109 | Jason Torchinsky on Our Self-Driving Future

It’s easy to foresee that technological progress will change how we live; it’s much harder to anticipate exactly how. Self-driving cars represent an enormous technological challenge, but one that is plausibly on the way to being solved. What will be the unanticipated consequences when autonomous vehicles become commonplace? Jason Torchinsky is a fan of technology, but also a fan of driving, and his recent book Robot, Take the Wheel examines how our relationship with cars is likely to change in the near future.

Support Mindscape on Patreon.

Jason Torchinsky is a senior editor at Jalopnik. His writing has also appeared in venues such as Boing Boing, Muck Rack, and Mother Jones. He is a producer and occasional guest star on Jay Leno’s Garage, and has been the host of the YouTube series Jason Drives.

• Articles at Jalopnik
• Articles at Muck Rack
• Robot, Take the Wheel: The Road to Autonomous Cars and the Lost Art of Driving
• Twitter

[accordion clicktoclose=”true”][accordion-item tag=”p” state=closed title=”Click to Show Episode Transcript”]Click above to close.

0:00:00 Sean Carroll: Hello, everyone. Welcome to the MindScape podcast. I’m your host, Sean Carroll. As long-time listeners know, we like to mix it up here at MindScape. There’s a lot of different topics we cover, but many of them are a little bit serious, academic, highbrow kind of things, but we also like to have some fun now and again, we’ll talk about cooking or TV shows, or whatever. Today is one of those podcasts, we’re gonna have a little bit of fun. Now the topic you wouldn’t guess is one of the weird ones because the topic is very MindScape-y, ‘Autonomous Vehicles, The Future of Technology, and Self-Driving Cars’, but it’s not gonna be about the technology itself. We’re gonna talk about the human side of what this means, what this means, what it will mean in the future, what it is meaning right now to give up our control over our vehicles from ourselves to the artificially-intelligent robots. So our guest today is Jason Torchinsky.

0:00:55 SC: Jason is not a PhD at Stanford, designing new radar systems for cars. He’s a writer and illustrator, a good friend of mine that I’ve known for a long time. Jason actually did the illustrations for my book ‘From Eternity to Here’. But more than anything else, Jason is a car-lover. He’s a writer for jalopnik.com, and his most recent book is called ‘Robot, Take the Wheel: The Road to Autonomous Cars and the Lost Art of Driving’. So what we’re talking about here in today’s podcast is what will it be like for us human beings to live in a world where cars drive themselves? A lot of it won’t even be driving ourselves. If you have autonomous cars, there’s no reason to have a person in them at all. They’ll be delivering things… At one point, in the podcast Jason fantasised about driving one of his cars that is liable to break down at any moment, but he has another car trailing behind him, carrying spare parts. I should also warn sensitive listeners this is another podcast where we have naughty language being used. Last week with Carl Bergstrom, it was right there in the title, we were talking about bullshit. Here, it’s just for colourful slice-of-life kind of purposes. So I think that you’ll enjoy this one. It’s both something that will make you think, because the autonomous cars are coming, but also something we can have fun with. And with that, let’s go.

[music]

0:02:30 SC: Jason Torchinsky, welcome to the MindScape podcast.

0:02:33 Jason Torchinsky: Thanks for having me, Sean. I’m very excited.

0:02:35 SC: Well, you’ve written this wonderful book, full of profanity. I’m saying that right out loud so that we know, listener-wise, that this is gonna be rated explicit in the podcast listings rather than a clean episode but, you know, not egregiously so, only when called for. So your book ‘Robot… ‘

0:02:51 JT: Yeah, I’m a big fan…

0:02:52 SC: Yeah, exactly. We’re…

0:02:52 JT: Of profanity.

0:02:53 SC: It’s a tool in the toolbox, right?

0:02:56 JT: Exactly.

0:02:57 SC: ‘Robot, Take the Wheel’ is about autonomous cars, but I… It’s not either a book full of predictions or… Although predictions are there, nor is it really a book mostly about the technology. It’s more… Is it too much to say it’s a book about our attitudes toward this oncoming thing that we’re gonna have to face pretty soon?

0:03:17 JT: Yeah, I think that’s a lot closer. I mean, there already are books out there that covered the tech aspect of it. And that’s changing so rapidly, the idea of writing a book about that just wasn’t appealing to me, and frankly, I’m just not that qualified, really. And as far as raw predictions go, they rarely turn out the way we think they will. And I feel like so many of the things that I see predicted are usually coming from the companies themselves that have a vested interest in these things, and I find them wildly optimistic, and I think they tend to downplay how complex and interesting a problem this really is. So I just wanted to avoid all that, make something… People were hearing a lot about it. I wanted to give them some perspective and get them thinking about it in a broader sense, and not just in an automotive sense, ’cause they are robots. This will be a mass deployment of robots into society like we’ve never had before.

0:04:15 SC: Yeah. Yeah. No, and you make that point very clearly, and we will get there, but I think we have to do at least a little bit of predicting here. I mean, there are probably people who think that, within 10 years, most cars will be essentially autonomous. And there’s probably other people who think that, within 100 years, they won’t. So what is your… The landscape in your mind that you’re using to sort of feel around what the important questions are? What do you think it’s gonna be like 20 or 50 years from now?

0:04:42 JT: You know, I think, actually, the most important thing, as far as that goes, is, before I get into 20 or 50 years, we have to talk about what people think right now, at this moment, because there’s actually a lot happening right now that’s extremely dangerous. [chuckle]

0:05:00 SC: Okay.

0:05:00 JT: And that is there’s a lot of people right at this moment who do believe that autonomous cars are already out there and you can go buy one, and that, I think, is the biggest issue we have to overcome, even before their predictions. I think we’re a ways away, we’re probably a good decade away from these things becoming actually usable and safe enough to just generally have in the road, and maybe 20 years away before they’re actually common. And I think, even then, there’s still gonna be a lot of restrictions. But the biggest issue right now is that they don’t exist at this moment, not really, not in the way everybody likes to think of them. And companies like Tesla is, I think, the biggest offender here, are doing a lot to muddy the waters. Their autopilot setup, they have on their website, it says, “Full self-driving,” that’s an option you can check when you’re buying yourself a new Tesla, and that’s dangerously not the case.

0:05:56 SC: That’s the label that they give it. Those are their words.

0:05:58 JT: Yeah, those are the labels they give it. I mean, it’s called autopilot, which is a confusing name for most people. Autopilot, when people hear that, they think autopilot, the car will drive itself. They think maybe about a plane on autopilot, or something like that. And the level of autonomy that a Tesla has… Actually, that any… Volvo has systems, GM has a system, Tesla has systems, Nissan… They’ve all got autonomous-like systems, but they’re just level… What would be known as level two autonomy, which means the driver has to be aware and in control, or at least ready to take control at a moment’s notice. It is not fully-autonomous by any stretch, and that confusion is why you see news reports of people in Teslas on the highway sleeping and reading, or something like that. And there’s a number of wrecks that have happened where these cars have been on autopilot and encountered a situation that they didn’t know how to deal with, that normally, you would, as a driver in the car, should be ready to take over, but people believing these things were more than they are, weren’t able to.

0:07:07 SC: Well, also, even at the best of times, I mean, that just sounds like a very difficult position to be in, to be in a car that is mostly driving itself but you are nevertheless expected to stand ready to jump in. That’s harder than actually driving, in some sense.

0:07:24 JT: I absolutely agree. In fact, there’s a whole chapter in my book I have about how semi-autonomy is stupid, which is, by semi-autonomy, I’m referring to the level two systems that cars like Tesla and GM’s Super Cruise use right now. And I think it’s fundamentally-stupid because it’s incompatible with how humans actually work, like if something’s doing 80% of the work and there’s a demand on the person who’s in this system, where a machine is doing 80% of the work, to be alert and ready to take over any moment, that’s just not how humans operate. We don’t work that way. If something’s doing most of the work, once it shows us that it’s probably basically okay, we relax, we turn off. That’s why there are people reading or watching movies behind the Tesla, ’cause it is, on a clear open highway, it’s doing about 80% of the work of driving, and that’s kind of enough for an awful lot of highway driving, but crucially, it is not absolutely everything, and it is not enough to go on, and people just… We just don’t work that way.

