The worst driver? An average driver? The best driver? There's a pretty big difference between all of those. Also, just like with machines being "just as accura

The worst driver? An average driver? The best driver? There's a pretty big difference between all of those. Also, just like with machines being "just as accurate" at translation from one language to another, doesn't mean the machine will be as good as the humans at exactly the same things. A robot that translates from one language to another with equal level of accuracy as a human would probably sound "off" to a human, because it would make different types of mistakes. Similarly, a self-driving car that achieves as few/many accidents per million miles as an average human driver may also seem like it's driving pretty weird compared to a human. > “Regulators may require some significant margin above human capability in order for a full autonomy to be engaged,” said Musk. “They may say it needs to be 50 percent safer, 100 percent safer, 1000 percent safer, I don’t know. I’m not sure they know either.” I would say it needs to be 1,000 percent safer. Better safe than sorry, and at the rate machine learning chips are improving these days, that would only take like a car model generation to achieve. So why not wait a few years? And if human safety is not the primary concern of car makers, because they'd rather be bragging with their "self-driving feature" for their cars, then they should consider that if they "screw-up" this self-driving thing initially, it could ruin it for themselves as well as everyone else for a decade or more. So again, better safe than sorry? Also, the next time someone mentions their cars are autonomous, I want to hear them say in the next sentence how secure software digital systems are against hackers.