For those like me who, though not a pilot, enjoy watching Aircrash Investigation and similar programs, and finds the investigations and findings inspiring/sobering/educational as a software engineer, and like me weren't aware of the "Children of [the] Magenta" phrase, this article and audio piece [1] looks a good place to start. The phrase was used in a 1997 presentation - available publicly [2] - by American Airlines captain Warren Van Der Burgh, who said that (quoting the article's summary of his words): "The industry has turned pilots into “Children of the Magenta” who are too dependent on the guiding magenta-colored lines on their screens." He makes his case with some very funny lines, too. Quite the humorist: "I got stuff to worry about, don't I. Heads up stuff. It's probably not a good time to be typing." [3] immediately precedes his explanation of the phrase.

[1] http://99percentinvisible.org/episode/children-of-the-magent...

[2] https://youtu.be/pN41LvuSz10

[3] https://youtu.be/pN41LvuSz10?t=510

It's an area of ongoing research as well, e.g. the Humans and Autonomy Lab at Duke. (Missy Cummings, who is the director of the lab, used to be a Navy fighter pilot.)

I expect we'll hear a lot more on this topic. Assuming you're at all skeptical about the timeframe for full automobile driving automation, it seems a certainty that we'll have a (significant) period when stuff works most of the time--until it doesn't. And, given that lots of people drive distracted today, what's it going to look like when the car can drive down the highway most of the time on its own?

Isn't this a false dichotomy, e.g., shouldn't it be "The problem is relying on automation in addition to real piloting skills"?

Look at the space program, for example. Extremes of forces and very short windows for decision-making have moved spaceflight in the direction of automation for most mission critical moments. The data that feeds in-flight automation is the same that feeds simulators, so I am not sure where the difference lies in your differentiation between the two.

I can see a point where doing everything by software rather than human is safer and makes the most sense (see the direction Tesla is taking us), and at that point there will be no issue. But that is not the case now.

I don't see your point

> The data that feeds in-flight automation is the same that feeds simulators

The problem is when the data sources fail, like in AF447.

Or when you have people crashing an airplane in perfect weather, like Asiana Airlines 214 in San Francisco

Yes, automation makes things easier, but they fail eventually. And when it fails you have to fly it manually (or with a lower level of automation). But some modern pilots are incapable of doing that.

Well, I am not arguing for or against automation (although ultimately as an optimistic programmer I land very much in the for camp.)

What I am saying is that simulation time alone is not good enough. Nor is flying-only time. Both are needed, and for different reasons. Sure, time spent in automation mode in the air is not incredibly useful. But it is reality, and enduring that, only to have to take manual control in a panic after 8 hours of boredom is something I don't see simulators re-reproducing very well any time soon.

You don't put the pilot in an 8h training flight. You train for specific failures in different situations in a simulator (take off/cruise/landing). And you're right, there's no training for "doing nothing" for 8h and then having an issue

Level of automation you're using has nothing to do with flying on a simulator or an actual plane.

What you're saying is absolutely correct but it has nothing to do with the issue being discussed in the linked article.

You can fly in a real aircraft for hours doing nothing but twisting autopilot knobs. OTOH you can practice basic airmanship skills in a sim.

EDIT: sorry, I realized that you were basically saying the same thing.