I disagree with your overall summation of human intelligence, although I do agree with some of the human scenarios you posit regarding driving behaviours.

AI, however, cannot become competent because competence is behaviour judged against a set standard and, by definition, AI will set its own standards and, in fact, already does so.

AI also begins with human input and algorithms and although it may be able to self-learn over time, as with all learning, if there is an error in a prior step then it can affect all subsequent ones.

The reality is that AI is only a mimic of human intelligence and intellect, it simply works much more rapidly. I suggest that whilst that may have many benefits when there is appropriate human control, when there is not events may be triggered so rapidly that irreversible catastrophe is both possible and likely.

If that's not something of which we all ought to be wary then I don't know what is.