Google is heavily promoting their driverless car, suggesting that it's the future of mobility, that the driver is going to be unnecessary to get around and a driverless future will be safe and easy. Having navigated around a city using their mapping software, I am not quite so confident in their declarations.
There are possibilities with a driverless car that are not possible with people behind the wheel. The cars all can talk to each other in ways that people cannot, which would mean that accidents could be prevented somewhat. The majority of accidents involving people are basically examples of miscommunication, as a person does not know what the person going in the other direction is actually doing. With that out of the way, it is theoretically safer.
I use the word theoretically deliberately because I am not sure that it would actually be safer, given that computers work in their own little world that is not always aligned with what makes sense in reality. For example, Google maps loves suggesting people do U-turns in places where a U-turn is either completely illegal or just generally very unsafe. Now, this might be possible in a world with no drivers, though I suspect even then you're just not going to have an opening. It is not at all possible in a place where there are people behind the wheel, because it is simply not safe in any sense of the term. Yet, this is something a computer thinks is a great idea. It's also not necessarily precise, getting you within the general area of your destination. If you are driving that's fine, because you can see house numbers and business names and know you're not far from where you need to go. If you have no control, that becomes a major issue, because it will need to park and you might be several blocks away from where you actually need to park.
These are minor bugs, you might say, that these little hiccups with the maps themselves could be ironed out and everything will eventually become safe and reliable. The problem is, these are the bugs we know about because we have the software they already have available. Software is only as smart as the people programming it, and the problem with these self-driving vehicles is that we're always going to be encountering bugs, always going to be discovering things that software simply does not know how to handle, and finding out ways that a driver would be very handy for actually getting from place to place.
Cars themselves often need a driver just to go from place to place. Snow confuses most traction control systems, for example, and while it continues to get better, without a driver you're just going to have streets filled with driverless cars smack into trees and stuck in the middle of the street.
Google claims that they haven't had any accidents where their driverless car has been at fault. That's believable, but they also have largely been testing in California on highways, and in areas where the team is familiar with the terrain. That's clear enough when their navigation systems make incredibly stupid decisions throughout Saskatchewan cities, places where Google employees don't live, work and know intimately. I want to see them throw one of these machines into Saskatoon in a snowstorm, and then we'll see if they are still touted as everybody's future of mobility.
