Will Robo Cars Ever Be Intelligent Enough, Or Is That the Right Question?

For those, of which there are plenty, who have been pooh-poohing the idea of robo-cars and predicting confidently that autonomous cars would never develop the necessary intelligence to handle the complex interactions on our streets, they might be barking up the wrong tree.

The problem with this skepticism is not so much the fact that navigating our streets is not complex and full of un-predictabilities which could overwhelm the most robust software algorithm. Rather, it is the lost opportunity to solve a simpler problem by re-defining the problem domain. That is, what are the causes of such complexities and what problem are we trying to solve.

Before we look in the mirror and point fingers at ourselves, first let’s use some numbers to set the context.

Some statistics from the 2013 report by the US National Highway Traffic Safety Administration (NHTSA):

32,719 fatalities in traffic accidents

2,313,000 injured

Fatality rate per 100,000 population: 10.35

Fatalities per day: 90

Pedestrians injured/killed per day: 181 / 13

94% of auto accidents can be attributed to humans (i.e. not due to mechanical problems with the vehicles)


32,700+ fatalities is actually a good number – the number, which has decreased by 25% since 2004, would have been much higher if you sample those from prior years. So congratulations all around to human drivers!

For those who were unfortunate enough to get killed or injured on the roads, they just sadly become a statistic. We collectively manage to glance over these stats or rationalize them with phrases like “collateral damage” or “that’s the cost of the way we live”. Unless the victim happens to be you or someone you know and love, that is.

With these grim statistics as a backdrop, Google recently began releasing regular reports on its self driving cars as first reported here (check out The Future of Driverless Cars). Gleaning from Google’s self-driving vehicles log books and slow-mo-replaying how the robo-cars interacted with human driven vehicles, pedestrians, cyclists and other moving objects using the real time data gathered,  Google claims that it is learning as much about human driving behavior as they are providing road lessons to its vehicle software.

Since all accidents involving the Google cars were minor, which, similar to those countless minor accidents which do not get reported, Google is able to offer a somewhat unique perspective on human driver behaviors which are otherwise not quantitatively analyzed due to a lack of data available.  A lot of these behaviors are anecdotally and intuitively known to us as we have personally experienced from (or done to) someone in the past. Nonetheless, it is still refreshing to have those intuitions confirmed.


Google car right turn
Google car witnessing a two-lane right turn – A car (the purple box touching the green rectangles with an exclamation mark over it) decided to make a right turn from the lane to the left of the Google car, cutting sharply across its path. The green rectangles, which Google calls a “fence,” indicate its car is going to slow down to avoid the car making this crazy turn. (Image: Backchannel)


Below are a few snippets opined from a recent article  from a director of the Google self driving program.

  • Google cars were involved in 11 minor accidents during 1.7 million miles of autonomous and human driving. Not once was the Google car the cause of the accident.
  • Rear end crashes are the most common (got hit seven times), mainly happened at traffic lights when the human driver behind ploughed into it.
  • The majority of their accidents occurred in city street driving.
  • Lane drifting and running red lights are leading indicators of significant collisions, as confirmed by this Department of Transportation report.
  • Intersections can be scary places – 21% of all fatalities and 50% of the serious injuries involved intersections, and injuries are usually to pedestrians and other drivers, not the drivers running the red light.


Google car cut off on left turn – a car in the leftmost turn lane (the purple box with a red fence through it) took the turn wide and cut off the Google car. In this case, the red fence indicates the Google car is stopping and avoiding the other vehicle. (Image: Backchannel)
Google car cut off on left turn – a car in the leftmost turn lane (the purple box with a red fence through it) took the turn wide and cut off the Google car. In this case, the red fence indicates the Google car is stopping and avoiding the other vehicle. (Image: Backchannel)


Which brings us back to our original question at the top of this article: What problem are we trying to solve?

Denial and a bruised ego aside, it should be clear to us mortals by now that the complexities of our streets are by and large the results of human drivers – human drivers are the causes of the complexities.

Once you manage to condition your mind and re-frame the problem, you begin to see the solution in a whole different light:

Would you rather come up with a software solution intelligent enough to deal with the complexities, or try to tackle a simpler problem by removing the cause of the complexities in the first place – the human drivers?

Imagine the day when, in addition to car free zones, we have robo-car zones within the urban areas where us carbon units are only allowed to be passengers. You only get to punch in where to go and hit the start button. You no longer have to deal with reckless drivers and unsocial behaviors on the roads. You don’t get to burn rubber or show off your impeccable driving skills, either. Sorry James, you will have to find something else to impress your Bond girls.

So, again, what would you rather pick:

  • Buggy software occasionally suffering from the famous blue screen of death (or is it only a Windows ghost from the past?)


  • Drunk drivers
  • Distracted drivers texting or otherwise tending to their electronic devices
  • Male drivers with excessive testosterone trying to prove their manhood the only way they know – by going at excessive speeds
  • Sleep deprived drivers
  • Drivers who had a bad day or otherwise emotionally distressed and on a short fuse
  • Drivers who are in a great hurry, legitimately or otherwise
  • Impatient drivers
  • Just garden variety incompetent drivers
  • ….


This article is part of the Automation in-depth topic. Get a crash course and read the latest developments on this topic.


Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes:

<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>