Robocar "Shark Tank" issues: Cheap rides, vandalism, jobs, dystopia and more


For the second year at the Automated Vehicles Summit, we held a "Shark Tank" where there were 4 pitches on controversial ideas in robocars, and the 4 sharks (including myself) and the audience debated them. While these breakout sessions are on Chatham House Rules, I can certainly outline my own views.

How cheap will robocars get

The presentation did not forecast costs as low as I have speculated of as low as 29 cents/mile but suggested there were big effects at even 50 cents/mile. Private car ownership is estimated at between 40 and 60 cents/mile, plus parking, but Uber is a bit under $2/mile. It is true that having Uber style service for 50 cents/mile will have a big effect, since you can buy it by the mile, unlike a private car where miles are effectively bought "in bulk."

The 50 cent price becomes important in cities with significant cost or difficulty of parking, and when compared to the cost of a transit ticket. Transit tickets are ranging up as high a $3 in many cities (more in Europe) and a large fraction of transit trips are well under 6 miles. In addition, the robocar price is unsubsidized, while most transit trips are highly subsidized, costing from $5 to $8 per trip in reality.

We're already seeing changes to parking. Many people do not use airport parking any more on anything but very short trips. Stay more than 2 days and an Uber ride to the airport is cheaper, especially when you add the value of your own time and the ability to be dropped at the curb rather than take a shuttle.

Another recent report indicated that in LA, valet parking at clubs and bars is way down, to the extent that a located that used to valet park 80 cars might today do only 15 -- this is again due to Uber. It has gotten to the point that Valet parking companies are charging the venues to set up valet parking, where they used to make their money from the parking fees.

Also discussed were things like subscription models (Lyft just started experiments with that) and, as I wrote earlier, how employers might pay for your robotaxi commute.

Will robocars get covered with vandalism or art?

In many cities, things get covered with graffiti, particularly subway cars and buses. Private cars tend not to get this so much, so it's hard to say how robotaxis will fare when they "park." If there is opposition to the robotaxis (just as we have seen for Uber from taxi drivers) this could encourage memes of destruction.

Fortunately, our Starship robots have seen very little trouble from the public. If people do mess with them, we have operators in our control center watching and talking to them, and calling the police if needed, and so with robotaxis. Do you want to commit a felony on camera?

Other, less connected devices have not fared so well. Recently, a Chinese dockless bike-share company had to abandon Paris and other cities because a meme started spreading to vandalize the bikes and dump them in the river, posting instagram photos. We've also seen public objection to dockless scooters like Bird and Lime, which sometimes get left in annoying places on the sidewalk, bothering pedestrians.

Robotaxis can move, and if need be, they can wait in more secured locations at least overnight. And they can even start moving if somebody is trying to vandalize them, though a group of vandals could encircle it in a way that it would not be able to move, not wanting to harm people. (It is an interesting question about whether a car being vandalized, under remote control from a control center, might be told by an operator to keep pushing -- with warning -- if a vandal stands in the way, and whether that would be legal or accepted.)

Will there be major job displacement in trucking?

I covered this in a previous post. New research is suggesting the job displacement from driving jobs, while not trivial, will be much less significant than is often suggested. When it comes to taxi/Uber driving, this is usually not a chosen long-term career, but rather something done as a stopgap, and so the attrition out of such drivers will be smooth as they move into other gig jobs. There will still be many roads for decades not in the maps of self-driving cars, providing work for those who wish to stick it out.

A similar story appears in trucking, particularly long haul:

  • Currently in the USA there are 250,000 open job reqs for truckers. Of course this shortage is only at the current pay levels, but even so, the first 250,000 robotic trucks will not affect the labour market.
  • Turnover in long haul trucking is huge, sometimes more than 100% per year. That's actually a combination of some number of career truckers with even higher turnout on new entrants. As such, it's not hard to imagine the next million robotic trucks will not affect the labour market due to attrition.
  • Truck drivers perform many tasks other than driving. Some estimates (more on short haul) suggest half the labour is not driving. As with many other jobs displaced by machines, career truck drivers will move to other tasks. Even if they need to be with the truck, networking will allow many tasks to be done from within it.
  • Truck driving kills 4,000 per year in the USA. It's not going to be easy to get special laws to keep it going.
  • The change will be gradual, and people will slow down in entering the field. In addition, special efforts will be made to assist in retraining and movement to other lines of work.

