Gone But Not Forgotten

Hello, you've landed on DATA eh? - Open Data Toronto's original blog space for data discussions. This is not an active blog at the moment but legacy posts are still here. Have a read ... you can still provide comments.

Tuesday, May 15, 2012

Giving The Finger To The Google Self-Driven Car

By Keith McDonald
So, by now, you've probably heard the buzz about the Google self-driven car being approved with the first "autonomous vehicle licence" in Nevada right? If not, it's a car, now legal in Nevada, that does the driving for you using GPS, laser range finders, video cameras, position estimators, radar and an inertial motion sensor. In short ... data!

I thought this might be another example to help explain why data is valuable and why we are so keen to have open data out and be as acurate as possible.

one way sign saying DATA eh?

Clearly, if you are going to trust a car to do the driving for you, you are going to want it to have all the data it needs to do so.

Humans Use Data Too
I don't think you ever think about yourself - the human - as a data collector when you are driving but you are! Consider all the technology the Google car is using. This car is going to use all of it all of the time. And it needs all of it because it needs to replicate what a human sees and does - based on what is happening in the moment.

Of course, with all the technology, you could argue that the "robo-driver" will be able to see and respond to situations beyond the moment into the future - the things a human may not be able to process in sufficient time to do anything about. Cool!

That might be the positive around this vehicle. But, I still have to ask: Can we trust this thing to do the driving?

Trust The Data
I think the answer does come down to how reliable the data is. If the car gets accurate and timely data, it's going to be just fine. And here is where cities and open data come into play. In particular, the mapping side is going to be important.

The car needs to know where it's headed and know when it gets there. How many times have you found your own GPS not to be perfectly correct? Anyone been directed by the GPS unit's voice to turn down the wrong way on a one-way street or highway?

That happened to me while driving in Florida. Yep, wrong instruction at the wrong time as a State Trooper happened to be right there viewing the whole thing. I saw quite quickly the mistake (as cars were suddenly bearing down on me). So, I have to say, you really want that street information to be precise every time!

Fortunately, the State Trooper in my experience agreed with my "excuse" and let me go with a warning. But, what if your Google car happens to disagree with what you see as a reality and just keeps on driving? This goes way beyond simple trust in the technology to trusting data.

dead end sign saying DATA eh?

Sure, sure, you can read how the Google car has been tested on highways, city roads and residential streets and how going robotic allows faster reactions, 360-degree perception, no distractions or sleep deprivation issues (or drugs and alcohol effects). It all sounds great but are you still sceptical?

Another added bit of programming is the ability to select a driver personality. You can tell the car to be cautious all the way up to aggressive. That worries me a bit. Aggressive, by their definition is "more likely to go first". I'd like to have the data determine how aggressive you need the car to be! Maybe that's how it will work but if you've ever driven with an aggressive driver (or are one yourself) you know that it can lead to bad decision making. (As a fail safe, they do allow the actual human to take over by touching the wheel or hitting the brake. Sort of like cruise-control right now.)

This too is analogues to how data can be interpreted by many people in many different ways. The advantage to everyone having access to the same data, via an open data web presence let's say, is that you can always use the raw materials to judge how accurate interpretations are. I wonder how this gets calculated in the car and what actions the car takes. Really, the question is how much of driving is interpretation?

Use Data For Decisions
Using data for decision making is, of course, very powerful. Only when you have all the data (some would say facts) can you really make the correct decisions.

It may be easy to understand how important data is to a self driving vehicle but it shouldn't be that much of a stretch to see the same value in other contexts.

I also have to wonder if the Google car has the ability to give other drivers "the finger" or, in fact, what happens if the Google car is ever given "the finger".  Try explaining that to the Police Officer.

DATA eh? #24

No comments: