One problem with these comparisons is the misinterpretation of statistical data. For general policy making statistics are useful, but for personal decisions you

One problem with these comparisons is the misinterpretation of statistical data. For general policy making statistics are useful, but for personal decisions you need to take them with a grain of salt and switch on your brain. Human traffic accident statistics include the totality of reckless drivers, street racers, and drunken idiots. Would you readily be a passenger when they drive? I wouldn't. At the same time, the statistics also include the calmest, best, and most reasonable drivers you've ever met in your life. I think it's reasonable to expect self-driving cars to always drive better than your average drunk college student with 1 year of driving experience and instead perform more like the best drivers you've ever met. So yes, self-driving cars should perform above the human average.