their jobs too, right? Like, we need to think about not just how we design a better car, but how we design a better transport system.
And cities need to be focused not just on bringing the sexiest new piece of technology to their city. But thinking about how do I protect the people here? How do I lower air pollution? How do I decrease congestion? And focusing on the big system problems. It's really easy to think about the vehicle as a system that we're trying to design for. But actually, the problems that we say we're trying to solve operate at a very different higher system's level of the entire transportation network. And if we only look at solutions in an isolated capacity, such as for the vehicle, we're not actually going to be able to hit our targets at the larger system level.
[00:30:37] RS: On the other side of that equation, from individuals in local government, setting up policies to make sure that AVs work for them, you have the AI practitioners themselves. And in my experience, speaking to them on this podcast, they're very hopeful bunch, they believe not just that they're working on sexy, cool, new technology, but that that technology is going to create a more efficient, safer, better world for all of us, right? Particularly in the case of autonomous vehicles.
So to the AI practitioners out there working in AV specifically, what are some ways that they can go about their work in such a way that it brings about a safer world as opposed to a more congested, less efficient cityscape, which could be one outcome without thoughtful design?
[00:31:23] SB: Yeah. So I think the first thing I would say is having hope is great. Having optimism is great, but also being realistic about how far technology can take us is also really important. And getting really curious about what transportation systems, what cities actually need, and what they're trying to achieve is really important. And questioning if you're designing something at the right system level to actually meet that goal.
You might design the world's best shared electric autonomous shuttle vehicle that has every bell and whistle, is the safest thing we've ever seen that's ever existed. That's a beautiful thing. But it is going to be a fraction of the work that is needed for our transportation systems.
And so recognizing that transportation systems are mosaic, and there is so much work that we need from so many different sides. And that the best thing that you can do as an AI practitioner is to advocate for the best version of an autonomous vehicle, for the version that you think is actually going to solve city problems. And you can figure that out by, first of all, talking with cities themselves about the problems that they face, and what they're trying to achieve, and how your solution can fit in with it. Versus going to a city and saying, “Hey, design, the super cool thing. And you're not going to believe how it works. It's super fancy. And let me show you all of its bells and whistles. And I think it's going to work for you.” Starting with the cities and actually saying, “What's your problem? Does this work? Is this going to solve a problem for you? Do you think this could create more problems? And if so, can you tell me why so I can actually try to design that out?”
[00:33:15] RS: Yep, makes sense. Sarah, this has been really wonderful chatting with you. Thank you for being on the show and sharing all of your research and wisdom with us. I've really enjoyed it.
[00:33:22] RS: Yeah, it's been great to be here.
[00:33:24] RS: And if people want to hear more about the ever-changing world of autonomous vehicles and policy and whatnot, how can they find your newsletter?
[00:33:31] SB: Yeah. So my newsletter is called Along for the Ride, and I write it every week. So you are welcome to come and join me there. I promise to have plenty of spelling mistakes for you to edit.
[00:33:43] RS: I hope you get a bunch more proofreaders out of this episode. Sarah, thank you so much.
[OUTRO]
[00:33:51] RS: How AI Happens is brought to you by Sama. Sama provides accurate data for ambitious AI. Specializing in image, video, and sensor data annotation, and validation for machine learning algorithms in industries such as transportation, retail, ecommerce, media, medtech, robotics and agriculture. For more information, head to sama.com.