Image: Shaye Anderson

Greetings from the Taxibot Future of Los Angeles

In this postcard from the future, Eric Spiegelman, the president of the Los Angeles Taxi Commission, envisions how automated taxicabs will fundamentally reshape the city he loves.

Apr 15 2016, 2:00pm

Image: Shaye Anderson

In this postcard from the future, Eric Spiegelman, the president of the Los Angeles Taxi Commission, envisions how automated taxicabs will fundamentally reshape the city he loves.

Part 1: Getting Around

Friday, April 15th, 2044. I wake up and go about my normal set of morning rituals, which end with me dressed and clean and ready to go to work. The taxibot is already outside waiting for me. It learned my schedule a while ago.

The taxibot is a two-seat, basic model. It looks like a space helmet on wheels. More than half of it is window. The inside is sparse. There's no dashboard, no armrests, no controls. The floor is metal tread plate. The seat cushions are some kind of polyester twill. They have this godawful pattern on them, which I suspect was chosen to hide stains. The whole interior looks like it was designed to be rinsed down with a high pressure hose.

There are a few higher quality taxibots I could subscribe to. There's a bigger two-seater with leather seats and a Harman Kardon sound system. There's an entry-level four-seater, which is just like the cab I'm in, only bigger. There are also "select" and "luxury" models. Those have seats that swivel so everyone inside can face each other. I use those sometimes, on special occasions, but for everyday use I like the basic taxibot because it's the cheapest.

My taxibot subscription works like a TAP card. I prepay into an account and every time I get a ride the fare is automatically deducted. I could just hail a taxibot on the spot market but a subscription comes with certain benefits. My favorite is the priority scheduling. That's why there's a cab already at my door when I put on my shoes in the morning. The taxibot app also hooks up to my Google calendar and knows where I need to be and when, and it shows up right when it's time to go. The app tracks my flight arrivals and departures. The best is when you walk out of baggage claim at LAX and there's already a taxibot in the holding area with your name flashing on it.

Image: Shaye Anderson

My travel needs are all pretty much in line with those of the average Californian. Last year I drove 13,636 miles in basic-level two-seater taxibots. At 25 cents a mile, it cost me a total of $3,400. I have no complaints about this. The year before I opened my subscription, I owned a Toyota Camry. When you add up everything I spent on that car—finance payments, insurance, fuel, maintenance, registration, parking—it came to more than $7,000 a year.

I still own a car, but I don't drive it that much. She's a classic, an old Buick convertible from the 1960s that my dad bought when I was a kid. He gave me a set of keys when I turned 16. During the 1993 El Niño, a deep puddle splashed water up into the electrical system causing smoke to pour out from under the dash. Yep, the Buick used to be so janky, she once caught fire in a rainstorm. I've fixed her up since then, and installed the aftermarket telematics kit that California law says you have to put on old cars if you want to still use them. This allows the Buick to communicate with the taxibots and other robot cars on the road. You can barely see the sensors, so I don't mind. I pretty much only take her to the Friday night car show at Bob's Big Boy in Burbank, anyway.

Part 2: The Transition

Image: Shaye Anderson

The transition from cars that we operated ourselves to cars that did the driving for us was so subtle we didn't even notice it was happening. I used to own a Prius, and one morning I woke up and it could drive itself. I mean this quite literally. Toyota sent out a push update in the middle of the night, and my car downloaded and installed a new operating system before I woke up. A robot chauffeured me to work for the first time that morning.

My Prius had some automated features before the update. It had a collision avoidance system that automatically hit the brakes if someone stopped short in front of me. It had "lane keeping assistance" that took control of the steering if I started to drift out of my lane on the freeway. Periodically the car received an OS update that purportedly fixed bugs in these systems. Rarely were the changes noticeable. Before going full robot, the most significant change I can remember was the automated parallel parking upgrade. That one annoyed me. I used to be excellent at parallel parking. Then, one day, everyone was. My talent was obsolete.

In just a few weeks, every make and model of car manufactured during the previous eight years received an update that allowed full autonomy. Cars less than two years old had a wifi or cellular connection that updated automatically. Vehicles older than that needed a mechanic to do the upgrade, but dealers made this a basic part of regular maintenance. If your car was so old that it didn't have the laser sensors needed for any level of autonomy, you could have an aftermarket kit installed for just a few hundred bucks.

