Self-driving cars, or autonomous vehicles, are here. What does this mean for us, our world, and how we interact with it? What is the robotaxi revolution? What are the economic implications? We hope to answer these questions in this blog post as we try to stay ahead of the coming revolution in transportation.

When we say "revolution", we're not being hyperbolic. We believe that self-driving cars are going to upend the transportation sector so greatly in this country that it will fundamentally transform our relationships with cars, the urban and suburban environments, public transportation, the car industry, and many other supporting industries. Make no mistake, self-driving cars are not some party trick or driver convenience system. It is a total revolution of the transportation industry which we believe will be as great or greater than even the invention of the automobile itself in the early 20th century.

In this blog post we will discuss the history of human transportation, the development of the self-driving car beginning with the DARPA Grand Challenge, and then discuss the societal revolution which we believe will coincide with widespread adoption of self-driving cars and the robotaxi revolution.

To fully appreciate the coming changes to our world brought about by self-driving cars we should also briefly discuss the development of automotive technology in general. Transportation is vital to any civilization and any civilization which was ahead in the ability to transport goods or people more quickly than others had a huge advantage economically and militarily over others. The speed of transportation was the deciding factor in countless battles throughout history and helped develop massive trade routes which linked empires around the world. Transportation is key to communications and trade between nations and peoples and the speed and efficiency of that transportation is a key to a higher quality of life and the power of a country and its people.

The speed and efficiency of transportation first increased by humans domesticating and riding horses. There is some evidence of horses being used by very early civilizations around 6,000 BC, but most archeologists would say that true domestication and industrial work of horses began between 3,500 to 2,000 BC. Once domestication of horses was widespread this became the primary mode of land transportation for humans for the next 3,000 - 4,000 years until the invention of the railroad and the locomotive.

The invention of the railroad as a means of moving people and goods was the first true revolution in transportation since the development of horseback riding. The earliest railroads were developed in Britain in the 17th century, and began appearing in America in the mid 1800's beginning with the B&O Railroad. Railroads were a huge step forward for transportation, but humans still required horses because railroads were limited by the need for tracks and the space requirements of the train itself. Rail transit allowed the US to expand to the west coast and greatly enhanced the speed of transit cross country from 6 months by horse drawn carriage to merely a week via rail. This led to an explosion in economic activity which was exponentially enhanced along with other technological developments during the industrial revolution.

It wasn't even 100 years later that the next major revolution in transportation materialized in the United States and finally ended the need for horses in the United States. The first true mass-produced vehicle, the Model T, marked the beginning of the modern era. The automobile continued to improve rapidly throughout the rest of the 20th century, but it was still essentially the same concept with a human driver required to operate the automobile. Although the actual act of transportation was extremely efficient now, there were still inefficiencies in the concept of private automobile ownership and usage itself. Within 100 years of the development of the automobile, we are now at the point of the next major revolution in transportation, the self-driving car.

Self-driving cars have been attempted for a while, with some experiments in the 1980's and 90's. These, however, were not true attempts at commercially viable self-driving cars. The true proof that this could be possible was marked with the DARPA Grand Challenge in 2005 where a Stanford team AI driven car successfully navigated 132 miles of off road terrain to win a $2 million prize from DARPA. Although this signaled that the concept itself was possible, there were more technological advancements needed for a commercially viable self-driving car to be developed.

Soon after the Grand Challenge the next pivot piece of technology for self-driving cars was developed: the iPhone. the iPhone was pivotal to the development of self-driving cars because it also drove the democratization of high technology on a mobile scale including the development of mobile data via cellular networks, GPS integration, and mapping software. These technologies continued to rapidly improve over the next 10 years to where now we have more power in our pockets than a mainframe computer had from the 1970's.

Mapping software, such as Google Maps, was vital also to create and maintain common mapping platforms which could be used by car companies to cheaply and easily navigate through urban environments. Mapping software has also enabled ride sharing services such as Uber which will be necessary to power robotaxis in an urban environment and create a true commercially viable service. The car itself would register as a driver and behave as an owner operator.

