This week, the Chicago Tribune released an interesting article called “Will your driverless car kill you so others may live?” that presented a few thought-provoking scenarios. Imagine this: You and one other passenger are riding in your self-driving car down a two lane road. You round a sharp curve to find yourself on a crash-course with a family crossing the road (three people). Your driverless car assesses that it does not have enough distance to brake, and that there are three pedestrians in the road. Does the car choose to save the two passenger’s lives and risk hitting the family, or cross the double line and risk a head-on collision?
It’s Not That Simple
Now, the question above could be answered more effectively if you had more information. Are there cars coming in the other lane? Is there room to swerve to the right? Is the family moving quick enough to get out of the way?
But that’s the question: How effectively will self-driving cars be able to compute these complex nuances of everyday life?
Or, Is It That Simple?
The consensus on self-driving cars seems to be that they are going to be safer drivers. But their speed, awareness, and accuracy depends entirely on their computing power, and that’s got us a bit concerned. Computers have already surpassed the brain’s ability to compute and store data, but they are not nearly as efficient. For instance, the world’s 4th fastest super computer, the K, is four times quicker than the human brain at computing data, and can store 10 times more data. But the K also requires 9.9 million watts to function, would cost $10 million to run per year, and takes up an entire building. By comparison, the human brain consumes 20 watts, which is equivalent to a dim light bulb, and is stored completely within our skull.
That’s a pretty big deal, considering most car batteries produce around 500 watts. Will manufacturers be able to build cars with the computing power to out-think the human brain? If not, we can’t see much sense in allowing them to take the wheel.