Half the cars in Los Angeles had full self-driving capabilities within six months of the update. The ridesharing companies, Uber and Lyft, capitalized on this almost immediately. The first taxibots were privately owned cars that went out by themselves and earned money for their owners while their owners were at work, or on vacation, or home for the night. Each one earned, on average, $7 for every hour its owner put it into service. People initially assumed that this would be every bit as popular as UberX and Lyft were when they launched in 2013. This proved not to be the case.

A car often returned at the end of its shift with a mysterious odor or an unidentifiable stain. Upholstery was slashed. A name was carved into a seatback. A headrest vanished. A human driver, it turns out, also functioned as a chaperone. Her mere presence deterred bad passenger behavior. The damage ate into the revenue from the trip. Repeatedly getting things repaired was a real hassle. Ultimately, the psychological toll of having your car befouled kept owners from signing up with a rideshare company. "UberX Has a Vomit Problem," a Jalopnik headline read.

Checker Cab, founded in 1984, became the hot new startup. The company received a $500 million cash infusion from Andreessen Horowitz to replace its old sedans.

In a twist, these troubles gave new life to the few surviving Los Angeles taxicab companies. It wasn't so much the loss of passengers to UberX and Lyft that originally put the taxi industry on the endangered species list, it was the loss of drivers who stopped leasing their cabs. At the time of the autonomous car operating system update, hundreds of taxicabs were sitting idle in storage lots. When these cars got their upgrade they became the only reliable taxibot fleet in the city. Many were already equipped with passenger-facing cameras to deter vandalism, a feature added years earlier to protect the safety of drivers. Checker Cab, founded in 1984, became the hot new startup. The company received a $500 million cash infusion from Andreessen Horowitz to replace its old sedans with smaller and more efficient vehicles. The two-seater I'm now sitting in came out of this investment.

Most of the human taxi drivers had already disappeared by this time. Many had retired, while others found a new livelihood. Those who remained were put out of a job, but not immediately. Checker Cab gave them an interesting severance opportunity: teach the taxibots the best way to maneuver through Los Angeles.

Inefficiencies littered the maps used by taxibot computers. For example, the existing map regularly told a taxibot to turn left onto a major thoroughfare from a side street with a stop sign instead of a light. Taxibots often lined up on Rowena, waiting twenty minutes for a lull in traffic on Los Feliz Boulevard. Humans knew this was insane, but GPS systems did not. Former drivers rode thousands of miles of Los Angeles streets noting these failures and teaching the robots better routes. Checker Cab's map of Los Angeles is now the industry standard. All that knowledge of how to maneuver through the city became Checker's intellectual property, and former drivers own a share of the licensing revenue.

Part 3: The Math

Image: Shaye Anderson

When Los Angeles taxicabs had human drivers, they cost almost $3 a mile. Pre-robot UberX and Lyft fares were roughly $1 a mile, not counting surge rates. Taxibots, as I mentioned earlier, are only 25 cents a mile now. This is a relatively recent phenomenon, just the last few years. Fares started dropping almost immediately after the first taxibots started to appear on the road. The more that showed up, the lower the fares got. Prices bottomed out at a quarter per mile.

Taxibot fares lowered as taxibot designs evolved. It was a rapid evolution. The taxibot models of one year often bore little resemblance to those of the previous year. The first taxibot I saw on the road was a 2020 Toyota Prius sedan with no driver. A few years later the two-seaters were everywhere. Then they started making taxibots out of plastic instead of metal. I guess the number of taxibot collisions fell so low that we didn't need a lot of the safety equipment we used to have. You could say this reflected a natural progression as taxibot makers shed features that were no longer necessary. Or you could say the taxibots just kept getting chintzier.

Not long ago, I tried to figure out how a taxibot company made money on 25 cents a mile. What I discovered is that the business is fantastically lucrative, mostly because of the tremendously high volume of service they deliver. A single taxibot drives more than 150,000 miles in a single year. They are so efficient that for 90 percent of this driving, there's a paying customer. Humans were barely capable of half this amount of driving. Los Angeles used to cut its taxicab drivers off at 10 hours behind the wheel in a day, because driving sleep-deprived is as dangerous as driving drunk. Taxibots don't need sleep. They're robots.

The cost of any car spread out over that much driving looks tiny on a per mile basis. Also, taxibots are, in many ways, less expensive to purchase and operate than the hulking metal masses of old. A basic two-seater model, with its plastic, recyclable body, can be purchased for $15,000. Taxibot companies turn their entire fleet over every two years, partly to keep up with ever-improving technology, but mostly just because they can. Taxibots are disposable.