The Society of Automotive Engineers (SAE) has attempted to develop a classification system for autonomous driving to provide a standard set of guidelines in the industry. Most car manufacturers are searching for level 3 viability as of April, 2023, which is defined as conditional automation with the expectation that the human driver will respond appropriately to a request to intervene. This is a good target, but it isn't true autonomy, at this point it is still a driver assistance system. This isn't revolutionary. to be revolutionary, you need a level 4 or 5 vehicle, which does not require a human driver to navigate successfully and safely in the vast majority of cases.

As of April, 2023 there are really only 2 companies looking to develop a truly driverless vehicle, though their approaches are wildly different. There is Waymo, which is a specialty car service being run by Google, and Tesla FSD. Waymo was a more plug and play kind of system and seems like a stitched together hodgepodge of many different technologies from a variety of vendors. This includes technologies such as LIDAR and other radar-based systems, sonic transponders, and others. Waymo is forced to rely on highly detailed 3d maps of the roads to successfully navigate.

Tesla, on the other hand, is a vertically integrated electric car manufacturing company and is building their FSD platform in house from scratch. Their approach attempts to mimic a human driver by using only video cameras and image recognition AI to synthesize the world around the vehicle. One key advantage Tesla has is its massively scalable data gathering and sensor suite also known as Tesla drivers. Each Tesla with the FSD software installed is constantly capturing data about the road, driver reactions, the car's performance itself, and many other data points. These data are then fed back to Tesla over the internet using a cellular data connection (see how the iPhone enabled this tech?) which is then used to further train and refine the model. This results in a feedback loop between Tesla and its drivers where each improves the performance of the other. Every time a Tesla with FSD installed comes online this improves the quality of the overall data model.

Waymo does work well within predefined geographical boundaries, and while this is probably adequate for short term intra-urban transportation, we don't believe that this truly qualifies as an AI system. Tesla FSD, on the other hand, absolutely does. Tesla FSD gathers data using only vision, synthesizes a model of the world by stitching together images from a 360-degree view of the car, uses judgement to predict car and people movements, and then takes advantage of Tesla's navigation system to maneuver through the world. Tesla FSD is a true AI because it is not geofenced and can adapt to almost any urban and even suburban environment it comes across. Additionally, Tesla FSD is a silent AI, meaning that a given person on the street could easily spot a Waymo vehicle, but would not be able to tell an AI powered Tesla from a regular one. That is a true Turing test. Tesla FSD is demonstrating the future of how humans will function in an increasingly AI driven world. This raises a very important question: are humans comfortable with putting their lives in the hands of an AI system?

We also like Tesla's approach because the technology is not directly tied to Tesla vehicles. After all, it is simply a suite of video cameras and a computer. Tesla's FSD technology could definitely be licensed to other car manufacturers who could easily integrate the technology into their own vehicles. Given this, we see Tesla's AI division becoming more profitable than the actual vehicles themselves. With a vertically integrated model for vehicle production, FSD, and a branded robotaxi service Tesla could emerge as the most powerful force in automotive history.

Yes, this technology will disrupt industries, and it will displace certain jobs in the transportation sector. This is the price of technological advancement and always has been. We predict that it will likely take 5-10 years for the transportation sector to fully transition over to self-driving cars which should give people enough time to realize other career ambitions. We predict that this will dramatically lower both the opportunity and economic costs of transportation for both human and goods movements. There will likely still be specialist services where human drivers would be preferred to AI, but this demand will be greatly diminished for average consumers and would be highly specialized (such as VIP protection services).

When a car is not being actively used it basically becomes a sunk cost. What if instead you could put your car to use by loaning to a rideshare service until you need it back? Another question is, why own a car at all at that point? If self-driving cars were ubiquitous and always within 5 minutes of most Americans, and the cost of a trip was less than $1 per trip, then the entire value proposition of owning a vehicle is diminished for 60-70% of the American population.

