Cockpit Confidential: Debunking the Autopilot Myth

This is a guest post by commercial airline pilot Patrick Smith, who writes about the hidden side of the airline industry. You can read his writing for here.

Cockpit Confidential: The Autopilot Myth
By Patrick Smith

One evening I was sitting in economy class when our jet came in for an unusually smooth landing. “Nice job, autopilot!” yelled some knucklehead sitting behind me.  Several people laughed. I winced. It was amusing, maybe, but was also wrong.  The touchdown had been a fully manual one, as the vast majority of touchdowns are.

(Digital Vision)

I’ve been writing about commercial aviation for nine years – a job that entails a fair bit of myth-busting.  Air travel is a mysterious realm, rife with conspiracy theories, urban legends, wives’ tales and other ridiculous notions. I’ve heard it all, from “chemtrails” to the 9/11 “truthers.” Nothing, however, gets under my skin more  than myths and exaggerations about cockpit automation — this pervasive idea that modern aircraft are flown by computers, with pilots on hand as little more than a backup in case of trouble.  And in some not-too-distant future, we’re repeatedly told, pilots will be engineered out of the picture entirely.

Don’t look now, here it is again in a new issue of IEEE’s Inside Technology Spectrum, in an article titled, “When Will We Have Unmanned Commercial Airliners?”.

It amuses me that as cockpit technology progresses and evolves, people view the elimination of the pilot as the logical, inevitable endpoint. This is a tempting extrapolation, but is it realistic? Is it even desirable? By comparison, are advances in medicine intended to eliminate doctors? Of course not.

And there’s more to that analogy, because high-tech cockpit equipment assists pilots in much the way that high-tech medical equipment assists physicians and surgeons: it has vastly improved their capabilities, but it by no means diminishes the experience and skill required to perform at that level, and has not come remotely close to rendering them redundant. A plane can “fly itself” about as much as the modern operating room can perform a surgical procedure “by itself.”

“Talk about medical progress, and people think about technology,” wrote the surgeon and author Atul Gawande in a 2011 article for The New Yorker. “But the capabilities of doctors matter every bit as much as the technology. This is true of all professions.  What ultimately makes the difference is how well people use technology.”

Gawande nails it. And what exists in the cockpit today is already a fine example of how progress and technology have improved flying — making it faster, far safer and more reliable than it once was. But it has not made it easy, and it is a long, long way from engineering the pilot out of the picture — something we needn’t be looking for in the first place.

The proliferation of remotely piloted military drones and unmanned aerial vehicles (UAVs) is hugely responsible for fueling the idea that commercial pilots will soon be relegated to a room somewhere to control their flight remotely, the way drones and UAVs are flown today. These machines are very sophisticated and have proven themselves reliable — to a point. A drone, however, is not a commercial jet carrying hundreds of people. It has an entirely different mission, and operates in a wholly different environment.

You don’t simply take the drone concept, scale it up, build in a few redundancies, and off you go. Adapting this concept to the commercial realm on any large scale would require, in addition to huge technological challenges, a restructuring of the commercial aviation infrastructure. We’re talking hundreds of billions of dollars, from the planes themselves to the facilities they’d rely on.  And for what? You’d still need human beings to operate these planes.

And what do terms like “automatic” and “autopilot” mean anyway? The autopilot is a tool, along with many other tools available to the crew. You still need to tell it what to do, how to do it, and when to do it. I prefer the term autoflight system. It’s a collection of several different functions controlling speed, thrust, and both horizontal and vertical navigation – together or separately, and all of them requiring regular crew inputs in order work properly. On the jet I fly, I can set up an “automatic” climb or descent any of about six different ways, depending what’s needed in a given situation.

A flight is a very organic thing – to use a word you’ll almost never hear being applied to commercial aviation. Things are fluid; change is constant; decision-making is critical. For all its scripted protocols, hundreds if not thousands of subjective inputs are made by the crew on any given flight, from deviating around a cumulus buildup (how far, how high, how long), to troubleshooting a mechanical issue, to performing the takeoff and landing.

Emergencies are another thing entirely. I’m talking about the run-of-the-mill situations that arise every single day, on every single flight, often to the point of task-saturation. Duties ebb and flow, it’s true. There are long stretches in which, to the observer, it would seem that very little is dependent on the crew’s attention. There are also periods of very high workload, and they can arise at almost any time. You’d be surprised how busy a cockpit can become – with the autopilot on.

Getting back to that crack about the automatic landing, yes, it’s true that most jetliners are certified for automatic landings. Known as “autolands” in pilot lingo, they are designed for use in extreme low-visibility situations. In practice, well under one percent of landings are performed automatically, and the fine print of setting up and managing one of these landings is something I could talk about all day. If it were as easy as pressing a button I wouldn’t need to practice them twice a year in the simulator or need to review those tabbed, highlighted pages in my manuals.