Is it utopia or dystopia?

While the planned presentation was a no-show, the room turned out to be highly biased, with few thinking a dystopia was coming, at least in the long run. In the short run, increased congestion is quite possible as new vehicles enter the road, replace public transit and make rush hour more tolerable. In addition, as vehicles make traffic jams more tolerable and as more people get access to car transport

Down the road, there is hope for different types of congestion management -- congestion charging, road metering and other policy incentives, along with private tools like Waze naturally redistributing traffic away from congested routes.

Driver's licence for a car

At first when I heard the idea, proposed by Gerben Feddes at the Netherlands road authority (RDW) for a "driver's licence" for self-driving car software. Physical cars have to go through a "type approval" process which involves a long checklist of things that need to be present and working as described. But software is a different story, both because regulators don't know how to certify it, and because it is a complex and constantly changing thing that the hardware methodologies can't work on.

While not fully fleshed out, the idea would be to have special examiners who exercise more human judgment to confirm the software is designed well and works properly. Such a system could never be perfect either, but nobody will ever be able to test a car as much as its own creator will test it. (That is the reason that in the USA, most thinking is around self-certification, where the vendor certifies it meets safety levels and suffers legal consequences if they lied.)

The driving test given to humans is of course very simple (even in the countries where it is stringent) and relies highly on assumptions about what humans can do. In 2012, I actually researched if the Google car would be able to pass the California driver test, and I concluded it mostly could -- at the time it was not ready to do an unprotected right turn onto a high speed street -- other than the "interface" where a driving instructor gives verbal instructions like "Take the 2nd left up ahead." The car could pass but it wasn't anywhere close to ready to deploy. It would have been a publicity stunt.

Old ideas persist

While the field and conference are maturing, several ideas are still very popular among many of the attendees. These include:

  • The SAE/NHTSA "Levels" -- though more and more people are rejecting them and one of the academic posters was a refutation of them.
  • A strong transit-oriented style of thinking, viewing robocars as mostly a "last mile" solution for transit.
  • A believe that if robocars are not shared (ie. multiple riders) that they are bad. Shared rides will play a key role, but it's not nearly as important as some people suggest.
  • Strong faith in the importance of V2V "connected" cars, in spite of security risks.


Re the eventual cost/km of TAAS, what is the IT infrastructure to support it? If that is not a large proportion then I wonder if people making the estimates are considering the possibilities of radically smaller lighter vehicles?
Re Congestion management, TAAS gives a much more realistic possibility for users to respond to both congestion pricing and HOV lanes. HOV lanes obviously don't work because car pooling is not easy to organize for privately owned vehicles. Congestion charging likewise is just sucked up when there is no public transport alternative. As you have pointed out before the cost/mile difference without congestion charging is not going to be sufficient to encourage people to share rides, but roading authorities stand to make massive savings if they can encourage sharing and the resulting decrease in new road building.
On Utopia/Dystopia, leaving aside that happiness is largely independent of these things, I think one area it should improve drastically is noise and pollution levels simply because robocars lend themselves to being electric. That alone should make the city a more pleasant destination.

I have another piece coming up on that -- without necessarily an answer, though.

You don't need "public" transportation -- by which I refer to government mass transit -- to reduce congestion. Automated ride sharing, like UberPool, is private and not in large vehicles, and it can also do that. Technically, UberPool is "public" transportation when that is defined as a transport service available to the public, though at present UberPool requires a credit card so not all can use it.

I agree with you that public transport is not required for people to respond to congestion charging.
I think the reason that the response is not an overwhelming shift to Uber pool is because Uber Pool is not yet cheap enough to use as every day transport.
even then people grossly underestimate the cost/km of their private vehicle the overall cost .

The reason this could change with robotaxis is cost and critical mass.
Cheap means more people will use it.
More people using it means more opportunity to pool.
Means there is an easy way for people to react to congestion charging.

I am hopeful this will happen, but who knows?

Uber Pool and Lyft Line don't have critical mass yet, but in the cities they report 30-50% of rides being pool which is pretty good. The price does vary, but a typical modest ride in San Francisco seems to be about $4. Muni is $2.50 (if you don't have a pass) but the real cost of a Muni ride is much higher. In 2016 they had 26% Farebox Recovery, meaning that the fares only paid for 26% of the total cost of the system. What this says is that even solo Uber is already cheaper than Muni, and UberPool is much cheaper.

Add new comment