0:08:36 JT: I find it harder, like when you’re driving, just driving, you go into kind of this state of where muscle memory takes over, it’s something you’ve been doing for years, and so you react before even thinking about it, like when you drive and something jumps in front of your car, you don’t immediately have to do math and process where it’s going, your body just does it. Your car is very much a prosthetic and that it’s an extension of you, you make a motion very quickly and instinctively, you turn the wheel, you hit the break, and you move. If you’re in a system where the car is expected to be handling most of that and you’re kind of… You have to have your hands on your wheel, but you’re also not supposed to be steering, it’s a weird space to be in. It’s an unnatural state, and I’ve driven in these cars, and I hate it. I hate being in this autopilot mode. I would… If my hands have to be on the wheel and I have to be sitting and looking and paying attention, I may as well just drive the damn thing, like just fucking drive, like you’re already there. You don’t get to sleep, just drive the damn car.

0:09:31 SC: And I bet that there is actually a neuroscientific explanation for this, in terms of both sort of system one versus system two or unconscious versus our conscious, but also the map in our heads of what constitutes our physical self that we can use to explore the world. I think it’s more than just a metaphor when you say that we’re in the car, we think of it as an extension of our body, like we get sensory input from it, and it reacts to us, and so forth. And if you break that connection, it’s a different kind of interaction, it’s much more cognitive, much more system two, and that’s way harder to maintain over long periods of time.

0:10:10 JT: I think that’s a really good point, yeah, ’cause when you’re just normally driving… I actually talk about car as a prosthetic, as a body extension in here a lot, ’cause that’s what driving does become. You wear a car, in a sense. You are physically doing things and they have direct actions. You push your foot down, you push down your right foot, and you go faster. You push the break or the clutch or the other foot, and you feel something happen immediately. You move your arms, it’s all… And your body gets used to these reactions. And I think you’re right, you’re not… It is much more of system one kind of thinking. You’re just doing things, and they happen, and it’s physical, yeah, but the semi-autonomous space is a completely different thing, you’re managing a system in a stepped-back way, and you’re not gonna have the immediacy of reaction, it doesn’t feel natural, and I don’t think it’s a good way to fundamentally pilot a moving thing.

0:11:06 SC: So, just so… ’cause I’ve never really followed up or follow these news reports very carefully, I mean, there are stories of people in Teslas or other vehicles getting into trouble, but on the other hand, there are many stories that don’t get published of people driving cars getting into trouble, right?

0:11:21 JT: Sure. Oh yeah.

0:11:22 SC: Is there some feeling for whether, at the current state of the art, these actually are safer or less safe than just driving a regular old car?

0:11:30 JT: That’s tricky, ’cause car-driving is dangerous, no question about it. People get in accidents all the damn time, every bit of the time. We absolutely do.

0:11:39 SC: I’ve met people and I’m incredibly surprised that people don’t die even more often than they do, like the idea of just giving people a ton of metal moving at very fast speeds and hoping for the best, it’s a little bit crazy if you didn’t already have it established in the world.

0:11:53 JT: Oh yeah. I mean, yeah… The reason I’m an automotive journalist, and the reason I love cars is because they are fundamentally irrational things, in every possible sense. If you were an alien and landed here, and you sit and you went to explain to this alien how the personal transportation system works in the world, they would be stopping you after the first five minutes and go, “Woah, woah, woah. What the hell are you guys doing?”

0:12:22 SC: Or just laughing, yes.

0:12:22 JT: Why would you do it this way?

0:12:24 SC: Yeah.

0:12:24 JT: I mean, if you think about the way we choose to buy cars, it makes no sense. The fact that cars… For example, I drive a car, my daily driver is this little weird Japanese market Nissan Pao, it makes 53 horsepower. It’s fine. I get everywhere I need to go just fine. I can go on the highway. I could do 75 miles an hour.

0:12:43 SC: But our audience needs to know that you are special in this particular way. You are not representative of the car-driving public in any way?

0:12:51 JT: No, I am not. And that’s gonna be linked to my point, which is that most cars today have three to 10 times the power of this car. It is we don’t buy cars for rational reasons. We don’t drive rational vehicles. If it all made sense and it was rational, there would be five different kinds of cars, parts would be interchangeable, the styling wouldn’t matter. But cars aren’t like that. There’s something much more human about them. They’re closer, in some ways, to fashion than they are other things we buy. You can have a refrigerator for 30 years, and when you get rid of it, you don’t care, but pretty much everybody I know has either shed a tear or come close to shedding tears when they’ve gotten rid of a car they’ve had for a long time. They’re not like…

0:13:35 SC: I did. I did. I recently changed cars and it was a heartbreaking experience. There’s a lot of goodbyes.

0:13:41 JT: Is this the Jag you used to have? Did you get rid of your Jag?

0:13:43 SC: I did, yeah. Yeah, yeah, yeah. I traded it for a little Electric BMW i3, which is the opposite end of car land from my V8 convertible Jaguar, but fun in its own way. But either way, there is a strong emotional connection to a…

0:14:00 JT: Absolutely.

0:14:01 SC: And you also spend a lot of time in it. A lot of people half an hour or an hour commute, either way, that’s a lot of your life that is spent in that vehicle.

0:14:09 JT: Yeah. No, it’s a big deal. And even people who claim they don’t care about cars. There’s plenty of people who say, “Oh, I don’t care what I drive.” I guarantee you we can find a vehicle that they do not wanna be seen in. There’s someone if they say they don’t care about cars and they say, “Okay, great. I’m gonna loan you this Hummer with gold wheels that I’d like you to drive around for a day,” they would go, “Ugh, I don’t really want anyone to see me getting out of that thing.”

0:14:31 SC: Like you say, it’s much more like wearing clothes. It is definitely part of who you are, even if you disclaim that, yeah.

0:14:36 JT: Exactly. You are making a statement about who you are with your vehicle. And now I’ve forgotten what… [laughter]

0:14:42 SC: We’re trying to get to whether or not they really are safe right now compared to the terrible lack of safety in regular cars.

0:14:50 JT: Yeah. Look, driving is not safe. Driving has never been 100% safe. It probably never will be 100% safe. You’re moving a human body at speeds far beyond the warranty. We were never meant to do any of this crap. We were good at running in fields and chasing gazelle or whatever the hell we did, but this is all, all outside of the warranty. So it’s not safe. Now, autonomous cars, the big argument that’s always made and the hardcore Tesla stans and the people who get mad at you on Twitter whenever you say anything bad about Elon Musk, one of the things they always come back to is that he’s saving lives, and if you’re not a 100% behind these autonomous vehicle plans, then you’re killing people, but maybe that’ll be true in the future.

0:15:34 JT: At this moment, that’s not true at all. And even for all the miles that… We don’t… There hasn’t been a deployment of 100% autonomous vehicles on the road yet. There have been tests and Google has [0:15:47] ____ things like that but it’s all under pretty controlled circumstances, really. There’s people in there watching and guiding. And even though they may have millions of miles racked up, it doesn’t even hold a candle to how many just person miles happen every day in the world of millions and millions of people driving cars everywhere. And it’s… At this moment, we’re not there yet. Look, okay, you could argue, somebody falling asleep at the wheel of a conventionally-driven car, it’s going to end badly no matter what, but it’ll end badly quickly. It will end badly probably within minutes of them really falling asleep at the wheel. You’re not gonna drive very far before you’re gonna run into something or hurt someone or kill yourself. And something like a Tesla or another vehicle with level two autonomy on an open highway, you can fall asleep and there’s a chance you will be okay. There’s a chance you’ll be okay.

0:16:38 JT: There’s a chance you’ll go two hours of driving before you wake up and realise, “Holy crap. I’ve been asleep.” There’s also a chance you could drive for a half hour and then run into a much bigger group of people. At best, I think where we are right now, you have maybe a buffer, maybe there’s a bigger buffer of safety that you’re allowed because there’s some compensatory work that the vehicle is doing. But I don’t know if we’re really at the point where I can 100% say you’re absolutely safe because, look, it could be that you fall asleep at the wheel of your car, you run off the road almost immediately, it’s done. You fall asleep at the wheel of a Tesla, you could end up going from an area of sparse population and traffic density into one that’s higher and then causing even more trouble. And people have been killed behind the wheel of cars that were like Teslas, driving themselves in level two. It’s happened. It’s not a hypothetical. So it’s a complex… It’s complex. And right now, I don’t think we’re at the point where we can 100% say that cars are doing a better job of not killing people. We’re not there yet. They could, but not yet.

0:17:47 SC: That’s fair. The other thing is you have been using this phraseology level two. And I don’t know if that’s just your classification system or if that’s universally-acceptable. Why don’t you let us in on what the classification levels are?

