The Car Ownership Model is not Broken

Some self-driving industry leaders say, “passenger cars … are designed for people to own and only use 4% of the time”.  Four percent of the time! That’s shocking! But this is true: my car gets used about as much. It’s a really strong, snappy pitch, like whoa, 96% waste! Surely, given the data, it’s obvious that a shared robotaxi service would eliminate this waste and improve our lives. But what if…

What if we apply this to other things in our lives?  Let’s try to put the 4% figure in perspective and apply the same metric to the things commonly used in our household that most consider “essential”, “integral” part of their lives. Here we go.

Hygiene

My toothbrush gets used 0.3% of the time. I wash my hands probably 10 times a day, 15 seconds per wash, which means my hand soap is used 0.2% of the time. The shampoo is used about 1 minute a day, which is 0.07% of the time.

My electic shaver is used 0.2% of the time, but other body care items (like, the pumics stone) are used 0.002% of the time. It’s actually shocking how little usage they get.

My towels are probably used 0.03% of the time. They are literally being laundered more than they are used!

Speaking of laundry, my washer and dryer are used 0.8% of the time. Shared laundry equipment is a whole industry in the US though, but when I have to use those, I have to literally plan my day around doing laundry. Not what anybody wants to do.

Furniture

The household items that are notorious for high usage is bedding. My mattress gets used 25% of the time (excluding the times I sleep elsewhere).  My bedsheets and pillows have slightly weaker track record since I rotate them, but they get 10+% usage too.   It’s even a trope in the financial advice literature to invest in high-quality bedding (e.g. “How to live like a Billionaire for 1/3 of your life for under $1,000”).

When I’m not working from home, I’m maybe spending 2% of the time using my desktop computer (and the desk it stands under).  This ranges from 1% on weeks where I spend most of my day at work to full blown 25% when I need to get something done, like the Stanford studies.

It’s difficult to define usage for storage drawers.  Are they used at all time because they store stuff, or are they used only when I open/close the drawer?  If the former and they are considered used, then let’s also look at my car as a storage device.  In addition to storing the car maintenance supplies (essential if you’re traveling out of the cell phone coverage area), it stores bicycle parts, spare gym bag, travel essentials (that corkscrew really saved the day a few times), some spare clothing, a first aid kit.  Essentially, no matter how you define the usage of “storage” the ratio of usage of the car to the usage of other items would be unchanged.

Clothing

Each of my suits I wear to work gets used ~1% of the time (by “used” I mean worn, not hanging on my chair).  That is, when I’m employed; currently the entire “business” portion of my wardrobe sits completely idle (Rent the Runway?)

Really warm clothing (thick sweaters) on the other hand gets used maybe 0.1% of the time.

How often does underwear get used?  If one rotates underwear, then each individual piece is probably worn 2-5% of the time.  The “concept” of underwear is however used almost all the time by all people.

What about clothing maintenance? My iron gets used 0.02% of the time. And yet, 87% of households own an iron. Puprle Tie and other similar services are just … expensive and slow (4 days of latency on your laundry, jeez).

Entertainment

For about half an hour of watching TV / Netflix / Movies per day, my TV is used 2% of the time. Not that little but definitely less than my car.

Each piece of my home gym equipment gets used 0.002% of the time.  It’s a backup for when the gyms are unavailable.  I was really happy I had it when all dumbbells went out of stock.  Now that I’m actually working out at home, it’s used 1.1% of the time.

My ski boots fare a bit better.  Every day you spend skiing, only 1-2 hours are actually spend shredding; the other time is spend standing about.  For 15 days of skiing a year, this yields 0.2% usage; not bad! I use my ski boots about the same amount of time as my toothbrush. This is frightening.

Each of my bicycles gets used about 1-3% of the time.  If I only owned one bicycle, it would get used about 5% of the time.  Although, people who spend 5% of their life on a bicycle rarely own just one.

My backpacking gear gets used 1.5% of the time.  That’s kinda a lot.  It seems I do sports more than an average person does though. Importantly though, a person who goes backpacking at least 6 days a year (which is the 1.5% usage) would definitely own, rather than rent, their backpacking gear.

Cookware

My fridge is pretty much the only item in my household that is used 100% of the time no matter how you define “used”.

My stove gets used 0.3% of the time.  And while I’m sure Travis Kalanick goes around saying soemthing akin to “stove ownership model is broken”, should we get rid of the stoves in our apartments?  The same goes for my dishes; say each is used for 10 minutes a day, which results in 0.7% usage.  Should I rent those too?

Perhaps, the plates, spoons, and other dishes are used most often, with probably 1-2% of use for the “concept of” dishes, and maybe 0.1-0.5% for each individual item.

Most of the cookware is probably used for 10 minutes a week, which amounts to 0.1%.  Still, everyone who cooks owns a set of measuring spoons.

So what does it have to do with the car ownership?

