I agree with most of what you are saying here, but didn't see anywhere you credited it with the ability to stop making mistakes, so am unsure of your stance there, do you think it will always make mistakes, or does it / can it learn from mistakes in your opinion?
Personally I can't see why not, and this is something it has supremely in its favor over humans, we are almost incapable of truly learning from mistakes, if we are honest, no human has perfect memory, our human memory pretty much always twists things, no matter who we are, it will happen to each of us sooner or later, when we get older or just ill. I can see how machine learning can be made completely robust to that.