0:18:01 JT: Sure. That’s from the Society of Automotive Engineers, is where this comes from. It’s basically a five-level setup. So, yeah, this is a good thing to explain. So level one is a completely manual car. There’s no autonomy features at all. Maybe level 1.5, you could say cruise control because the car is handling the throttle aspect a little bit. And then dynamic cruise, we’re probably edging on level two, where it keeps the same distance. Level two autonomy is the highest you can buy commercially right now, and that would be like Tesla Autopilot, GM Super Cruise, that kind of thing, where the car can do things like… It’s basically a combination of lane-keeping. It does some visual cognition where it has a camera feed. It reads lines, it can see if something is coming in front of it, and it can slow down or stop, it does dynamic cruise where it keeps a set distance between another vehicle. So it does some, but level two also means it can fail at any moment with zero warning. That’s why it’s in that weird space we were talking about where you have to be ready to take over at any moment.

0:19:07 SC: Yeah.

0:19:07 JT: And that’s why Teslas require people to have their hands on the steering wheel while the system is active, although people have done things like you can shove an orange in the wheel and that could override it. And somebody was actually selling this thing called an Autopilot Buddy that you just hung on the steering wheel and it mimicked the weight of hands on the wheel. So there’s all kinds of really…

0:19:26 SC: Oh, that seems safe. [chuckle]

0:19:27 JT: Stupid ways people… Yeah. And GM is doing something different, where they have a camera pointing at your face, where they’re watching your eyes and if your eyes deviate from the road, then it sets off warnings and stuff. And I think also, I bet if they ever go to court, that they would use that in their favour or something like that too.

0:19:42 SC: I don’t even know. Are these…

0:19:43 JT: So that’s what they’ll do. Yeah.

0:19:43 SC: Are these supposed to be able to also park the car, stop at traffic lights, and start, and things like that?

0:19:51 JT: Level two could encompass stopping at traffic lights, and things like that. Tesla just… I think the recent beta just allows for traffic light-stopping. Otherwise, that’s… And actually it’s funny because it was… They added in a thing for it to read stop signs and traffic lights, and things like that, and somebody had… Was recording, where it was reading a Burger King sign as a traffic slow down thing, and it was slowing down by this Burger King sign, and then Burger King used that in some ad campaign…

[laughter]

0:20:23 JT: So if you were in a Tesla and you could prove that your car automatically stopped at a Burger King, they’d give you a whopper, something like that.

0:20:30 SC: Very good.

0:20:30 JT: And I was playing… And actually, the traffic sign recognition shows up in a lot more cars that even don’t have these systems. Mazda uses it, and then on their little heads-up display, if there’s a stop sign, they’ll show an icon of the sign that shows up, and just with magic marker, I drew a sign that said “Slop”, and it put it up and it fooled the system. So these are… There’s all kinds of potential for shenanigans with this stuff because it’s very easy to fool these things. Artificial intelligence is still pretty stupid when it comes to a lot of things. Okay. That’s level two.

0:21:02 SC: But can it also… And can it park? Can you, in principle, do the…

0:21:06 JT: Yeah.

0:21:07 SC: If you were just throwing caution to the wind, could I get into the car in my driveway, tell it to drive to someone else’s driveway and not touch the wheel the whole time? Is that current technology?

0:21:18 JT: I don’t think you can go driveway to driveway, but you can… There’s a bunch of cars have self-parking systems, where they can go into parallel parking spots or other spots. Tesla even has a summon mode where you can be in a parking lot, a big parking lot in front of a supermarket or something, and you could summon the car to come out of the parking lot and then come to you, tracking you by your phone, which actually that… It amazes me that this is allowed to be done at all because it’s insane, if you think about it.

0:21:47 SC: Oh yeah.

0:21:47 JT: ‘Cause, okay, a parking lot is not technically on the road. This is just a beta system and they’re letting people do it, and this is… And when it came out, a number… Mid last year, there were people putting all kinds of videos of these things going the wrong way down these sections. It’s very impressive technology, but at the same time, it’s not really finished, and they’re basically beta-testing this in public, live.

[laughter]

0:22:12 JT: If I had a 2500-pound remote-controlled car and I said, “Hey, I wanna play with this in the parking lot of a Target and just fuck around with it and see what would be fun,” I would get arrested in seconds. They wouldn’t let me do that. But that’s literally what you’re doing with this Tesla summon mode. It’s not different, but somehow they do it. And it’s… No one’s been killed or anything, but there have been plenty of close calls, and it’s…

0:22:41 SC: It’s a remote-controlled car, but a big one. Yeah. Anyway…

0:22:43 JT: Yeah, but it’s a full-sized car.

0:22:46 SC: So then what’s level three?

0:22:49 JT: What’s that?

0:22:49 SC: Level three.

0:22:51 JT: Yes. Level three. So level three is very similar to level two, but it doesn’t… It can fail safely. Level three means you can… It can basically drive on its own, but if it encounters a situation where it fails or it gets bad input or it can’t deal, where level two would require the driver to immediately take over, level three will have some kind of system for it to fail over in an elegant way. Either pull over carefully to the side or something similar. I don’t think… A bunch of companies have come pretty close to level three. I know Volvo was playing with a system that they claimed could safely pull over, but if you’re on a highway, this is still a difficult problem.

0:23:34 SC: Yeah.

0:23:35 JT: If you’re on a highway and you don’t have a good shoulder, being able to fail over safely is still tricky. You can’t just stop. So if something happens, remember ’cause these things are relying on LiDAR sensor… Tesla doesn’t use LiDAR, but cameras, and some of them are using LiDAR, if mud gets flung onto a camera lens and you can’t see, where humans are pretty good at squinting through the wipers and… We have millennia of evolution to make us good at figuring out what we’re seeing. These things don’t have that. And if they lose most of their vision, they don’t have a lot of choice. They just have to stop safely. So level three is like level two, but you don’t have to take over. It can handle stopping without everyone ending up dead.

0:24:22 SC: And this is… It recalls a couple of conversations we’ve had here on the podcast before on… About artificial intelligence and its struggles and shortcomings. We don’t have the ability yet to teach computers common sense or a feeling for the world around us that somehow takes into account what we all are born knowing about solid objects and how they behave. I remember seeing one demonstration of a machine learning programme that was really, really good at telling the difference between dogs and wolves, but it turns out that it wasn’t even looking at the dogs and the wolves, it was just looking at the background. And the wolves are always in nature and in the snow, and things like that, and the dogs are in backyards. And so taking a picture of a husky out there in the snow, and it’s like, “That’s clearly a wolf.” And those things are…

0:25:10 JT: Sure, you put a wolf on a couch and it’s gonna be a dog.

0:25:11 SC: Exactly, yeah. And that’s very… That’s a little bit worrisome.

0:25:14 JT: It is, ’cause a lot of these are black box kind of systems. We don’t really know what’s going on in those weird brains in there. And it’s… Honestly, artificial intelligence is always a term I found questionable ’cause it’s not really… It’s not even artificial intelligence. It’s more like mimicry of intelligence, or something like that. It’s… The processors are so different. And I think, in driving, we get this even more because there’s so many human aspects to driving, which we should talk… After I go through these levels, we should talk about all those subtle things because I think it’s… That’s what makes it really fascinating. And I think if you wanna say like, “Okay, the first 20 years of autonomous driving development got us to like 70%, this last 30% or 20% is where the really strange and subtle stuff is gonna happen and I think that’s what’s gonna really take a while.”

0:26:03 SC: Yeah, that’s true with everything, but good, okay. Level four.

0:26:05 JT: Level four is it can basically do anything, drive just like a person, you get in, say, “Go wherever I wanna go,” but it’s in a geofenced area. You’ve basically established a corridor of a situation where you know you can kind of control the circumstances, but the car, in that circumstance, can pretty much do anything. I’ve been in a system like this, a test system where they basically had a retirement community in these little golf cart-like vehicles that were autonomous, but it was an extremely controlled situation. After we were in there, we took it out into… It was just outside of Boston. We took it into Boston traffic and the thing freaked out immediately at the first roundabout we went to because it’s programmed to be extremely polite, and part of driving involves taking risks, risks you don’t know the output of. Whenever you go on… Like in a roundabout, there’s always a point where a car is coming at you and you just have to decide, I’m just gonna go. You just hit the gas and do it and you take a little bit of risk.

0:27:04 SC: Hopefully they can deal with it.

0:27:04 JT: A car… What’s that?

0:27:06 SC: Hopefully they can deal with it, the other driver.

0:27:08 JT: Exactly, and you are relying on the other driver to not panic, and all that stuff. And most of the time, it works fine. Occasionally, it doesn’t. But this system, used to its geofenced area in a retirement community, surrounded mostly by other vehicles like it, that were friendly and polite, it got along there fine, but once it got in the real world, it freaked out. It’s like a kid who’d been homeschooled, and then you throw him into a brutal high school situation, and the first day, he’s gonna come home in tears and feeling pain.