They're also exponentially safer than cars used to be, which makes insurance premiums far less than they were 30 years ago. Taxibots communicate with each other and with all other vehicles on the road (they also talk to the road, itself). They form a network that knows, collectively, where all the potential dangers lie, and adjusts vehicle behavior accordingly. Crashes are mostly a thing of the past. It used to cost over $500 a month to insure a Los Angeles taxicab, but now it costs only $150 per month to insure a taxibot.

Fuel costs and maintenance are the only things that have remained constant. The first taxibots were hybrids, because it took too long to charge the battery of a purely electric vehicle. It didn't take long to figure out that if taxibot batteries were swappable, they could be entirely electric without charge time being idle time. Electricity costs basically the same as what it cost in 2016, roughly 3.5 cents per mile.

Maintenance costs haven't changed since 2016, either. Taxibots do so much driving that their batteries burn out entirely every six months. Each battery costs $2,589, same as an old Prius battery. Mechanics conduct 10 days of major service at a cost comparable to the maintenance of an old Prius, roughly $3,000 per year. Taxibots also need three full tire replacements every year. Los Angeles taxibot companies prefer the Michelin Defender. Twelve of those cost $984.

Taking all of this into account, here is the math I came up with:

Annual Mileage:

Average length of a taxibot shift = 17 hours (use tapers off in the middle of the night as Angelenos drift off to sleep).
Average speed of a taxibot: 25 miles per hour.
Average operational days per year: 355 (10 days in the shop).

17 hours x 355 days x 25 MPH = 150,875 total miles driven per year
x 90 percent occupancy rate = 135,788 miles driven per with a paying customer.

Annual Costs:

$15,000 taxibot purchase price, spread over 2 years = $7,500 per year.
$150 monthly insurance premium x 12 months = $1,800 per year.
3.5 cents per mile in electricity x 150,875 miles = $5,281 per year.

$2,589 per battery (x 2) + $3,000 service costs + $984 in tires = $6,573 per year.

All the costs of operating a single taxibot come to $21,154 per year. Spread this out over the 135,788 miles a taxibot drives with a passenger, and a taxibot company needs to charge just 15.58 cents per mile to break even. For every mile I drive in a taxibot, the taxibot company makes that back plus 9.42 cents.

I read recently that taxibots are so popular now that the number of people who own a car is 80 percent lower than what it used to be. If everyone is saving as much money as I am, that makes sense. Back in 2010 there were 3.47 million car owners in Los Angeles County. An 80 percent drop means that 39.85 billion miles per a year shifted over to the taxibot industry, enough demand to keep almost 300,000 taxibots busy. The revenue on 9.42 cents per mile for all those miles? More than $3.5 billion dollars a year.

On second thought, maybe 25 cents a mile is price gouging.

Part 4: The New Equality

I live in Leimert Park, in a hundred-year-old Spanish style two-bedroom with a nice lawn. The lawn was eight feet shorter when I moved in. When the City started to reclaim street parking and turn it into green space back in the 30s, my lawn unfurled another 400 square feet. This house has a garage, a quaint feature you don't see in new homes anymore. My daughter turned it into a dojo.

The demand for taxi service in South Los Angeles shot up more than a thousandfold very quickly, and taxibot companies found themselves with every incentive to blanket the area with cars.

Leimert Park is on the northern edge of South Los Angeles. When I was growing up, we called this neighborhood "South Central." In taxi-speak, it was called "hard-to-serve." Cabs wouldn't show up here as quickly as they'd show up in, say, Hollywood. The LA Department of Transportation spent years trying to improve taxi service in South Central. They actually did a great job. Former city administrator Tom Drischler, for whom the Drischler Taxibot Safety Algorithm is named, deserves much of the credit for this. But even as response times improved south of the 10, they were never equal to the more affluent areas of Los Angeles.

Zone C is Hollywood, Zone D is South Los Angeles. Chart design by James Stanek

There were two reasons for this. One, some taxi drivers were simply afraid to go to the 'hood. It was prejudice, pure and simple. But the other was basic economics. Taxi drivers didn't like to wait for passengers in a neighborhood where fewer people could afford a ride. They also didn't like to hang around in neighborhoods with fewer people, in general. (The population density in other parts of Los Angeles was higher than in South LA.) Drivers had families to support. They couldn't waste time with an empty back seat.

Taxibots solved this problem. First, they have no racist tendencies, because they're robots. Second, taxibot fares got to be so low that everyone could afford them. They're cheaper than Metro buses for rides shorter than seven miles. The demand for taxi service in South Los Angeles shot up more than a thousandfold very quickly, and taxibot companies found themselves with every incentive to blanket the area with cars. For the first time, taxi service in South Los Angeles was equal to that of Hollywood.

