Autonomous cars — an outdated vision.

Plenty of current engineering efforts aim at realizing an old dream which stems from a world before climate change awareness.

Here’s my hypothesis in a nutshell: Companies and scientific research groups around the globe work like crazy on mastering what seems the last steps towards autonomous driving. It feels to me as if the focus of attention has nearly entirely shifted onto the technicalities, the engineering part. It’s all about the how and when.¹ I will here argue that we should start to seriously question the industry’s vision (or rather it’s apparent lack of having a really innovative vision). Could it even be that self-driving cars are not a solution for the future of transportation?

Wait. WHAT?

We finally (nearly) got what science-fiction promised us for decades: self-driving cars! Who wanted to play the party pooper here?

There is a lot of science-fiction visions for the future of mobility. Often involving cars. Flying, self-driving, time-traveling. But it’s also good to remember that much of what we now consider iconic science fiction (movies, books, comics) were created in a time before widespread knowledge of human-made climate change. Emissions and energy efficiency were not an issue (at least not compared to killer robots, aliens, super-viruses, or zombies). Picture of the legendary time-traveling DeLorean from “Back to the Future” (picture link, attribution: JMortonPhoto.com & OtoGodfrey.com, CC BY-SA 4.0 license)

Sure, self-driving cars and flying cars are all super cool. Many of us grew up with those dreams. That’s what we saw in shiny movies, books, comics. But I believe it’s time to question those car-related dreams and realize that most iconic science fiction scenarios which really made a mark stem from times in which human-made climate change has not been on our radar. I was reminded of that when I recently listened to a quote of Larry Burns (former GM executive, now at Waymo) describing with verve what this magical self-driving car could all do:

I think that vehicle would get a lot more usage than my personal car now: When I arrive at work, it drops me off at the door, and then I could dispatch it in the middle of the day to go pick up my dry cleaning, and I could dispatch it again to go get takeout dinner and then go pick up my kids and then pick me up at work and take me back home. This whole world of a robotic personal valet is very intriguing to me. (Larry Burns, interview with Derek Pankratz for Deloitte, published in 2019)

To be honest, I found this vision quite repelling. In essence it means that everyone with a car can then behaves as if they were living like the super-rich in those 90s Hollywood movies, ordering their drivers from A to B to C to D to pick up a slice of bread from B, but butter from C, and cheese from D (it really must be D, that’s simply the best cheese in town, no way around that).

Cartoon by Florian Huber, licensed under CC BY 4.0.

While that sounds like the self-driving car could offer us a lot of luxury, it also sounds like an IMMENSE WASTE OF RESOURCES! If we get such a scenario, self-driving cars will not mean less, but much more car-based traffic. And it made it clear to me that this entire dream of self-driving cars comes from another age. A time when individual transport was key; when driving from A to B in a car was considered such a great thing that many, many, many cities around the globe made the horrible mistake to tear down old, pretty residential areas and parks to make room for big straight lanes and parking slots .²

Just have an imaginary look at it yourself. Hold on for a few minutes at a busy road intersection in the place you live, say during rush hour. Then count the passengers per car. Few cars with 3 to 4 persons. Some more with 2. But most often you will find ONE person in the car. ³ Now, picture that a fraction, say 1/3, of all those cars that you just counted could in a not-to-distant future have exactly zero passengers. Doesn’t that feel odd?

One of the key problems with deep learning models is that it often remains unclear how they will handle rare events. They will hence always remain alien to some extend. So, we shouldn’t think that autonomous cars will be perceived by us as “normal” cars. The fact that they are driver-less and alien is likely to alter our behavior on public streets. Cartoon by Florian Huber, licensed under CC BY 4.0.

There are the voices from industry promising us a brighter, cleaner future if only we would finally get self-driving cars. But I believe that the key problem we should focus on is not the DRIVING, but the CAR!

Individual transport might be the most luxurious one, the most comfortable one. After all it seems to be the form of transport that is most tailored to our personal needs. But it also means that we have to move 1 to 2 tons of stuff (the weight of a car) to move ourselves .⁴

Industry promises

There are a lot of huge promises that I hear from people (or companies) working on self-driving cars.

1People will ‘waste’ less hours driving.
→ Probably true.
Though mostly for the upper middle-class I guess.

2 Self-driving cars will reduce the number of casualties from traffic.
→ Probably true.
There will still be accidents, maybe very different types of accidents compared to now (none from falling asleep or driving after drinking, but some from weird, unpredictable rare incidents that the machine learning algorithms misinterpret…). It still seems likely though that there could be notably fewer severe incidents.