Perhaps, the most profound economic effects from self-driving cars will come from the evolution of ridesharing tools such as Uber or Lyft. I say evolution because self-driving technologies will not displace the need for these apps. Quite the opposite in fact, we believe that self-driving cars will greatly increase the demand for rideshare services as part of the robotaxi revolution. What it will do, however, is greatly diminish the need for human drivers. Currently labor accounts for the largest cost of an rideshare trip by far. If you could remove that cost while simultaneously ramping up supply of self-driving taxis, then we would estimate that the cost of an Uber trip could fall by 90% or more. If you could simply pay $100/month for a ride share subscription and get unlimited rides anywhere in the city (or country, even) then I would be hard pressed to imagine the value proposition for continued widespread car ownership in urban environments.

We instead predict that the commoditization of self-driving cars will result in consolidation in the industry and that a few large enterprises will own fleets of self-driving cars and increase or decrease the number of cars on the road in accordance with demand.

If unlimited rides throughout the city would cost less than $100/month, then why would you need to own a car? And if most people no longer need to own a vehicle, then what does this mean for the automotive industry including salespeople, mechanics, financiers, insurers, and all the other supporting industries associated with private commercial car ownership? We predict that the overall economic demand for transportation will always grow, but the demand for car ownership will greatly diminish, at least within dense urban environments. Self-driving taxis will always be available, could theoretically be able to run 24 hours a day, and would stop only to refuel/recharge or for maintenance. This becomes interesting, because now although there is more demand for transportation there is less demand for the total number of cars. The average vehicle is only used for about 5% of its total life whereas self-driving cars operating in a robotaxi network would be the exactly the opposite and would be active for 95% of their life and only dormant for 5%. This means that many fewer cars would be required to perform many more trips.

Now this isn't a sunk cost, this is real savings and increased wealth for the average American. Any loss experienced within the automotive industry would theoretically be offset by potential gains in other sectors of the economy. That doesn't do much to alleviate any pain in the automotive industry, but we believe these changes are inevitable, and it is best to get ahead of the curve. One possible benefit is that although volume of car sales will decrease, the actual price per car would likely increase due to the FSD suites and ongoing subscription services.

Another huge revolution from self-driving cars will be the effect upon the car insurance industry. If no one owns any cars, then no one will need car insurance. Another huge improvement will be the massive increase in safety of the vehicles. If there are no accidents, or if the negative effects of accidents are diminished, then this also diminishes the economic case for car insurance. There will still be car insurance, but the industry will shrink precipitously. Once again, though, the economic loss experienced by the car insurance industry will be offset by spending elsewhere in the economy.

Do self-driving robotaxis need to park in the city? We don't think so. Robotaxi cars would only ever need to park to pick up or drop off passengers or to return to "home base". If this is the case, and 95% of cars in downtown sectors become self-driving, then there will be no need for the huge lines and rows of cars parked on the side of the road and there will no longer be a need for large parking lots to house parked vehicles. This will directly translate to more drivable or usable space in urban environments, which will increase traffic flow and human throughput in downtown environments. This will virtually eliminate the need for parking meters and paid parking lots. This means even more savings for the average consumer and more money into other sectors. It is possible that this loss to governments could be offset by a per milage tax on robotaxi consumption or other mechanisms.

If this trend continues, and the demand for parking real estate greatly diminishes as well, then those parking lots could be reclaimed by urban developers seeking to construct new buildings and offices. This is an interesting calculus because, on one hand, this will have a negative effect on commercial real estate prices in major cities. There will be an almost instantaneous and dramatic expansion in available sqft for new development. Simultaneously, this could also greatly increase the real estate value for parking lots, whose real estate now has the potential for much more efficient uses. The value for the parking lot business will drop, but the value for the real estate of those lots will increase. In any case, we once again translate this as a net benefit for the average American both living and working in the downtowns of American cities. The cities will become cleaner, safer, less crowded, and more economically dynamic.