0:27:41 SC: To be fair, I kind of freak out when I’m asked to drive in the Boston area also, as someone who was there for many years. It’s a little bit anomalous.

0:27:48 JT: Yeah, it’s not easy, but I mean, there’s plenty of cities like that. And then level five is the hypothetical Platonic ideal of just the complete robotic car, you get in, say where you wanna go, and it figures everything out.

0:28:02 SC: And for the… For even at level four, we could imagine maybe one of steps along the way is we have specialised highways that only autonomous vehicles are allowed to go on, and the speed limit is 250 miles an hour, and we don’t expect them to stop at traffic lights, or anything like that, but is that… Do you think that that’s a reasonable stepping stone along the way to autonomous vehicle-hood?

0:28:26 JT: I think that definitely could happen. I think… And I covered this in the book too kind of from a perspective of someone who wants to keep driving as a human because I feel like that kind of compromise offers… That’s the biggest threat to human driving, in the sense like there may come a point where the tipping point is that more people would rather have their cars drive for them than drive themselves. And then, all of a sudden, I could see human drivers getting forced out of highways that get turned into autonomous vehicle-only highways, which are marked and designed for autonomous vehicle travel, relying on the car-to-car communications that would be happening. You could have very high speeds. In a lot of ways, it becomes something close to a mass transit system, really, just with independent pods or cars because that’s kind of what makes the most sense for highway-like corridors, all of these things should be ganging up together and communicating with one another to form, basically, mass transit kind of setups, and then people still have individual ability to break off of the main group and then go their own separate ways, but that would necessitate no humans around.

0:29:38 SC: Yeah. Well, no humans controlling anything. Yeah. And you also… You invented your own level six to this list of levels, right?

0:29:45 JT: Oh, yeah. My level six… What was my level six? I’m trying to remember.

0:29:49 SC: Level six is where the car decides where it wants to go.

0:29:53 JT: Oh, yeah, where it actually decides where to take you. Right, where you don’t even have… You don’t even tell the destination anymore, level six, where you just trust the car for anything ’cause you just assume it’s gonna know better than you where you should be going, and then you’re done.

0:30:05 SC: We laugh, but that might be coming up. Who knows?

0:30:09 JT: Yeah, it could be.

0:30:10 SC: I think also I do wanna get into the human aspects of driving and interfacing maybe, maybe even driving isn’t the right word, but we should also share with the audience a little bit about the technology that enables these things. We mentioned AI. You mentioned the cars talking to each other, is that… I take it that is not part of most current systems.

0:30:33 JT: Not to the extent that a lot of people are saying it should be. They usually call this car-to-car or C2C sometimes, but the thinking is for this to work really well, every… There has to be a constant web of communication between vehicles. We do it to some degree now. Turn signals are a version of this, and even looking… If you ever… Well, you live in LA so you know in LA, LA has its own few unspoken rules about driving. I remember when I lived in LA, if I wanted to merge into a line of slow-moving traffic, or whatever, the trick was you got… If you made eye contact with the person in the car, there was an unspoken rule, and it worked 99% of the time, the person would let you in in front of them, but sometimes they would really avoid eye contact because they didn’t want to do it, but they knew, once that eye contact was established, it happened. And I remember staring at one guy once, really wanting to get in, and he finally broke and made eye contact and he made that kind of exasperated, “Gah!”

0:31:29 SC: Now I gotta let you in.

0:31:31 JT: Let me in ’cause he knows that’s the rule. LA also has the rule of three people can turn left at a red light. You know about this, Sean?

0:31:42 SC: Well, see, I’ve always heard it as two people. So you are trying to push the idea that three cars can go through a red light, turning left, after it turns red.

0:31:50 JT: Yeah. I’ve always heard three. That was my… That’s what I had thought.

0:31:52 SC: Three is a lot, man. Okay.

[laughter]

0:31:55 JT: It’s a lot, but I feel like I’ve consistently… I lived in LA 20 years almost, and I feel like I consistently would see three going at the red. But it’s not a rule that’s written anywhere, it’s just some degree of that rule, two or three, some degree that you allow some people to go through on the red, is an unspoken agreed-upon rule among the drivers. That’s part of what’s gonna be fascinating to teach autonomous vehicles, ’cause they won’t understand this. These folk laws that we have are things that they don’t get. All the gestures that we make through windows and through… When we look at each other while we’re driving, which we do all the time, all of those are things that these cars won’t get. And I know I’m getting away from the technology, but…

0:32:41 SC: No, no. But it relates immediately back to the technology because if I push back a little bit, and I play the role of the techno-utopian now, once the cars are talking to each other, and we can let them all do a little bit of deep learning on what happens as they go through the streets, the cars will evolve their own ways of signalling and implicit rules that are not written into the traffic code, but nevertheless maybe make things more efficient.

0:33:08 JT: I think that is true. If you had a system where all the cars were talking to each other, we’d need agreed-upon standards, which is tricky, and that also leads… This is the thing I love about autonomous vehicles, is that the more you dig, the more complex this gets, ’cause we’re about to enter an even more sticky discussion, where if these cars are talking to each other and they’re gonna come up with their own set of criteria and rules to make for more efficient traffic flow, that means there needs to be an overarching set of rules and criteria. And this is edging on Asimovian laws of robotics, like do the cars prioritise their own occupants or do they prioritise the good of the people around them? How do they make these value calls? And there’s been a lot of fascinating research done on all this stuff. You get into the trolley problem…

0:34:00 SC: The trolley problem, yeah.

0:34:01 JT: Your flip a switch trolley problem. Yeah, exactly. It comes up… I had to address it in the book because it always comes up, even though it’s absurd, and you’re very unlikely to have to actually encounter it. But it’s a thing…

0:34:12 SC: I’m not sure what happened, ’cause I’m pretty sure that, 10 years ago, no one had heard of the trolley problem, and now it’s just part of pop culture. I’m not quite sure how it got there. I know it was in The Good Place, but it predates that.

0:34:23 JT: Oh yeah. It predates that. I think autonomous cars did a lot to popularise it because the rise of…

0:34:29 SC: Probably, yeah.

0:34:30 JT: ‘Cause that’s the big thing where people start thinking about it. Why would we have brought it up before then, really? Although the trolley problem is hilariously… It’s a hilarious thing to read about ’cause there are various levels of it, there’s one where they talk about pushing a fat guy under the tracks to stop the train.

0:34:44 SC: The fat man, yeah. Yeah, that’s right. Well, this is what you’re saying out loud. I think it is fascinating. The trolley problem is not supposed to teach you something about ethics. It’s just supposed to remind you that we have competing ethical instincts and it’s not clear how to resolve them. And I think that it’s almost misapplied for self-driving cars because, hopefully, if a self-driving car has an option of killing one person or killing five people, it will kill one person. And this is probably, hopefully, very rare that it even has that choice. But if you read Asimov’s stories about the robots, where the real tension comes, because all the stories are about how his laws of robotics don’t work in some sense, there’s some way in which they fail.

0:35:32 JT: Hole, loophole. Yeah.

0:35:33 SC: So what if the completely-autonomous vehicle sees a terrorist who is about to do something terrible? Does the autonomous vehicle have the positive responsibility of taking out the terrorist to save a whole bunch of people rather than just letting the terrorist do the terrible things? That’s where things get a little bit freaky, I think.

0:35:54 JT: Yeah. And actually, I address this directly in the book. Should we make them heroic? Here’s a thought, okay? And I go through this little thought experiment in the book. Okay. Let’s say we’re… At some point in the future where autonomous cars form a population of like they’re 20% of the traffic of general life. And let’s say there was that situation in Toronto where that guy drove a rented U-Haul truck into a huge crowd of people. What if you could sign up when you register your car, in the same way you sign up as an organ donor now, you could check a box that says, in certain situations, my car is allowed to be conscripted for the greater good. So if there could be something where they see a guy in a rented U-Haul who is gonna drive into a big crowd of people at a street fair, or something, if, say, six cars are in-between the path of that truck and the group of people, and one of them is yours, and you could have said, “Okay,” what if your cars could be used to form a road block to keep that truck from getting to where it wants to go? That’s a thing that we could do.

