Airline pilots depend too much on automation, says panel commissioned by FAA

Before the inevitable "why do we even have pilots anymore" posts, let me share a story.

The tldr is: we have pilots so there is a trained human for redundancy when the automatic systems fail.

My father was a pilot, first in the military then commercial. He hated flying commercial, "it's boring, most days I'm a well paid glorified bus driver, unless something goes wrong"

He is since retired, but he likes to say "I only earned my paycheck once in my last 10 years flying".

The story goes, they were just over the Atlantic between Maine and Iceland when there was an electrical fire that threw out some of the automation and controls.

His job that day was to manually fly a piece of metal with about 200 people on board to the nearest airfield where they could land safely and do repairs.

Every other day his job was to run checklists, smile, and drink coffee.

Same thing just happened in Kazan Russia. A pilot was on the landing path and decided that he was too low, too early so he aborted landing, and pulled up the nose to go around again.

Problem was that he was also too slow at that point to safely pull up. The result of pulling up the nose was that he lost airspeed, stalled, and plunged nose first into the runway.

It's counterintuitive to go down and increase airspeed ( or at least remain level while you gun the engines) and that is why these pilots make this kind of mistake. They don't have enough practice flying in full control of the airplane.

An additional factor is age. All of the air force pilots that fought in various wars from WWII onwards have retired. The world actually relied on these guys to fly the commercial passenger planes and build the air industry. Now the industry has to train their own pilots and the gaps are beginning to show.

We're living in an era of unprecedented airplane safety, as no American airline has had a mainline crash since 2001. I'm speaking largely as an outsider, but it seems probable that the increased automation has helped facilitate that trend.

We all have a desire to romanticize and glamorize the jobs we do and resist the belief we can be augmented by automation and process. Some of the pull quotes from this article seem to indicate that mentality. To wit, "Now that is being in command of a system, of wonderful computers that do a great job — but that isn’t flying."

Just like a good programmer should understand manual memory management even if he or she always will be writing in Python, I sympathize with the need to understand the underlying principles. But can we really say safety is at risk?

At some point N years from now, we're going to be reading the same article, but this time about drivers and self-driving cars...

I have a couple friends in the biz and whats (carefully, to meet an agenda?) not discussed is why the automation exists in the first place. Most of their time and brainpower is expended on what HN would call hacking the system, such as taking a decade of experience on that route and more than a decade as a pilot, plus inadequate out of date weather info, plus gossip from other pilots, plus a close reading of the manuals to request a FL 4000 feet lower which under these peculiar wx conditions results in lower fuel burn. Or hacking ATC by figuring out there will be a backlog at the destination because of weather reports 2000 miles away and experience, so instead of burning 2000 extra dollars of extra fuel in a holding pattern, he decides to burn 1000 extra dollars of fuel getting there before the rush begins. Of course this kind of stuff requires the cooperation of ATC etc etc its not cowboys and the wild west, but suffice it to say he does a heck of a lot of thinking about flying while some automation keeps the wings level or whatever.

Another thing is he has a very maint engineering type job continuously prioritizing all possible issues while working on the most critical one. This doesn't mean he flys thru tornados every flight, its more like there has never been a truly flawless aircraft. He won't take off if there's anything on the critical list but there's always "something" non critical or needs special observation because it was just serviced/replaced/needs servicing, basically he's the partner of the mechanics and has something of a "trust, but verify" relationship with them.

So its hardly a one way street. Take away the automation and force him to keep the wings level, manually by hand, for the entire flight, and he is NOT going to be able to spend 15 minutes putting together enroute weather reports and pireps and maps and gut level experience with the plane in that climate and on that route, so that minor rainstorm that turned into a sudden t-storm enroute kills everyone in a downdraft because he was "too busy flying the plane". Or the mechanic replaced fuel pump #2 on engine #2 but something went wrong, and he doesn't have time to watch the pump pressure and temp soar out of control because he's "too busy flying the plane", setting the engine, then wing on fire and they ditch in the ocean and everyone dies. So yeah, automation kills some people, but lack of automation would obviously kill MORE people.

There are also financial issues, where the money you save by efforts at fuel economy might buy you a TCAS system that prevents a mid-air collision... or maybe the money saved just buys some crooked exec another Ferrari. Either way the pilot is going to be required to spend his day on heroic fuel economization efforts.

As in many areas of the economy, there's a zillion applicants for each job. Being a programmer in SV is not like any other career in the country. So there's always some slacking and some slackers, but most pilots are in fact pretty hard working. The bottom, say, 50% will never, ever get hired by the majors...

is the faa american? how do politics factor into this? isn't there a "difference of opinion" between boeing and airbus about the amount of automation that is appropriate?

i see these things discussed from time to time, and i wonder just how neutral institutions are...

In recent years, programmers have come to depend heavily on syntactic sugar. In the old days programmers flipped switches on a front panel to directly hand program opcodes in machine code.

However, in recent years, programmers have gone away from front panels and direct machine language coding. One recent programmer reported on one project, she only wrote assembly language subroutines seven times during the entire project.

Several recent programming disasters have coincidentally happened during this recent fad of syntactic sugar and object orientation and virtual machines. Surely changing the entire way modern programming is done, would have changed the outcomes of those very recent individual modern disasters.

A panel, convened by Hacker News, has formed for journalists, managers, and the general public (no programmers of course) to discuss the safety issues of modern coding practices, and a possible pr campaign encouraging modern web developers to use more IBM mainframe BAL basic assembly language, 6502 assembly language, and/or perhaps PDP-8 direct octal machine code to improve program safety and security in the startup industry.

Some recent reports include "I stole this existing product, wrapped it in bootstrap, in only 3562 lines of IBM 1620 'cadet' assembly language" and "Show HN: Tic Tac Toe in 5219 punch cards on 1920 era unit record equipment"