So you are implying that software, created by buggy humans, will somehow be more reliable? Consider also that self-driving cars will depend upon hardware. That hardware will break or become degraded by dirt, debris, etc. Those promising self-driving cars are greatly over-simplifying the challenge. At this point in time we flawed human beings are far superior to any self-driving car. I suggest humans will remain superior for quite a long time...
I'm beginning to think your job is "car salesman to the uneducated". You spend more time posting about your hate of self-driving cars or electric cars and how much you hate "progressives" than you do on any other topic here (I just checked to make sure it wasn't because I primarily see your drivel in the front page forum, apparently it's the only place you really spend time posting).
Yes, software is more reliable than us. Whether due to Bayesian models leading to reduction in variance, engineering processes that allow for individual shortcomings to be handled better by many, or just the fact that we have late nights, stressful situations, frustrations caused by the actor-observer asymmetry bias (e.g. "that asshole cut me off!" versus "oh man, I didn't realize that lane was the right one, glad that person is letting me in"), and other such situations that make our bugs not only common, but unpredictable.
You claim to be better than "any self-driving car". If we were to put together a test of that, what would be something that would convince you otherwise or affirm that you are? Would the same test apply to the driving population?