3Self-driving cars will make better use of the already available streets.
→ Probably true.
Self-driving cars would increase the capacity of many streets since they are much better in keeping distance and hence cause fewer traffic jams (there are actually many scientific studies on this, for instance this one).

4We will need much fewer cars.
→ Not sure.
It’s actually a frequent argument for self-driving cars. Since they can be shared more easily we will need much fewer cars and hence less parking spaces in our cities. In urban areas I can believe that shared self-driving cars could indeed result in less cars per household. But if there use becomes cheap and easy, people that now don’t have cars or don’t use them frequently could make additional use of them. So it remains to be seen to what extent the number of cars will really drop. On top of that: Yeah, parking spaces are not the greatest part of our cities, but I somehow doubt that this development will be a dramatic change of our world. ⁵

5It will be much more efficient, it will save miles and emissions.
→ Totally not!
I totally don’t buy that. In my opinion this point will outweigh the possible advantages listed above. The quote by Larry Burns above is a good example of what could happen instead. Sending self-driving cars around could become so convenient and cheap that people will be seduced to use a car MORE and not less. And whether or not it is their own car, a leased car, a hourly-booked car, or follows a fancy subscription scheme, really doesn’t make that much of a difference.

Sure people could SHARE self-driving cars, there could be fleets of self-driving cars that work like a smart swarm of taxis. And those could in principle pick up more than one passenger at a time if the route doesn’t change too much…

To be honest, I think that’s a big marketing lie. Only wait till riding in self-driving cars has become cheap enough and many passengers will rather pay the little extra to have their own car tailored to their schedule and demand for privacy. To me it all sounds like the self-driving car is nothing else than another (huge) luxury consumer market. And that’s why so many companies want to be part of it.

Who needs self-driving or flying cars if we can have a “Schwebebahn” (suspension railway). It is in place in Wuppertal, Germany since more than a century and might still have more of a future than fancy google cars. Like most public transport it is safer and much more efficient than cars. And guess what: this way of transportation is indeed efficient in terms of emissions as well… Picture by Max Grobecker under CC BY-SA 4.0 license.

Let’s focus on what really matters: public transport

And you know what? If it’s really about efficient transport, about saving the climate, about bringing multiple passengers from A to B, than there is something else that is much older than self-driving cars but much more effective in terms of avoiding empty miles: public transport!

Sounds a million times less fancy.
But imaging what could happen if we would invest the same amount of money, hours, brains, and devotion that’s now spend on autonomous cars into developing the public transport of the future! Autonomous vehicles could obviously play a role in it as well, though maybe more to bridge gaps than to be a main component. I am sure that we could create a much faster, cleaner, safer, more reliable, and smarter public transport.
I would join the effort.
What about you?

Footnotes:

¹ Maybe a small note of doubt as this point as well. Many voices currently make it sound like it’s only a matter of how soon self-driving comes will take over our street. And, yes, no doubt that there has been a lot of progress towards autonomous vehicles. But there are also plenty of reasons to remain at least skeptical about whether this will really come so quickly. A typical problem in deep learning is that models have huge problems covering ‘rare events’, i.e. situations that do not occur frequently or look very different every time. And traffic — in particular in chaotic inner-city areas — is full of rare events. If you have ever driven a car in the city center of Amsterdam full of anarchistic bike drivers, swarms of tourists, narrow streets, bridges, construction sites, trams, … you really need to do WAY BETTER than on an average training dataset.

² Even Amsterdam where I live would nearly have done the same and teared down big parts of old residential areas to make room for bigger roads (see for instance 2015 article in Guardian). I guess today there is few people living here that regret this decision…

³ Average values for persons per car are typically around 1.2 to 1.5 per/car. See wikipedia, or EEA (data only up to 2005, but fairly stable around 1.45 persons/car in Europe).

⁴ And don’t be fooled by other purely technological solutions like “zero-emission” cars… Electric cars and alike are in many cases probably better than there gasoline-burning counterparts. But not by as much as one would like. Driving around as the single passenger of an SUV that weighs about 1500–2000kg seems intrinsically inefficient to me. No matter on what energy carrier the motor is run. And no matter who’s driving (you or your car).

⁵ This could be different in the US where many cities are much more build with car-related infrastructure in mind. Still, I believe that there are bigger problems to solve.

Thanks

Thanks to a very nice (German) podcast from Deutschlandfunk Kultur on the auto-auto era for pointing me to the Larry Burns quote. Thanks also to another entertaining podcast on the same topic by Adam Conover for making a strong case for public transport.

Many thanks to cultural sociologist Dr. Sophie Pfaff for stimulating discussions.