0:36:56 JT: If an ambulance is coming down the street, it could be sending a signal to all the autonomous vehicle traffic to either get out of the way or, possibly even like a sheepdog, herd other traffic out of the way to form a path for the ambulance so it can get to its location quicker. There’s actually a lot of applications where heroism could be programmed into these vehicles. We just have to decide do we wanna do it? Are people willing to potentially risk… If it’s a privately-owned car, do they wanna take that risk? Or do we mandate, if a company runs a fleet of self-driving taxis, that they all have to agree to be heroic in certain extreme situations? I think it’s a fascinating concept.

0:37:39 SC: Well, and it also… It does… I think it’s psychologically-illuminating, ’cause even without knowing what the answer is, when you say, “I could imagine being in the self-driving car that is programmed to get out of the way for ambulances,” that you think, “Oh yeah. That’s not so bad. I could probably live with that.” But then when you say, “What if all the other cars on the road are self-driving and they are programmed to sheepdog me out of the way when an ambulance comes, suddenly, you feel like your personal autonomy is being a bit threatened here.

0:38:07 JT: Yeah. And it is… You would be giving up a little bit of autonomy with that. If your private property could be destroyed in the service of the greater good, I think most of us would probably say they’d be willing to do that, but you are giving up some degree of autonomy. And the idea of prioritising the occupants, are we gonna have one set of rules or is it gonna be each manufacturer gets to decide? Is it just gonna be the kind of thing where you know if you buy an expensive enough car, the people inside will be prioritised more than the people outside? Or can you make these settings on your own, in like a settings option panel in the dashboard, or something like that. And how will it work if we allow people to make their own… If we don’t have a hard and fast set of rules for something like that? It gets messy fast, and…

0:39:00 SC: We live in a country where people have trouble wearing masks in the middle of a global pandemic, so I’m not at all optimistic they’re gonna turn over the autonomy of their cars to companies.

0:39:09 JT: I know, and when I wrote this book… I wrote this book, I guess, year before last… It came out last year, but when I wrote it, we weren’t in a pandemic…

0:39:16 SC: Yeah.

0:39:16 JT: Nobody was talking about masks, and I think I was a little more optimistic. But yeah, seeing how people react to masks is very disheartening. These are not the people who are gonna prioritise the greater good when it comes to their car. If there’s a setting for them to kill everyone around them to make sure they get out through okay, they will.

0:39:36 SC: Anyway…

0:39:36 JT: And then there’s fear… Yeah?

0:39:38 SC: LiDAR cameras, how do these work? How do the cars know what’s around them and who to hit and who not to hit?

0:39:45 JT: Okay. So a lot of different systems in these things. Right now, the most common way is visual, there’s cameras, and almost every modern new car has a camera setup in it that could be used for semi-autonomy. Every car that does lane-keeping, for example, which is extremely common now, almost every car-maker, I think, makes a car lane-keeping. Basically, behind the rear view mirror on your windshield, there’s a camera. It’s looking at the street and it’s basically using visual processing algorithms to see where the lanes are. On an autonomous car, it’s using similar kinds of AI to take images from the camera, identify what’s likely to be a car, what’s likely to be a person walking, lane markings, street signs, that kind of thing, and it uses that to help drive and steer the car. A lot of the images of cars and people are done kind of proportionally. Cars tend to be a certain scale and a certain kind of proportion, and a guy on a… Guy walking or a woman walking is another basic, taller rectangle. Someone on a bike is a certain kind of rectangle. Speed and distance, all these things are factored in, so there’s a big visual component. And that’s… A lot of the hardware in the car, a lot of the image processing, computing power is for reading what the camera’s sending.

0:41:05 SC: And yet, I was surprised to read in your car that the cameras aren’t that great.

0:41:10 JT: No, they’re not.

0:41:11 SC: They’re nowhere near as good as human eyeballs, and that seems like an obvious place for improvement. I don’t know, maybe it just is not worth the cost. I’m not sure.

0:41:19 JT: Well, I think the quality is basically from a webcam 10 years ago, really. I think if they’re doing 1024 x 768 resolution, I think that’s even on the high side. I don’t think they can do much more than that, really. The image quality isn’t that great, and it doesn’t… But it doesn’t need that much detail, really. Although, there have been cases where these things get fooled. Like, if you have a photograph, there was a study I think MIT did, where they had a van with pictures of bicycles on it, and that was very easy to fool these autonomous cars, they would immediately think these were bicycles on the car. They’re…

0:42:00 SC: That’s just nefarious.

0:42:01 JT: Again, these are all things that humans would never be fooled by.

0:42:04 SC: Yeah, but they’re trying hard. And like you said before, shenanigans, right? There are ways that people with not completely noble intentions could make life hard for these autonomous vehicles.

0:42:17 JT: Absolutely, it’s like having a very competent toddler driving some of these, in a way. They can do it, but they’re easy to fool at the same time. And then there’s LiDAR. So, LiDAR, famously rejected by Elon Musk. He didn’t wanna use LiDAR, I think possibly for aesthetic reasons, ’cause it usually requires a funny-looking dome on the top of the car, but it’s…

0:42:38 SC: Yeah.

0:42:39 JT: It’s like laser-based radar, it’s basically… It shoots lasers in a 360-degree… Well, I guess a circle or dome around the car, and it’s recording the time to bounce back. So LiDAR’s very good for doing distance-ranging and seeing things even if the visual conditions aren’t necessarily that great. LiDAR gives a good overall view of everything around the car in a very three-dimensional way. And it would be working with camera input to approximate that, give the car a sense of where it is. And then you have things like ultrasonic sensors, which are mostly used for close quarters kind of thing. Parking, reading what’s very close to you. There’s also radar, which is usually used for keeping the distance between the car and a vehicle in front of it, usually those are just front-mounted radar emitters. These are… That’s most of the systems. And of course, there’s all the computing power needed to run it, and ideally, in the future, there probably will be communications equipment. I mean, there already are. There’s GPS also. It is talking to satellites to know where it is and know where to go. GPS is crucial for this, and it can use GPS and pre-existing data about the roads, so it kind of knows what’s coming up and can predict certain things about the roads.

0:44:02 JT: And even that could be updated. If there’s systems in place to give it like information, and that changes rapidly, like there’s construction here and it’s down to one lane, that information hypothetically could be sent to cars prior to them driving. I think Cadillac was using a system for a while, where they were doing a sort of higher-level autonomy on roads that they had vetted, that they knew, and were in a database that was sent to the car so they could drive them with more confidence. But of course that doesn’t help for random things that could happen, other drivers, logs that fall on the road, cattle trucks overturning, anything like that.

0:44:39 SC: I mean, these days, most of us are carrying around personal GPS devices that are in constant communication with the rest of the world in the form of our phones. I could imagine very easily hooking up the phone to a car and letting it share that info. You must have seen the guy who was carrying around a whole bunch of iPhones in a little toy wagon, and then Google thought it was a traffic jam on that street, because…

[chuckle]

0:45:04 JT: Oh, yeah.

0:45:04 SC: All those phones moving slowly down the street.

0:45:06 JT: Yeah, it didn’t know… Yeah, that’s right, which is amazing. Yeah. Yeah, that guy was awesome. It’s… So, yeah, I mean… And the other thing is some of these other things that don’t seem to get thought about very often though, even with all these sensors being really good, you have to keep them clean. We’re really good, as people, at driving through hard rain, and snow, and icy conditions. Like, even most of us, driving in cold weather ice, we scrape out a hole in the windshield, and we’re pretty much good to go. It’s not necessarily the best way, our… We’re driving more carefully. We’re maybe trying to be a little more aware, but a good ice storm will keep most autonomous vehicles from going anywhere, unless they have systems in place to heat and melt ice off radar emitters and camera sensors, and things like that, like you’re gonna have to keep these things clean or they’re gonna need to…

0:46:00 SC: Yeah, I know, that could be a thing. But so it sounds like a big part of the challenge then is, just like human beings, the cars are multi-sensory, right? You have the LiDAR and… Do you personally think that LiDAR will be part of future autonomous vehicles or will Elon be right and you can get away with just cameras?

0:46:16 JT: No, I think you’re gonna need LiDAR. LiDAR does provide a lot more information than just a camera’s feed, just a better three-dimensional map of what’s around you, elevation changes, things that are also confusing, visually, LiDAR wouldn’t get fooled. For example, one of the… There’s been at least two fatal wrecks of people in Teslas on autopilot, where a… It’s basically been a… Down a open highway in bright, clear weather conditions, but a white truck has been going perpendicular, like coming out of a gas station or across the road. And the camera wasn’t able to discern the white truck… Body itself, like a tractor trailer truck.

0:46:54 SC: Yeah.