Taxibots had a number of tangential benefits to the neighborhood, as well. The time it takes for one to show up is so low, and the travel time so much quicker than other options, that parents now have 30 minutes to an hour more to spend with their kids every day. The recent drop in truancy rates and the rise in retention rates at local public schools correlates with the growth of taxibot service.

Also, South Los Angeles used to be considered a "food desert." There were no sources of fresh, healthy groceries at all nearby. Alongside the taxibots grew a whole class of delivery robots—basically, taxibots that moved stuff instead of people—and their delivery fees followed the downward trajectory of taxibot fares. Everyone in the neighborhood had access to healthy food even though the nearest store was a dozen miles away. Most people didn't mind a $3 delivery fee for two weeks of groceries.

Leimert Park is having a moment right now. It's the new Pico Union, which was the new Lincoln Heights when Lincoln Heights was the new Echo Park, after Echo Park became the new Silver Lake. (We all remember those days!) In terms of access to transportation, we're no better than any other neighborhood, and no neighborhood has it better than we do.

Part 5: The Prepared Food Boom

Image: Shaye Anderson

I want to tell you more about the food thing, because I'm really fascinated by it. Taxibots turned Los Angeles into the center of the nation's prepared food industry. You know how fast food was basically invented in Los Angeles a hundred years ago? This was like that, only much, much bigger. To explain how it happened, I like to tell the story of Silverlake Ramen, one of the first beneficiaries of the prepared food boom.

Silverlake Ramen became an unexpected success almost immediately after it opened in a strip mall on Sunset. There was nothing contrived about the place. It had a nondescript decor and a boring name, bucking the trend of that era for behipster-monikered concept shops. Their ramen was known for being good enough, a notch better than the other ramen places nearby. The price point wasn't cheap, but it wasn't more than you'd expect to pay, either. The place had a strong foothold on the middle.

All this proved to be a magic combination. There were always lines outside, even at two in the morning. I used to joke with friends that the secret to getting rich in LA was to open a ramen spot across the street from Silverlake Ramen, just to capture their overflow. Their popularity turned out to be a master class in restaurant economics. Their cooks became experts in bulk ramen preparation. Their servers became quite good at turning over tables, strongly hinting to patrons that they should leave within seconds of finishing their meal. They operated at close to perfect efficiency.

Postmates nearly put Silverlake Ramen over the edge. Without delivery, Silverlake Ramen patrons came from a roughly two-mile radius. Postmates expanded that to more than six miles. When Postmates switched to taxibots customized for food delivery, or "dbots" as we soon called them, they were able to cut delivery fees by 90 percent. Demand exploded again. Why wait in line when you can have ramen delivered for basically the same price as eating out? Silverlake Ramen couldn't keep up. They moved all their delivery prep off-site, to a large building in the Warehouse District of Downtown Los Angeles. This turned out to be a good idea.

The dbots that brought Silverlake Ramen to customers were a delightful piece of engineering. Ramen deliveries were separated into into two styrofoam cups—one for broth and one for the ramen—and each deliverybot could hold up to 32 styrofoam cups. Most importantly, the dbot kept these cups warm the entire time, for the belly of the bot itself was an electric oven. A robotic arm removed each customer's order and placed it into a receptacle box that the customer supplied. Then the dbot said "thank you" in Japanese and sped off.

Image: Shaye Anderson

People in Santa Monica were able to get Silverlake Ramen delivered to them, hot and fresh, in 30 minutes, for a delivery fee of $3. You could get Silverlake Ramen in Long Beach and Calabasas, both almost 30 miles from Silver Lake. The combination of bulk prep and dbots allowed Silverlake Ramen to scale in ways never before possible for a mom-and-pop restaurant. Soon they franchised their system nationally. You can now get Silverlake Ramen in Boston. It's delivered from an officially sanctioned prep facility in Waltham.

The original Silverlake Ramen is still there, on Sunset, in the strip mall containing the liquor store that the band Silversun Pickups was named after. I think they keep it open for sentimental value. They have headshots of famous robotics engineers on the wall.

Eric Garcetti, back when he was Mayor of Los Angeles, always used to talk about this idea of "the city as a platform," that Los Angeles was a proving ground for innovative ideas. Never was this more true than for the prepared food industry. If your restaurant had an efficient way of cooking, Los Angeles dbots could raise you to the precipice of a global brand. In that regard, restaurant startups in Los Angeles were like media startups in New York. If you can make it here, you can make it anywhere. And deliver it everywhere.