In at least two of those cases they had a point.
In January 2009 Captain “Sully” Sullenberger saved the lives of 150 passengers by making an emergency landing on the Hudson River.
No, they don't have a point with this case, for two reasons.
First, computers could easily be trained to do this. Sully taught pilots in simulators for years after the incident. Computer systems can learn from that data.
Secondly, Sully could land the plane because he had very precise extra skills as a glider pilot, which commercial pilots aren't trained for. In other words, out of all the pilots in the world, he was one of a handful, or even the only one, who could land that plane safely. Every other pilot would have ended up killing someone. If a computer is trained in this landing, then any aircraft that is running with that software can survive.
And this doesn't even go into the fact that most aircraft fatalities are due to pilot error. If machines can eliminate those, there will be far fewer fatalities, even counting the extreme cases that the software can't handle.
Other parts article are congruent with your second point.