Robotaxis will have a large negative effect on public sector transportation demand. If the average commuter could pay the same or less for a private robotaxi compared to a bus ticket or metro ride, then why wouldn't they? We believe that most public transit will disappear as robotaxis become more prevalent. We also predict that governments will begin to shift resources towards ensuring access to robotaxis via simple trip subsidies supported by the state. It is also possible that states could run their own self-driving fleets which would help alleviate any diminished access for poorer citizens. We predict that the dollar/person-mile cost of robotaxis will fall so greatly that even public transportation will be unable to compete on a dollar/person-mile basis. The added convenience of a robotaxi that can go anywhere, pick anyone up and drop them off at any location in a private and secure environment is simply insurmountable for something like a bus or metro system.

We also see some possible changes to the suburban environment, though these would likely be less dramatic than urban commercial real estate restructuring. For one, car ownership will likely linger for longer than in urban environments because there simply isn't as high of a competition for the drivable spaces in suburban environments. We do believe that self-driving cars will be present in suburban environments, though these would be a mixture of privately owned vehicles and vehicles owned by a robotaxi service. Robotaxi vehicles which are constantly in motion would be odd in a suburban environment, so those cars would likely remain at a local home base until needed, increasing their time to pick up for drives. We see potential use cases for self-driving cars in the suburban setting, for example, there might be less of a need for multiple cars per household if those vehicles could drive themselves to various pick up and drop off locations without needing two drivers. If families did decide to only use self-driving cars from a robotaxi service, then garage space could be converted into residential space, effectively increasing the livable square footage of the house and thereby increasing it's marketable value.

Probably the most important change to suburban drivers would be the simplification and convenience of a given commute. Suburban drivers would likely be much more willing to commute farther distances if they didn't have to deal with traffic themselves and were able to do work on the way to the office or engage in other non-driving activities. This could expand the suburban footprint even farther away from city centers while increasing worker productivity. self-driving cars will return time back to middle class people which will improve their quality of life and expand economic activity.

Self-driving cars will make long distance travel by road much quicker, more tolerable, and more enjoyable. This will greatly increase the desirability to travel long distances by self-driving car rather than by air, especially if this tavel is a fraction of the cost of air travel. With self-driving cars passengers can engage in many other tasks besides driving such as sleeping, electronic entertainment, working, or any other number of activities which would more enjoyable and profitable than monitoring a steering wheel and pedals.

It will now be possible to travel for 24 hours straight while only stopping for short term rest. The cars could drive overnight and provide a comfortable and safe sleeping experience for passengers. The convenience of self-driving cars would probably outweigh the speed of air travel for some regional traveling such as from the Bay Area to Los Angeles. This would also have the benefit of easing air traffic for regional hops. Although it would be a longer journey by a couple of hours, we believe that the convenience of self-driving cars would make the extra hours much more tolerable than with human driven vehicles. We believe that many passengers would be willing to trade a few extra comfortable and easy hours in order to save a few hundred dollars in air travel costs. Traveling overnight from Bay Area to Los Angeles would be a sinch as you could easily sleep on the way.

This will also make cross-country trips much easier by road as well. Air travel would still be quicker than road for many passengers, but it would now be possible to cut the number of days required for overnight trips by half because the car can now drive safely and comfortably while the passengers sleep or perform other activities. Currently, a human driven trip by car from Los Angeles to Houston requires stopping and sleeping in El Paso but, a self-driving car can make the same trip within 24 hours by stopping only for essential activities while costing less than 1/3rd of the cost of air travel. As self-driving cars become ubiquitous high-speed zones on open highways will emerge where self-driving cars could comfortably and safely travel on open highways at speeds of 100 MPH or more, similarly to the German Autobahn. The cost savings coupled with the ease, comfort, and time savings of self-driving cars will increase the appeal of driving long distances as opposed to flying.