I keep seeing the automation topic come up. I think most folks don't realize that the lunar lander, with Neil Armstrong, operated one of the first autopilots using essentially similar logic as modern automation (and, or, nor, etc.) It is also a well known example of an autopilot becoming incapable of execution--Neil manually overrode the autopilot and landed. The attempts to make flying automated go back at least that far. Computers are really good at "learning" repetitive tasks given enough iterations to perform it at an acceptable tiny margin of error. However, when one factors in those pesky non-normal events of low occurrence, the computer is going to fail in those scenarios at a rate of 100%--only the human brain at this point has the ability to think outside the box when something different happens. I think of examples including UA 232 where it was "impossible" to loose all hydraulics, US Air 1549 with CA Sullenberger and WN 1380 CA Shults. How steep is that autopilot going to execute an emergency descent with its "default" response? Maybe I don't want an ultra rapid descent because there is a hole the size of Aloha Airlines 243 and the rest of the plane could break up...

$h!t happens...passengers want human pilots to deal with it.