0:46:55 JT: The trailer part, it wasn’t able to discern it from the sky, and the car just basically slammed right into it. LiDAR would have prevented something like that, because it would have read it as a large three-dimensional object in the way. So I think the more systems you have, more sensory systems you have acting as backups to one another is always going to be better.

0:47:16 SC: But so what you’re saying is future cars are all gonna have laser domes on the top. That doesn’t sound like a bad thing to me. I mean, that sounds like an improvement.

0:47:22 JT: Yeah, I think that would be kinda cool. And I’ve seen… They’re getting smaller. And there have been some designs I’ve seen where they’re incorporating them better. And if you don’t have a central one at the top, they’ve been working on ones at each corner that do each… They just work together to form the full arc… They’ll figure it out. I mean, make the stylists work a little harder, they can figure out how to incorporate these things in there, they can do it.

0:47:43 SC: I think… I don’t… So I completely appreciate the worries that you’re mentioning with these… The difficulties with all these sensory modalities, but it does… I’m gonna guess that these are solvable problems. A lot of technology… It might be cautionary for the timescale that it will take us to solve them. I guess what I wonder is, it sounds like a complicated problem, are we… Do we have really smart programmers trying to write code that will help the car turn all this information into a map of its environment, or are we sort of doing the deep learning thing where we just train it, and the car, the AI system, or whatever, figures out for itself how to convert that data into a map? Do you know?

0:48:30 JT: Yeah, I think… Well, I think it’s a combination of both. From what I’ve been able to see the way the industry is moving, it seems to be a combination of both. There’s definitely very clever programmers working on these problems. But there’s also an awful lot of importance put into the number of miles of training that these cars do. Tesla likes to talk about how many millions of miles their cars have racked up with their autopilot setup. Google and Waymo have been driving cars around constantly, mapping and trying and testing, and they’re doing the deep learning method for that. So I think it’s a combination. And I feel like right now the big players are putting a… It seems like they’re putting more, at least more visible effort into the deep learning, training, getting them to drive, just brute forcing as much data in there as possible. That’s what it looks like.

0:49:15 SC: What could probably go wrong? Yeah, I can’t imagine [chuckle] how this could be terribly disastrous.

0:49:22 JT: And the other thing is, like we were talking about before, there’s so many… I feel like there’s a lot of things that aren’t being addressed, like… You’ve driven in New York, I imagine, like in New York City?

0:49:33 SC: I have not. I don’t think I ever have.

0:49:35 JT: You haven’t?

0:49:35 SC: I’ve been in New York plenty of times, but look, I drive in California. I’m a wimp when I move back to the East Coast, where I’m from. I leave that for other people.

0:49:45 JT: Well, driving in New York or any… There’s a lot of big cities like this. It’s a different kind of thing. If you’re at a intersection with pedestrians around, which in like Middle Manhattan is pretty much everywhere, there’s an interesting process that happens. So, I haven’t even noticed, I don’t even know if it matters what the light is showing or the ‘Walk or Not Walk’ lights, and sometimes it just doesn’t matter. But there’s a dance where some mass of pedestrians will be going across. And if the driver has a green light, I guess, and if they feel like they wanna go through, they kind of inch forward into the scrum of people. And they’re basically driving into the crowd of people until there’s some breaking point where people decide, “Okay, let’s let the car through,” and they back off. But for some period of time, you’re driving into a crowd of people very, very slowly, but you’re doing it. So how do you programme a car to do that? [chuckle] How do you programme a car to say, “Drive like you’re gonna run over a whole lot of people, but don’t really mean it”?

0:50:40 JT: And the way it works is, again, there’s a lot of communication between driver and people through the windshield. You’re looking at the eyes of the person in there, at some point, there’s this negotiation that happens within seconds, and then one or both goes through. So like the driver ends up going or then they pause and then the people go by. But there’s a constant negotiation that’s very human. And it’s… There’s communication there, human-to-human communication that I think is going to be really difficult to programme into a car. And also, when it comes to people crossing the streets, we immediately see like a mom with a… Holding a little kid’s hand at the corner of a street, like a crazy little toddler, we know, ’cause we’re people, that toddler could do something bonkers at any moment.

0:51:23 SC: Exactly, much less predictable.

0:51:24 JT: And you go… Or you see someone who’s clearly crazy, who’s just acting weird, and all of a sudden, we’re like, “Be ready for someone to act weird,” that’s a lot to ask out of a computer learning system. And that’s the part where I think things get dicey.

0:51:39 SC: Well, it is…

0:51:39 JT: I’ve also driven in India, which is completely bonkers.

0:51:41 SC: Oh, yeah, that’s…

0:51:41 JT: And I have no idea how the hell that would work there.

0:51:44 SC: I’ve been in places. I’ve been in… Paris and London are pretty bad, to be honest, we don’t need to go to very exotic places, but I’ve been to Southeast Asia also, I’ve not been to India, but yeah, the local norms are all different. LA and San Francisco have different local norms for driving.

0:52:00 JT: Absolutely.

0:52:01 SC: And it does matter a lot.

0:52:03 JT: India is basically Brownian motion somehow [laughter] It’s just… If you think of the cars as molecules flowing, and things… It’s somehow… Not everyone’s dead. It works, but I’ve driven in India, and I’ve been driven in India, and it’s terrifying. People will drive into oncoming traffic in the other lane because the road is smoother on that side, and they won’t even move over until you can read the watch of the guy coming towards you, and then… I was on this off-road rally there, and the only time in my life I’ve been driving this Mahindra THAR, which is like a Jeep, in these sand dune-kinda conditions that was actually very tricky to drive, then there were brief stretches of road. It’s the only time where I’ve ever gotten off a paved road onto dirt and rocks, and thought, “Oh, thank god. Now I can relax.”

0:52:48 SC: But it’s all because those are not our local conditions. I had a friend in Boston who’s like… It’s all the tourists who make it dangerous here because they drive at the speed limit, and try to obey rules, and it’s just not what everyone else is doing. They’re fish out of water.

0:53:01 JT: Exactly.

0:53:02 SC: But anyway, I think I do wanna get to the…

0:53:04 JT: Yeah, go ahead.

0:53:04 SC: The fun part of your book because it’s sort of… There’s a couple of not-very-hidden agendas there, and one of them is to think about the design, the fact that, once you have a truly-autonomous vehicle, it’s not a car anymore. And we are being held back by this idea that we’re thinking of autonomous cars as basically cars, but they’re gonna drive themselves. And what you’re saying is, once you’re not driving it, there’s a whole new universe of possibilities for what to do that opens up.

0:53:34 JT: Yeah, it’s a robot. That’s the thing we have to really… We have to stop thinking about them as cars. Once we get to these level four or five vehicles that don’t even have provisions for a human to drive, it’s something completely different. It’s no longer a car. It’s a transportation robot, and really what people want has very little to do with how cars are designed now. If you think about it, the way we design cars now, it’s like a set of bleachers, two rows and they’re both facing forward. Why would you want that if you’re not driving? You want a little room, really. You want a little room on wheels and you wanna be able to do anything in that little room. You’ll wanna be able to sometimes have it open windows and sometimes you’ll want complete privacy ’cause you’re doing filthy things in there, ’cause people will, because it’s gonna be your own private space. You’re not gonna wanna sit facing forward, you would want couches or people facing each other, maybe you’re gonna wanna sleep in that thing, you’re gonna work, do whatever. It’s just… It’s no longer a car. It’s a mobile room, that’s a robot, that takes you where you wanna be.

0:54:37 SC: I was definitely surprised to learn in your book that Honda has already made a love car.

0:54:43 JT: Yes. Yeah. Oh yeah, I love that car, that Honda. And what’s amazing is that it was the S-MX, and it was actually… So you have to understand the circumstances of the Japanese market to really get it, ’cause it was…

0:54:56 SC: When was it?

0:54:57 JT: Marketed… It was a K-class car, which is a special class of small car that you don’t have to pay the same taxes on in Japan, so it’s designed to be sold in dense cities like Tokyo. Privacy is a precious commodity in a city like Tokyo, especially if you’re a young person who’s not quite out on your own. Let’s say you’re 18 to 22, or something like that, you have a girlfriend, but you still live at home, the idea of a small volume of space that is just yours is extremely valuable for all the filthy reasons you’re probably thinking of. And Honda knew this, and they actually advertised this car with love-based terms, and it was all tongue-in-cheek, everything… It would fold flat into like a very nice little double bed, with all the seats folded down, and they knew exactly what they were doing. And all the advertising was basically mobile love hotel, unashamedly. That’s what this thing is for.

0:55:51 SC: When was this made?

0:55:52 JT: What’s that?

0:55:53 SC: When was this made?

