The Impact of AI on Transportation: An Interview with Sam Sklar

“There’s a lot of good AI can do to help transportation and infrastructure professionals create value and make people’s daily lives and overall experience better” -Sam Sklar

Transportation is the lifeblood of civilisation. It enables the moving of goods from one place to another and as a result, facilitates trade and economic growth which then ensures an increase in the standard of living for citizens. AI has already begun revolutionising transportation as we know it through; self-driving cars, Intelligent Traffic Management, Public Transport, etc.

To fully understand the scope of AI’s impact on transportation, I enlisted the help of an expert. Sam is a trained transportation planner and a journalist with well over a decade of experience in breaking down transportation infrastructure. In this interview, I hope to pick Sam’s brain on the potential Pros and Cons when it comes to implementing AI in transportation.

Enjoy!

You should also check out Sam’s Newsletter; exasperated infrastructures.

Edem: Hey Sam! I’m super glad to have you. Could you tell us more about yourself, your qualifications and such?

Sam: I’m a trained transportation planner and journalist with over a decade of building, communicating, and critiquing projects and processes. I’m currently an independent consultant and lead author of Exasperated Infrastructures, where I am consistently unimpressed with the state of the United States’ ability to build anything at all, even though we just said we’re going to spend $1 trillion on infrastructure. On what exactly. I’m also a quasi-sceptic on AI as a solution in and of itself to problems we haven’t yet identified, so that should make this conversation fun.

Edem: Let’s Jump right in! When People hear about AI's impact, they almost never think of Transportation. Why is that?

Sam: The one place people currently think of AI and transportation is autonomous vehicles. We think about how cars will one day algorithmically drive themselves. It’s such a huge topic, especially beyond the tech itself. There are ethical and political implications that intersect with the idea of autonomy, especially when it involves economics and safety, too. It’s a whole dissertation to try and untangle the mess. The big theme here is that AVs dominate the space, no one has answers, and we’re expected to just hobble into the future, hat in hand. I think that’s a mistake and problematic.

Edem: I see your point. How do you envision AI will impact the future of transportation?

Sam: There’s a lot of good AI can do to help transportation and infrastructure professionals create value and make people’s daily lives and overall experience better in the OEM world. A few things:

Just because we don’t have fully self-driving cars, yet, that doesn’t mean that training our 4000-pound metal death machines to avoid hitting each other at high speeds isn’t worth pursuing. It very much is. Tech like this already exists today—cars that can park themselves, cars that recognize an imminent crash and can break faster than a human can react, and cars that can, for a short time, perform all the adjustments needed to maintain a car’s position and speed. These are huge quality-of-life improvements that can’t and shouldn’t be ignored. We should expect leaps in these safety and reliability improvements in the coming months and years.

There’s a push in non-car technology coming, too. Think long-haul trucking and transit/trains, and planes. Lots of these modes also already have a lot of AI tech built into their operations, but I can imagine a future where human <> control interaction is minimized.

There’s also a big push for ancillary software to help enhance the bus space. Buses are unique in that they often blend into other street traffic with bikes, pedestrians, trucks, vans, and cars. Even with marked bus lanes, a bus traveller can reasonably expect to experience traffic delays, which in turn cause other knock-on effects, like bus bunching and degradation of reliable service. There’s tech to help make sure traffic laws are enforced—cameras that are programmed, and that can reprogram themselves to learn more about their environment so that agencies can automatically issue tickets to offending motorists, for example (Hayden.ai).

Edem: In what ways can AI-powered data analysis aid in improving traffic flow and efficiency in busy cities?

Sam: It’s about speed and recognition of patterns. Computers that process millions of times more data at millions of bits per second can detect anomalies faster than humans ever could. It’s up to us, the programmers, to determine what counts as a problem and what doesn’t. Let’s take a look at an area of interest: traffic signal timing. Right now, aided by computers, traffic engineers and planners can decide the exact timing of intersectional traffic signals—when lights change, how long each light (red, yellow, or green) lasts, and how long a light’s “cycle” is. They can do this across a whole network to maintain a consistent flow of traffic (in a vacuum).

