Mystery Chart
Below is a chart of the valuation of two things over time.  Both of these things are revolutionary. Thing 1 is going to change transportation, while thing 2 is going to change finance. 


However, with any new technology, both of these things have problems.. Thing 1 is only available in two markets, needs significant regulatory approval, and is very costly to scale. Thing 2, well, try telling people you own Thing 2 at a dinner party and see what the reaction is.

Thing 1 - Morgan Stanley’s estimated valuation of Waymo, Google’s self driving car unit.

Thing 2 –  Bitcoin.

The question I pose
Which do you think you will do first: buy a coffee with Bitcoin, or get a ride via an autonomous car?

The answer may surprise you
We (myself included) have gotten ahead of ourselves with autonomous cars. It always felt like they were right around the corner, ready to pick us up in a few years. The valuations of companies stoked our beliefs. We mentioned Waymo’s ludicrous valuation, but others are similar. GM bought Cruise, an autonomous driving company for $561 mn in 2016. Three years later, GM brought additional investors into Cruise, at a valuation of $19bn.   

For all of the hype and fanfare, no one has an idea when autonomous cars will come, how safe they will be and if people will use them. The conversation has turned the opposite way, with some debating how self driving cars could be as difficult to solve as artificial general intelligence. One autonomous car CEO thinks self driving cars will be phased in over the next 30 to 50 years.

If self driving cars are so far away, why did we think they were so close?
First, autonomous cars do operate in the United States. Aptiv, in partnership with Lyft, has been using about 30 self driving cars in Las Vegas as of May 2018. The cars have two Aptiv engineers in them at all time, just in case something goes wrong, but the cars are driving themselves.  Waymo has an undisclosed number of self driving cars in Phoenix, Arizona, and those on a beta list can summon a car at any time.

These accomplishments come with enormous caveats. Both of these locations are relative “grids”, meaning there isn’t a lot of complexity to driving. Also, both of these areas are warm and have sunny environments, where things like snow can’t get in the way. Once you add in more complexities like bad weather, roundabouts or moose on the road, things become a lot more difficult. We just assumed since self driving cars are available now, soon they will be in our backyard. Not the case. 

Questionable safety metrics have also mislead us
The best metrics we have on how safe self driving cars are comes from “disengagement” reports. Companies testing cars in California are required to release disengagement reports, which show how many times a human driver had to intervene in a self driving car. We were encouraged to see these rates drop dramatically. Waymo reporting a disengagement rate of 0.64 per 1,000 miles in 2015, which dropped to 0.09 in 2018.

But even at 0.09 disengagements per 1,000 miles, self driving cars aren’t everywhere. When you think about what this metric means, it doesn’t mean anything. Is a autonomous car with a 0.064 disengagement rate better than a human driver? Is a 0.00064 disengagement rate better than a human driver? I really don’t know.

And then politics...
Even if self driving cars were completely ready, since there is no “federal” approval criteria, approval needs to happen at the local level. Local government relationships with startups like Uber and Lyft hasn’t been smooth and Uber has managed to get into conflicts everywhere. Part of this was Uber’s general conduct, but any time taxi revenue is threatened, politics arise.

Finally, Elon Musk
Elon Musk’s job is to hype autonomous cars. In 2015, Musk said it would be two years before autonomous cars came to market. Two years later he made the same statement. Two years after that, he made the statement again, this time predicting a million autonomous Teslas will be driving around. I wonder what we he will be saying two years from now...

Even if it takes 20 years for self driving cars to arrive, there is one mega-trend which will lead to the inevitability of self driving cars.

Young people don’t want to drive
Teenagers are giving up the ability to drive. In Detroit, about a quarter of 16-year-olds had a driver's license in 2017, a sharp decline from nearly half in 1983, according to an analysis of licensing data by transportation researcher Michael Sivak. Think about that - the greatest thing you get as a teenager - freedom to go everywhere, teenagers are turning down.

Instead of turning driving down, they just use Uber. The article cites Uber/Lyft as one of the main reasons teenagers are putting off driving. Survey data confirms this. A total of 45% of 18-29 year olds used ride sharing, which was the highest of any age demographic, according to a poll done by Gallup.

It all makes sense. First, if someone can drive you instead of driving yourself, that’s great. Two, cars are expensive. Without the help of parents purchasing a car, teens are unlikely to afford it. Instead, if rides are $10 each way, thats a different story. 

In conclusion. 
No one knows when self driving cars will become mainstream.Maybe in two years, maybe in five, maybe in twenty.

But two things are for certain: teens aren't learning to drive and you can buy a coffee with Bitcoin.

Like this article? Have insights and analysis about everything machine learning/artificial intelligence delivered to you weekly, by signing up here