You’re driving down a residential street. A toddler darts into your path. Your only choices are the toddler or an oncoming car.
Not every human will make the right choice. But every human knows the right choice. Even the toddler.
Computers know nothing at all. When they do seemingly intelligent things, such as drive a car, they are only blindly following a predefined algorithm. That algorithm must contain not just the decision above, but every other possible decision.
And they aren’t really decisions. The computer has no intuition. It all boils down to a binary choice at a threshold.
This illusion of intelligence is just fine for ordering groceries or playing music. Not so much for life-and-death situations.
Humans have an innate ability to process human error. We apportion blame, reflect on our own fallibility, and wrestle with forgiveness.
But what about computer error? In the situation with the toddler, we’ll blame the only form of intelligence involved, the programmers. At least they’re paid well.
Remember, the toddler is actually intelligent, yet still needs 14 more years of training before taking the wheel.