But conditions change: there’s bus bunching because of any number of reasons, likely congestion; there’s a traffic crash or incident; there’s a parade or other motorcade; any number of reasons. There’s a way to mechanically change signal timing, but a trained AI would be able to adjust accordingly, almost instantly. There’s a huge time/economic value savings that we can capture. The AI-powered light system would be able to, should we let it, prioritize transit or bike traffic flow to ensure that our streets are prioritizing the most efficient vehicles.

I do want to say though that safety is the most important benefit from AI-powered signal timing. The ability to slow traffic to ensure more vulnerable users can use our streets safely is the biggest benefit of all AI-powered transportation tools. Safety first, dignity second, efficiency third, congestion fourth.

Edem: What are the long-term effects of AI on Urban Design and Planning?

Sam: This is a good question. I’d lean toward how we go about collecting and processing data. We already make decisions based on algorithmic collection and processing of data. Statisticians call it regression—best fitting a trend based on disparately collected data—and then make policy decisions based on either the idea that makes the most people’s lives better or affects people’s lives negatively the least. Planners use our experience and expertise to decide and prove what needs to be included and what doesn’t—what variables affect one another—and what variables don’t matter at all. AI will be able to process these variables almost instantly to find the line of best fit that takes into account all available data.

If we let it. It’s still up to us to decide what our values are and it’s up to us to determine that even if some policy is the most efficient if it doesn’t help us achieve our goals as we decide them to be, then it’s all moot.

This is a long road and I’m appropriately skeptical each step of the way.

Edem: How can we reduce the implementation costs of Transport AI systems in order to allow their implementation in poorer countries?

Sam: This is a huge question! We’re still in the “Blue Ocean” stage of development for most of the transportation AI products and services. These technologies have relatively large upfront development costs and require a huge existing infrastructure to help refine and innovate. The tech development will likely continue in established tech markets—the US and East Asia come to mind—and manufacturing will continue globally. As the tech develops, we’ll start to see developing nations dive more deeply into the space to be able to right-size them for their economies.

Right now, however, emerging economies that often have large bus networks, whether formal or informal, are relatively excellent candidates for networking technology that can be added modularly. Personal AVs will remain relatively expensive across the world for a relatively long time, so it’s important that we use all our resources to build the parts that can help developing countries skip levels of development right now.

Innovation can come from anywhere so the more pervasive the ideas and technologies are, the better off we’ll be.

Edem: How might AI assist in designing more inclusive transportation systems for disabled citizens?

Sam: This is also an essential question and I think I know where you’re going with this: operability of personal vehicles for people with low/no vision or people with mobility challenges, like limited limb movement or some form of paralysis. There are huge gains in mobility options and accessibility choices for these people with the advent of AI-powered vehicles.

But AI might be integrated into other forms of ability-granting tech—think wayfinding or getting around. With or without a personal vehicle, people need to navigate a space; everyone’s a pedestrian for some part of their trip. So enhancing mapping features with up-to-date, real-time features might help a blind person make a safe choice instantly. Or building signage that can give real-time traffic conditions to know when it’s safe to walk or cross a street. There might be a more tactile way to engage—curb cuts might be built with AI-powered changes to colour or feel depending on road conditions.

At this point, if we can think it and keep all users of all ages and all abilities in mind, then we can make sure that we’re building roadways for people.

Edem: How do strike a balance between employing the use of AI to design and implement better road networks and preserving the cultural heritage of Cities?

Sam: I’m going to change the last part of your question. I don’t think we’re out here to preserve the cultural heritage of Cities. Part of what makes cities livable is their mutability—we can change things and we can keep things. Part of what makes cities desirable is their history and their ability to help citizens connect to the past. I think preservation is a relative bunk as a lens to see cities through, but I also think too rapid a change can kill a city. So that’s the balance I want to talk about.