0:55:54 JT: Oh, this? When it was made? It was made in the… I think this was in the… Recently. This is a late ’90s, early 2000s vehicle.

0:56:01 SC: ‘Cause, man, it sounds like ’70s made, but…

0:56:02 JT: Like, yeah, it was… What’s that?

0:56:04 SC: It sounds so ’70s to me, but I guess, okay, it could be more recent than that.

0:56:09 JT: We had that in the ’70s too, the big van, the custom van craze in the 70s, with the shag carpet, that was basically the American version of the same thing…

0:56:16 SC: But the love car…

0:56:17 JT: 100% the same thing.

0:56:18 SC: It was a regular car with seats, but then the seats would basically fold into a bed, this is just what I think is amazing, and maybe I’m surprised it didn’t catch on, I don’t know.

0:56:28 JT: Yeah, I think they sold a lot of these, and the fundamental design is still popular for this class of car in Japan, which is… They call them Tall Boys, they’re basically like little cubes, they have set restrictions for the length and width of this class of car, so they basically just maximise that volume and you just drive these little cubicle cars, and they hold a surprising amount of space inside. The volume in the interior is great, so I’m sure they were used for this purpose long before Honda just thought, “Screw it, let’s just make a car specifically for it.”

0:57:00 SC: Lean in. But this is an inspiration when we’re thinking about how things are gonna change if you don’t need to worry about driving the car. I presume that, for the love car, the idea was you would park, but now you’re saying that the whole idea of engine in front, forward-facing passenger compartment, then trunk in the back, you don’t need to think in those terms anymore, you can be much more open about what you want the space inside to be like.

0:57:26 JT: Exactly. And I think flexibility is gonna be crucial. If you are actually… There’s a lot of people who believe that autonomous vehicles will be something that people don’t own. They’ll all be like Ubers, or something like that. I think, in certain areas, that’s probably true, but I think that’s more true for places where car ownership is rare now, big cities like New York, London, Paris, that kind of thing. But I think, in places like LA, having your own car is… It’s different. People store stuff in their cars. It’s like a locker. If you work far from home, you may take a break in your car because it’s like your own space. I think people will still desire their own private personal car. If you have kids, you know that this is true because there’s so much crap that’s associated, especially when they’re really little, that you just have to keep with you, and it doesn’t… It’s not feasible, the idea of lugging all this crap and kids seats and stuff to put in Uber. You want your own car. So in that sense, I think there will still be private ownership of autonomous vehicles. I think they’ll be very flexible, boxy things, exterior-wise, ’cause that’s just the easiest way to encapsulate the most volume in a given area. And I think the interiors will fold out and do all kinds of different things. All kinds of crash protection’s gonna have to be figured out.

0:58:48 JT: Volvo’s already working on [0:58:50] ____ seat belts, so restraint systems for when you’re laying in a bed. Maybe we’ll decide, at some point, these things are safe enough, we don’t have to worry about that anymore. Maybe we’ll just have more advanced supplemental systems, like airbag systems, to just flood the interior with a foam so you’re not gonna move, something, but people want… They’re gonna want flexibility, they’re gonna wanna maximise the time in the car, and they’re not even gonna wanna think about it in terms of travelling. It’s a room. Whatever you may do in a room, you’re gonna wanna be able to do while you’re in motion, and I don’t really see the point of any other way.

0:59:25 JT: Also, the idea of performance changes. Performance cars that we know about now we drive partially because we get a visceral feeling of enjoyment from driving the car, and there’s a statement that says, “I can handle this amount of power, or a car that handles like this.” If you’re not driving, who cares? That’s not a thing anymore. So I think status will always be a thing when it comes to how they’re styled, but I think the interior space will become arguably more important than the exterior. The exterior will still be important ’cause people still like to convey things with their cars, but these are gonna be vehicles designed from the inside out, as opposed to today where, while the inside is important, the variants of exterior design are huge right now. I think we’re gonna be entering an era of vans.

[chuckle]

1:00:17 SC: An era of vans. Well, I think the most basic psychological change is the idea that the front row seats can be facing backwards. And you maybe put a little table in there, and you can enjoy your coffee along the way, and do the crossword, whatever it is.

1:00:34 JT: We’ve seen this before in some things. Yeah, anything. Work surfaces, tables, eating, whatever, you’re gonna wanna do it. Why wouldn’t we? If you’re not driving, there’s just no point in staring and sitting facing forward. The reason we do it in things like trains and buses is just because of the amount of people you can cram into a given area in airplanes. If we had our way, and if space wasn’t such a big issue, you would be able to move around and do whatever.

1:01:06 JT: And when it came time to move from Los Angeles… And I moved my family to North Carolina. So, for that move, to make it a little more fun, I bought an RV in LA. I bought a $5000 RV, and it was in actually beautiful shape, and it had a kitchen and a bathroom with a bathtub and a toilet, and I put it in a little flat screen TV and hooked up a Wii, and all this stuff. And so we drove across country, and I ended up driving 99% of the time, but my wife and kid were just lounging in the back like it was a little room. And that’s the closest I can think of. In this instance, they had an autonomous vehicle. I was the computer and the cameras were my eyes, but I was basically that, and they just had the best time hanging out back there, making pizza rolls in the oven and using the… When it comes to a luxury vehicle, there is no greater luxury than taking a comfortable, satisfying dump wherever you feel like it in your own car. That’s far better than any Rolls Royce, and it’s… You see the appeal of it. A room on wheels is the way these things are gonna go.

1:02:08 SC: But I’m also interested in how design will reflect the fact if we do have something like the all-autonomous Ubers model, where you don’t… Maybe some people will own their cars for various reasons, both romantic and practical, but one of the huge weird things about cars right now is there’s an enormous number of four-seater cars driven by one person at a time. And if it’s just a set of autonomous vehicles that are basically like Ubers or taxis, and you can call them, then probably there’s gonna be a lot of one-person vehicles, and a lot of special-purpose vehicles that you can call for whatever reason. Is that a sensible, economically-feasible model, you think?

1:02:51 JT: I think so, because I think… It’s even worse now, if you think about it, ’cause not only are there four-seat cars, there’s usually, by and large, four-seat SUVs, which is even crazier if you think about one person in a 5500-pound off-road vehicle with gigantic tires. It’s wildly-inefficient, really. So I think the idea that there will be one or two-person little podlet things, that are Uber-style, that take you place to place, that’s very reasonable. And chances are most of these vehicles will be electric, which means that, from a technical standpoint, most electric cars right now are being built on this skateboard idea, meaning all the batteries are flat in the bottom, the motors are actually really small and can just be basically in-line with the wheels, either at the front or the rear, and they’re very easy to make modular.

1:03:40 JT: Almost every major auto manufacturer right now is developing a modular skateboard-type chassis for their vehicles. Volkswagen’s, for example, is called MEB, and they have basically one fundamental electric platform that’s gonna be everything from a two-door Golf to the new Microbus, the electric Microbus that they’re planning. So this is how it’s going to be. So the idea that making a small, two-passenger car, or a four or even six-passenger car for bus replacement kind of things are more… They’re all modular-type systems. It will make a lot of sense for them to do that, and I think you’ll certainly see a lot of it. And I think a lot of the dispersion of how many of each size of the car will depend on the particular market. Some markets are gonna be more communal group-type of… Like buses or sub-scale buses kind of thing, and some places are gonna be more one or two passenger thing.

1:04:33 SC: Yeah.

1:04:33 JT: I think a one-passenger one is unlikely. I think two is about as small as they’ll get because by the time you’re building for one, it’s easy enough to just add a second seat in there, and people are as likely to travel in pairs as they are singly. So I think you’ll have two… Two, four, six, eight, and then bus-size things. And I think it’ll be easy enough to build those in whatever quantities makes sense.

1:04:56 SC: Well, and also zero person things. This is something you talked about, which I, embarrassingly, had never really thought of, but there’s gonna be a lot of autonomous delivery vehicles with no people in them at all.

1:05:06 JT: Yeah. Yeah. And I think actually that’s going to be… I think that could actually come even quicker than human-carrying vehicles. And I think this is actually, in a lot of ways, a lot better, ’cause, if you think about it, all of the… Many of the errands we make, if a autonomous vehicle is taking you there, you really would start to wonder, “Why do I even have to go at all?

[chuckle]

1:05:28 SC: Yeah.