The point here is, most of the items considered “essential” are used less, way less frequently than one’s car!  The car is actually one of the most frequently used items in my household, only behind the fridge, underwear, bicycles, and bedding. So can we really reason that the car ownership model is broken on the basis of the low usage alone?

Not really. But there’s more.

The biggest point of owning a car is decreasing the latency to access transportation services.  Overprovisioning in order to reduce latency is the standard technique in software engineering.  We should not drive cars to the grocery store but we do because there’s no waiting for the bus.  There’s no need to wait for the uber to arrive. There’s no need to wear a mask in your own car.

Besides, the Ubers stopped accepting calls to places that require 5+ minute response times. A friend of mine who lived in Presidio of San Francisco would often not be able to hail an uber to their house at all.  This is what we hated taxis for; so Ubers are now officially like taxi cabs.

Thus, having a thing available at your disposal is valuable even if the things is not used 95%+ of the time.  This is why we buy warm clothing in the Bay Area and store it in the drawer 360 days a year.  It’s freezing right now but Amazon will only deliver you a warm sweater on Thursday.

Warm clothing is very relevant because a warm jacket is a transportation device. It’s an item that facilitates your transportation between places during the cold time of the year (and is used pretty much for nothing else).

Owning is not Driving

While there are reasons for me to own a car in America today, I sure as hell don’t want to drive it.

If there was a highway-only L4 self-driving system available for the car I own, I’d pay tens of thousands of dollars for it, and subscribe for hundreds per month. I don’t care about detecting traffic lights at the intersections, just please sit in the Infamous and Eternal Hwy 101 Traffic Jam while I write code, and then get me to Lake Tahoe on Friday.  Sadly, there is no such system available now.

No, Tesla doesn’t have full self-driving yet (see here and here).  But I believe that a fully autonomous highway driving system for trucking or personal use is within reach.  I believe, if any of the current self-driving car companies were focusing on one, it would be available.

This doesn’t need to be as difficult as city driving. A system that knows how to engage on a freeway / highway and safely disengage and pull off when the freeway ends would be valuable to the consumers and easier to build.

Taxi cabs are not just about driving

Another issue with completely driverless taxi cabs is that driving the car is not the only job that a driver does.  The driver currently fills the following roles:

1. Dispatcher.  Ever tried to describe how to get to your house and the dropoff point to a cab driver? (And then they still miss the turn!) Now imagine describing this to a self-driving car that doesn’t even have a capacity to talk to you. Google Maps? They work less reliably than you think they do, and they’re expensive to licence or rebuild.
2. Custodian who ensures the cleanliness of the vehicle and performs or orders sanitation work, if needed.
3. Steward of the infrastructure and cargo.  Defacing or destroying a self-driving cab is an easier job than defacing a taxi since it doesn’t require confronting a human.  In Breaking Bad, fictional characters managed to steal from a train whose security vulnerability was simply its gargantuan length. The robbers stopped the train by exploiting other vulnerabilities in the railway infrastructure.  I can imagine that self-driving truck heist will be a thing (albeit minor).  After all, unlike the giant freight trains, the trucks will be designed to safely stop when obstructions are introduced.
4. Security guard for passengers.  While sometimes the driver themselves poses danger, usually it is the party who ensures the safety of one passengers from the others.  Drivers typically have a lot to lose from an incident in their vehicle. And consider this: if you’re a woman, would you get into a self-driving cab late at night with two dunk guys who start to ogle you as you approach? Or would you prefer a cab with a professional driver who is tracked, insured, licenced and has a lot to lose?

Now, how would a self-driving system, however great it is at tracking obstacles, routing around them, and driving from point A to point B, clean a spill from a passenger seat? How would it ensure the safety of the passengers? I’m not saying it’s impossible (see the sidenote). This does, however, pose the question of whether “self-driving car” is a concept that only replaces the “driving” part of the vehicle operation.

And if we need a human present anyway, why not target the private car market still centered around the humans sometimes driving?

And yet

Does that mean that the companies that are not targeting and are building a robotaxi and driving on the city streets are wasting their time?  I believe these companies will come out ahead, and here’s why.

There is a proven Machine Learning technique that works called “transfer learning”.  You train a model on a difficult task and then unleash it onto an easier task (perhaps, with some fine-tuning).  E.g. you can pre-train your classifier on millions of images and the 1000 GT classes of ImageNet and then fine-tune the pretrained model on a, say, binary “hot dog” vs “not hot dog” classifier.

From that point of view, say, a highway-only autonomous system trained on city driving would be a better system than a highway-only autonomous system trained on highway driving only. In fact, the latter might end up prohibitively worse.

Despite that a private car is only used 4% of the time, very few items we consider essential are used more often than that. The car drivers are not just drivers, they perform other tasks. From that point of view, building a self-driving system that can be employed by private cars (or cars with a driver/custodian present) would be a good play. Just no more of this “the driver must be ready to take control at all times” L2 bullshit.