I think we have a choice to make relatively soon if we want to go all in on building out our cities to become the hearts of the Internet of Things (IoT) or the epicentres of Vehicle-to-Everything (V2X) connectivity. Putting a sensor everywhere is what we want. Marking more spaces for driving or parking—that’s what we want?

The question is do we want to spend billions or trillions of dollars to remake our cities in the image of “How can this car get around most efficiently?” rather than “How can this person get around safely, reliably, and with dignity.”

The answer is to pour our resources into a system that favours mode choice and emphasizes shared, carbon-neutral and shared, carbon-neutral options. Bikes, buses, trains, walking(!). How can AI help us achieve this goal?

Edem: What are the key infrastructural and Ethical Challenges that could come about from AI integration and how can they be addressed?

Sam: This is related to the last question. If we do decide to sensor everything or dive deep into data mining, there’s a lot of personal safety and autonomy we’re giving up. Even with the most stringent data privacy controls in place, we’re exposing ourselves to that risk and we must be clear about whatever we’re willing to accept it individually and as a society.

There’s also the ethics of empowering an AI to make trolley-problem choices: if there’s an imminent incident where a car will crash into Group A should the software be programmed to avoid Group A, given it will then hit Group B? Which is acceptable? Who says?

MIT has had a program called the Moral Machine up and running for years. It allows any user to see two choices about which group of people are vulnerable to a car that will hit them. The program describes the physical and personal attributes of a certain group—how many there are, age, perceived “value” to society—then asks the user to make a choice: continue and let the car hit Group A or choose to divert the car to Group B. It says a lot about you as a god player, but here’s the larger application. How do we program an algorithm to make this choice when we can’t even do it? Who is responsible for the outcome?

Edem: How can City planners collaborate with AI engineers to ensure that AI seamlessly adapts to different Urban Environments?

Sam: I don’t think AI-powered urban tools will adapt seamlessly to anything. It’s an unnatural harnessing of information that our ancestors could have dreamed of, but would have called magic—right? Anything sufficiently advanced technologically is practically indecipherable from the arcane. But it’s happening anyway and at a relatively breakneck pace.

I think it is about making sure that we don’t lose our humanity in the process of tech roll-out. We still have choices to make about how we, collectively, want the tech to interact with our lives and we have to make sure that we’re in control to the best of our ability what we want and that the opposite is not true. AI-driven-urban-tech is not inevitable, but if we don’t make sure that it’s not, then we’re going to give up our control to actors that don’t necessarily see cities as a place to live, but as a resource to mine.

I think it’s good that it’s hard to dump and run AI into our communities. It should be a thoughtful, challenging decision to make this choice for ourselves, let alone other people. A slow drip is the best course of action, even though the flip side is that there are imminent, deadly problems happening at an increasingly frightening pace—safety, climate, and privacy. What we have to figure out is how to balance our very own trolley problem.

Edem: How do we ensure that AI-powered Transport Systems adapt to the different local preferences and customs of communities and cities?

Sam: I think we have to make sure we understand the words “community” and “city” here. I’ve been relatively flip about the shorthand so far, but what makes “community” or “city” or “public” challenging is that they’re both the sum of their parts and also their individual parts separately; cities are both economic engines and products of them; there are many intersecting publics. They’re all waves and particles and it’s often hard to distinguish between them, especially when we’re talking about innovation that affects more than just the individual. Add in, as you said, culture/heritage differences, and then there’s another variable we have to account for.

This is why it’s essential that we build modular tech that can adapt to local conditions. Context-sensitive solutions are borne from practical design and making sure we’re asking the right questions to the right people, constantly, and we must be willing to drop what we’re doing at great expense if the pathway is one to destruction.

I don’t envy the decision-makers here, but I do want to help advise them.

Edem: Thank you for your time, Sam. Any closing words?

Sam: Yes! Please subscribe and share this interview.