1:05:28 JT: It might even be better to have an autonomous car that you can’t get in just for quality of life. There’s no reason why, if you had an autonomous vehicle that was basically the size of a pickup truck bed, it was small and it had remote opening compartments, and you could just tell it to go a place. You basically go to the website of a grocery store, you put in what you want from your grocery list, and then you send your car over and then the grocery store already has a system in place. When your stuff is ready, the car gets summoned, they put it in, you maybe get a camera that automatically takes a picture of what’s in there and sends it to your phone so you can confirm, you hit okay, it comes home. Like you have to go out for tampons at 2:00 AM, you shouldn’t have to do that. You just send the car out with some instructions, it goes to the CVS-automated pod area, it picks them up, and it comes back home.

1:06:19 JT: There’s no reason why this can’t happen. It’s the exact same technology. The engineering is even easier ’cause you don’t have to worry about the people inside. These things would be small, relatively cheap. I’ve even… In the book, I even talk about how I would love to have one, as someone who has a bunch of old cars. I’ve got an old ’73 Beetle that I drive around a lot, and it’s getting old enough that… On trips, it would be great if I could have this thing follow me with spare parts in it, if I wanted to take a long trip, and even to tow me out of situations if I just get stuck somewhere. It could be a huge boon to people with classic cars. These would just be literal errand robots.

1:06:52 SC: Yeah.

1:06:53 JT: And I think there’s all kinds of reasons why these make a ton of sense. Maybe even… I know I’m almost more excited for those than I would be… Just because I like to drive, but there’s plenty of times I just don’t feel like… If I wanna drive… And that’s the other thing. When autonomous cars come, if there’s human-driven cars still, they’re going to be fun human-driven cars because all of the mundane driving you can hand off to a machine, either an unmanned errand bot or your commuting machine to go to work. But if you… I feel like the only companies that might survive making human-driven cars are ones where it’s enjoyable. Lotus, perversely, may still be around, or Morgan, or these companies that make cars that are all about the visceral enjoyment of driving. I think it’ll be niche, but it’ll still be there because I think that’s still something people will want.

1:07:42 SC: You make the wonderful point that the design of actual human-driven automobiles over the last few decades has become more and more aggressive-looking. People want their cars to seem threatening and intimidating. And for sort of ex post facto obvious reasons, designers of fully-autonomous vehicles make them just look adorable and cute because they don’t want them to seem threatening.

1:08:08 JT: Yeah, and I think this has fascinated me for a long time. Personally, I’m just someone who unashamedly loves cute cars. I have no shame or guilt about this whatsoever. That’s why I always love Beetles. The little Nissan Pao I drive is adorable. And I’ve always liked that. I’ve never felt like… I don’t know, there’s people who get insecure about that thing, but not me. I love it. And it’s always been a little baffling, just everything now gets an aggressive face. And they claim there’s been focus groups that study this, although I feel like, psychologically, it’s gotta take a toll. Everything, every… Almost every modern car now, unless it’s based on a retro design, like the Fiat 500 or new Beetle, or something like that, has a face that looks like it wants to murder you. It looks like an angry vacuum cleaner that just wants to gut you and suck up your entrails. And they keep making things like this. But when Google made their autonomous car, it looks like a koala bear. They made it look like this adorable koala bear things. It was like a freaking cartoon. It’s just…

1:09:10 SC: Big round eyes.

1:09:11 JT: Ridiculous-looking, but they had to do it because once these things are driving themselves, do you really want a 4000-pound robot driving around that looks like a murder bot? And I think people would be unsettled by that.

1:09:25 SC: It’s true. Even the recent models of the Prius were styled to look more and more aggressive and threatening-looking. And as a previous older owner of an older version Prius, I’m like, “What is the point of this? No one is fooled. No one is threatened by this.”

1:09:41 JT: Honestly, Toyota’s design recently has… I actually… I came up with a name for it, it was Cyber [1:09:46] ____, ’cause it’s like there’s so much going on. There’s folds and things, but it has a mean angry fish face that just looks like a cruel piranha. “I hate you.”

1:09:58 SC: But okay, that’s a good segue into the final topic here, which is, I think, one of the not-so-secret agendas behind your book, which is that, safety and efficiency and visions of the utopian future aside, there are people out there, including yourself, who love driving for its own sake.

1:10:16 JT: Yeah.

1:10:16 SC: And some crazy people even love tinkering with their cars. I don’t really get that, but there are people like that. And how can we carve out a space for that, or is it just impossible? Probably there were people who loved riding horses 150 years ago, and too bad for them.

1:10:35 JT: Yeah. And the thing is the horse model is something I don’t wanna have happen. I don’t want car people to become like horse people because it’s difficult to keep a horse. It’s extremely exclusive. And I’m gonna miss driving. And I talk about… One of the things I think I’m gonna miss most, I call it like the death of the journey in here, because, in an autonomous car, you get in at your start location, you tell the robot where you wanna go, and then you do whatever the hell you want for however long it takes, and then you get out at your destination. It’s like flying, in that sense, or teleportation, hypothetically, could be, in that there’s no middle. There’s just two end points and some amount of time in-between that… It doesn’t really relate to the actual process. When you drive, though, you go on a six-hour road trip, you’re engaged with your journey the whole time. You’re engaged with the environment the whole time. You see the city fade out into farmland. You see the environment change. You see the landscape change. You see things on the side of the road. You have chance encounters, places. You are literally… It’s analogue, you’re taking in every bit of that journey incrementally until you get to your destination.

1:11:50 JT: If you had an autonomous car, you would be on your phone, or reading, or doing whatever the hell you wanted, masturbating, eating in that space of time, and it would just be point to point. And I think there’s something that’s lost there. The idea of a journey as an actual journey that you experience has value, and will always have value, and there’s no place for that in an autonomous future. And I think about… I talk about ways to… I think we’ve all just driven, sometimes, without a destination. And how do you do that in an autonomous car? The first question when you sit down in an autonomous vehicle is, “Where do you wanna go?” But I think… I don’t know anybody who owns a car who hasn’t, at some point in their life, just thought, “Oh, let’s just get in and drive and see what we see, see where we end up.” Or you can go in with a vague idea of like, “I’ve never been to Solvang, California, let’s just drive out there and see what we see.” But how do you do that in an autonomous vehicle that demands a destination the very first thing you do. And you could probably programme in artificial ways around it, you could probably tell it, “Computer, wander,” and then it could wander somewhere, but, cynically, I feel like companies would sell bits of their wandering algorithms to put you in places where you’d see certain restaurants or you’d end up in the parking lot of a Best Buy, or some crap like that, and I don’t wanna see that happen either.

1:13:13 JT: And I also think about potential things for autonomous cars, you could download road trips that are curated road trips from people, and all of these things are cool, but it’s still not the same. The process of just driving is also meditative, and it’s the last place we have in society where people will leave you alone. In an autonomous car, if you know you’re on a six-hour trip, somebody will be asking you to do something in those six hours, but when you’re driving, you’re engaged just enough to not have the burden of productivity on you, and yet, you still have enough of your mind free to wander, to listen to music or a book on tape or a podcast, or whatever, but you’re still… Nobody can expect you to do anything other than just driving that car. And we will lose all of that if we lose human driving, and I feel like we don’t appreciate how important those things still are.

1:14:10 SC: So, basically, what you’re saying is the real danger of autonomous cars is not that they will fail the trolley problem and kill people, but that we’ll be expected to be checking our emails and doom-scrolling on Twitter the entire time that we’re commuting.

1:14:23 JT: Exactly. Is there any context where you can imagine that that’s not gonna happen? If people know you have an hour-long commute, and they know you’re not driving, of course they’re gonna expect you to be working in that time.

[laughter]

1:14:35 SC: Yeah, those should be productive hours.

1:14:38 JT: Yeah, they should be. And I don’t think we… We still need refuges from productivity. We’re productive an awful lot, and driving is that perfect mix of you’re doing enough with your hands and your body but enough of that higher part of your brain is still free. That’s why people… It’s like you know how people will say they come up with great ideas in the shower, or something like that. I think driving a car is very similar, enough of your mind is freed that you can explore ideas. I do some of my best thinking while driving in a car.

1:15:09 SC: Oh, yeah. I agree.

1:15:10 JT: And I don’t wanna give that up.

1:15:11 SC: Yeah, absolutely. Well, I think, regardless of what the predictions are for what the real future is going to be, I think that these are important considerations to keep in mind, in the sense that, hopefully, we have some influence over the future and we can choose to do things. And I do, like you, hope that there is some space carved out for just driving even in our autonomous future.

1:15:30 JT: I totally agree.

1:15:31 SC: And Jason Torchinsky, thanks so much for being on the MindScape podcast.

1:15:34 JT: Thanks, Sean. This was a lot of fun. I really appreciate it.

[music][/accordion-item][/accordion]