One thing you’ll notice is how many of the more fantastical predictions about the future of flying tend to be from academics — professors, researchers, etc. — rather than commercial pilots. These people are surely intelligent and their work might be valuable, but often they are unfamiliar with the day-to-day operational aspects of flying planes.

And, yes, airline pilots too are part of the problem, and often our own worst enemies. In our attempts to explain complicated procedures to the layperson, we’re given to dumbing down. We wind up painting a caricature of what flying is really like, in the process undercutting the value of our profession. “Aw, shucks,” says a captain, “this plane practically lands itself.”

That’s a bit like a surgeon saying, “Well, we just cut in and yank out the tumor.” Granted some surgeries might be easier than others, just as some flights are more routine than others. But what does he really mean?

To some, I may sound like a Luddite pilot trying to defend his profession against the encroachment of technology and an inevitable obsolescence.

You can think that all you want.  I am not ag
ainst the advance of technology.  I am against foolish extrapolations of it.


According to the author planes are equipped to do autolands in conditions that exceed human capability. Autoland is used in less than one percent of landings because they are difficult to set up, but not because the system is inadequate. The author seems to be arguing that autopilot interfaces need improvement, not that onboard pilots will always be necessary.

[Insert here] delenda est

Posted on the Freakonomics blog and so soon after Kahneman to boot this is quite funny and deliciously ironic.

I was last in a cockpit when I was 9 in a DC-3(!). But I will bet with great confidence that 10 years from now no-one will be releasing widebody jets designed for manual piloting. The pilot will be relegated to an entirely supervisory role.

I am probably being overcautious, the Dreamliner may indeed be the last such. As other commenters have averted to, a computer, for example, absent sophisticated hacking that could readily be prevented in the constraints of an airplane and flight time, could never be compelled to fly into a building.

The Air France flight which crashed into the sea killing all aboard would almost certainly have never crashed absent human error. The list goes on.

Plenty of other commenters have made the point but the increases in computing power and computer software abilities will leave human judgement for dead in mere years from now in such a straightforward area as controlling a plane (yes, for the right computer, it is straightforward).



"But I will bet with great confidence that 10 years from now no-one will be releasing widebody jets designed for manual piloting."

I'll take that bet.

Consider: The 787 was 'launched' in 2004 (i.e. Boeing committed themselves to making it) by which time it was already the product of some years of design work. If *any* airliner model is to fly pilotless in 2021/2022, there have to be engineers starting to work on it right now. At the most optimistic, there might be one Boeing and one Airbus model flying pilotless in 2022. We could then expect the remaining models to convert at about one per 5 years, putting us at around 2035 before no new piloted airliners are coming off the production lines. (And this is all assuming no basic engineering, regulatory or public relations problems with converting to fully automatic flight.)


I think I'd like to see more of the autopilot:

Mike MacDonald

Anyone who suggests that an autopilot can land a plane filled with passengers should wonder why there's no such function built into their car. And no, "parking assist" doesn't count. Your car never flew into a downdraft or clear air turbulence.

Besides, once the pilot is removed from the loop it should only be a matter of months before someone hacks the autopilot.

I like "human in the loop" systems. Whenever I see those IBM commercials I think "Let's build a planet that can be hacked by a 15 year old in Islamabad."

Bu$y B

Mr. Smith, As a regular reader at I must say that you are absolutely obsessed with this issue; I've read at least a dozen pieces by you on this topic, all making the exact same points over and over. Whether you're correct or not has become irrelevant. This myth/urban legend has become your nemesis and you are wasting your time/energy/sanity trying to fight it; the truth will not prevail, the myth will not die. Give up already. Repeated attempts only call in to question your intelligence, and will do nothing (or very little) towards enlightening people.


Funny how you compare piloting to surgery at the end of the piece. I used to think that doctors and lawyers had the biggest god complexes, but I was wrong. When it comes to inflated egos and grandstanding, pilots give surgeons a run for their money.


great op-ed you got there... i would agree that human pilots are extremely necessary today.

what doesn't help your argument is that 20 years ago, people would have said it's utterly impossible for a computer program running on a laptop to drive cars, and yet google has developed just that and driven it all over california without accidents. a system that's so good that the only time it's crashed thus far, is when a human was controlling the botcar.

all i'm saying is technology does amazing things. we just gotta wait and see.


Hudson River, Chesley Sullenberger; I know where my preference lies.


He's absolutely right. As a flight instructor I can attest that flying an airplane, from the smallest Cessna to the largest Boeing, is an exhaustive workload most every flight. The autopilot, among other systems, is used primarily to do the monotonous job of keeping the aircraft upright and on course while the pilots run checklists, check weather along their route, brief approaches, make fuel calculations, etc. The captain of an airliner needs no less training, practice, and experience than any PhD. Just look at Captain Sullenberger and the "Miracle on the Hudson." Without his level-headed and timely thinking, the lives of 155 people would have been lost. No autopilot can do what he did, no matter how sophisticated.