Blog

The past few years have been quite exciting for the automobile industry, with the introduction of the classy, sporty Tesla cars, taxi-service Uber and also Google starting the development of fully autonomous vehicles. Furthermore, if we may believe Ford CEO Mark Fields, the near future could be even more exciting as he expects the first fully autonomous vehicle to be available around 2020. A while ago I decided to do some research into this as I find these developments interesting, especially the fact that a software giant such as Google is getting seriously involved. However, I’m not so sure what to think of the idea that a computer is driving my car. It feels unnatural and somewhat disturbing that my life is completely in the hands of a computer, and what about privacy? – keeping in mind the most likely increasing interconnection between devices in the near future. Despite my reservations, Google claims autonomous vehicles will be safer, more efficient and more enjoyable.

Ok, perhaps a computer can drive a car more efficiently than a human being as it can more accurately determine where to move and with what speed. But I fail to see how it becomes more enjoyable however, as I can’t remember the last time I actually enjoyed looking at the Dutch landscape from the passenger seat. But perhaps it holds some truth if I travel with more people, having the ability to just chill out before going to work or maybe even using the time productively.

What does genuinely concern me however, is the so called improved safety and reliability these autonomous cars are supposed to offer. The idea here is that by removing the weakest link in the chain – the human – the safety will drastically improve. However, Google is a company that has absolutely zero experience designing complex human-machine interfaces and systems safety of such magnitude. Yet they claim they will be able to make driving significantly safer. Even companies in the aviation industry – which have dozens of years designing autopilots and extremely complex human-machine interfaces – aren’t that cocky.

As a fully certified commercial pilot I have quite some experience in safely handling automated systems which (partially) take over controls, but I was trained how to handle these complex systems and know how to respond if they fail. The crux here is the fact that such complex automated systems were designed in such a manner that as a pilot I am required to always monitor the computer so I can take over controls when necessary. One of the reasons flying has become significantly safer is because of this approach to systems safety. When I look at Google and other companies currently developing autonomous vehicles, I don’t see the same approach to systems safety. I mean, what happens if the computer and its backup systems fail? Can the ‘driver’ actually take over controls manually? How is the ‘driver’ expected to suddenly take over controls if he is doing something else meanwhile, and what kind of training will be necessary to safely monitor the computer? Who is responsible for an accident? These questions beg for answers.

Finally, what about privacy? Will the car’s software allow the car manufacturers or software providers to extract more data from their customers? Considering Google’s terrible track record of data mining and software support this could very well be the case; just imagine yourself sitting in the car, booting the computer and being confronted with a start screen that forces you to accept the terms and conditions – which includes accepting your data may be shared with 3rd party companies – before the car starts.

I think there are quite some obstacles to be overcome if we want safe autonomous cars. As for the current concept from google; no thanks, I think I’d rather miss this boat.

Tim Kousemaker