The Uber Approach to Autonomous Driving

Uber has been in the news recently for its controversial rollout of a driverless pilot program in San Francisco. While I can applaud the zeal with which Uber wants to make autonomous vehicles a reality, there are several dangers to their approach that could end up hindering their cause more than helping it.

Levels of autonomy

Before delving too deeply into the pros and cons of Uber's approach to making driverless cars a reality, it's probably a good idea to review the various levels of autonomy in use today.

At one end are the smart features that many modern cars already posess, such as the ability to break autonomously if the car is travelling at slow speeds and detects a sudden obstacle in the way1. Autonomous parallel parking is another feature sure to find many users.

Then there is Tesla, with its controversially-named Autopilot system. It can automatically maintain and change lanes as well as go in and out of a garage by itself among other things. That's not to say that it isn't capable of doing more. Due to the complexity of regulation (and no currently established legal framework) for driverless cars on the road, it must necessarily be restricted in its level of autonomy.

Uber slots in here, at the more autonomous end of the spectrum, with a computer making decisions and a human as a simple fail-safe.

At the far end is Google's program (which was recently split into its own company called Waymo owned by Google's parent Alphabet), which aims at completely driverless cars, able to operate in any environment and situation. This is perhaps best exemplified by their original desire to build Google Pods, cars that would come without any driver controls, except for a navigation system. Though this plan is now axed with Google planning strategic partnerships with manufacturers.

The motivation for driverless

Uber has perhaps more motivation to go driverless than any other company. While Tesla and Google's stated aim is to reduce the shocking number of road fatalities across the world, Uber has perhaps a much more mundane aim: profit.

Far be it from me to demonise Uber for this, every compnay's final aim is to make profit, but it feels like for Uber driverless cars are a means to that end, rather than a by product of making the world better.

Uber employs around 1 million drivers worldwide in 300 cities. Drivers are self-employed, so they are technically not a cost-centre for the company. They do still take home around 75% of the passenger's fare. So it might be surprising that Uber is currently on target to end the year on a $800 million loss.

If Uber managed to replace drivers with a fleet of driverless cars, the full 100% of the fare would end up with Uber. 2

The problems with Uber's approach

Having established that Uber will do well by replacing it's driver population it's perhaps unsurprising that they are pushing hard for driverless technology to go mainstream. With public and regulator acceptance of driverless technology, Uber would be free to implement its plans. The problem: It's pushing too hard and moving too quick.

It begins with Uber's refusal to register it's program and obtain a testing license. At $150 for 10 cars and $50 for every subsequent 10 cars, it is likely not the monetary side that is holding them back. Uber itself states that it

does not need a license as the vehicles aren't truly autonomous

By which it means, there is a driver watching over the vehicle ready to take control in case of a problem. The issue here is that by the same definition Google's test vehicles are also not autonomous, since they all come with a safety driver.

The likely reason for refusing to get a test permit, is not so much the cost, but the need to report on any incidents the vehicles are involved in. Such as the [fender-bender][] one of Google's vehicles was involved in earlier this year. This is important to build public trust in these autonomous systems. When companies can run [red light][] or stop signs or dangerously cross bike lanes, it undermines that trust. This in turn will make it harder to pass legislation that allows widespread adoption of driverless cars.

Uber could very well be shooting itself in the foot.

Furthermore, Uber has jumped into the hardest of all environemnts: Dense urban traffic, including lots of bicycles. Tesla boats an impressive 140 million miles covered under autonomous conditions, but this is misleading. The vast majority of this is done on highways, where situations and actions are much easier to detect and predict.

Uber on the other hand is jumping straight into the dense urban jungle of winding roads, fast bikes, unpredictable pedestrians and impatient commuters. Uber's cars are unlikely to spend much time on highways, rather continously crossing San Francisco all day long.

Final thoughts

Every new technology comes comes with two competing approaches: theory and practice. Spend too long on the theory and someone else will get there before you. Spend too little on theory and you end up with a product that is too green for mainstream consumers.

In the case of driverless technology, perhaps a little more time on the theory will ensure a green product doesn't kill the practical applications before they're even on the market.


  1. This is said to have prevented even more deaths during the mass murder at the Berlin Christmas Market in December 2016 

  2. I imagine Uber would end up with a franchise-style business of providing software, cars, etc to local operators, rather than managing every car itself. This would also allow it to distance itself from any operating issues, much like it has done with its drivers currently.