Aircraft have more mechanical failures then hijacking attempts. The labor savings are not significant enough to warrant autonomous passenger/cargo jets.
> In the meantime, it seems unlikely that we'll see mass unemployment of pilots. Mandatory retirement will see 42% of US pilots retire in the next ten years. If anything, this shortfall will push the automation case forward faster than ever.
This is a misnomer. There is no pilot shortage. Next time you fly a commuter flight, ask your pilot how much they make (it's about $20k-50k/year). There are an overwhelming number of available pilots in the training pipeline.
Just as with truck drivers, its not a labor cost problem. Its a corporate disdain for cutting into profit margins.
"The annualized driver turnover rate at large truckload fleets was 74 percent in the first quarter and the industry was short about 48,000 drivers at the end of 2015. That shortage is expected to balloon to almost 175,000 by 2024, according to the American Trucking Associations."
"Raising salaries is another option, but it would cut into profit margins. The University of Pennsylvania’s Viscelli said companies would have to double driver pay -- now about $41,000 -- to reduce turnover. About 30 percent of new drivers quit in the first three months, according to Stay Metrics."
The article suggested that the price savings would be (relatively) most significant for private jets, which makes sense. I'm wondering if this might finally make for a viable small-jet-hire industry that caters to "normal" people? It seems that every few years we hear about a "jet taxi" startup that promises cheap(ish) flights between general aviation airports but they never seem to materialize...
My wife is an Airline executive and I just showed her your post, she had a good laugh about your comment on labor and wages. Labor is the 2nd greatest cost after fuel. There are a lot of pilots making well over 100k - 250k+ a year. The problem is not cost. They can pass that along to the customer. It's union contracts.
The top 25 legacy airline executives make, cumulatively, a bit under $90 million/year. The average annual wage per airline employee is about $60k/year. The median pilot wage is ~$120k/year (the average is ~$80k/year).
EDIT: (crew is the lowest component of per hour flight time)
B757-200 FOC
Costs per block-hour of operations (avg. 186 seats):
Don't ask your pilot what they make. Honestly! This is basic etiquette. Just realize that regional carrier pilots and flight officers (as well as crew) work really hard for their money. They make respectable money on major carriers, though, and a lot of the regional pilots are probably trying to advance to those positions.
Also, how many years of lead time and training do you think is necessary before piloting a commercial aircraft? Based on what I've read it is 4 years if can devote most of your time to it. You can't immediately ramp up pilot numbers no matter how much you increase salaries.
> They make respectable money on major carriers, though, and a lot of the regional pilots are probably trying to advance to those positions.
Only after spending 10-15 years gaining seniority.
> Also, how many years of lead time and training do you think is necessary before piloting a commercial aircraft?
About 1500 hours, and close to $100k (I have a private pilot, IFR, and multi engine certs). Only within the last few years did Congress increase the mandatory number of hours from a few hundred.
I know because I flew a few hundred hours and spoke to too many pilots before deciding to stay in tech and continue making a tech wage, instead of having to wait another decade to get back to that pay grade (with no guarantee you make it, based on macro air carrier finances; air travel slows and the public can't afford to fly? you get furloughed indefinitely with no pay).
Securing the communication links between the ground and the aircraft -- along with other signals such as GPS -- needs to be properly addressed before such a system can be "hijacker-proof".
I'm going to have to disagree with the conclusion that 'self flying aircraft may be well ahead' of self driving cars. I think automation of cars is a necessary precursor to automation of air travel.
Many people have issues flying due the feeling of not being in control; I expect we will first need the automation of cars to acclimatise people to the concept of trusting in an AI driver/pilot, before automated flights will be accepted.
Cargo flights strike me as the much clearer entry point for the technology.
> Many people have issues flying due the feeling of not being in control
You lack control whether it's a pilot or an AI in the cockpit, so rationally this shouldn't make much difference from this point of view. Of course, people are anything but rational. :)
I think it would be amusing to let an AI fly some commercial routes without telling people -- after all, I bet there's nothing in the carriage contract about having humans at the controls. I'm sure the FAA wouldn't approve, though.
> Many people have issues flying due the feeling of not being in control
If there's one consistent view Americans have expressed about flying, it's that they'll choose the cheapest flight over almost anything else. Bring a cheap self-flying route to market. People will get over their qualms.
> Cargo flights strike me as the much clearer entry point for the technology
Agree. That said, a helicopter seat from Manhattan to the Hamptons costs about $700. If you cut that to $200 or 300 the demand would be massive, pilot or not.
> they'll choose the cheapest flight over almost anything else
Almost anything else. I agree people will eventually get over their qualms; I just think this is a big qualm to get over.
Though what you've said about cheap helicopter flights has me thinking - perhaps the way in will be via the scaling up of quadcopters and drones. Once smaller drones are a familiar thing, larger passenger drones would be less of a stretch.
I'd argue it's easier psychologically to hop in a drone based on an ancestry that's never had an on-board pilot, than to hop in a plane that's had the pilot removed and replaced.
I don't think that's true, about low priced tickets. What you actually mean is that once companies collude to create a situation where it's virtually impossible to compare one flight to another, or one airline to another, by any criteria other than cost then cost is what drives customer decisions.
I'm sick to death of this meme that Americans are only interested in low prices. The fact is airlines were one of the first of many industries to use fees, line by line cost breakdown, obscure deals and a baffling purchase process to strand customers on an island of price comparison and nothing else. See also: cell phones plans.
If you knew what you were getting and could hold airlines accountable when they didn't deliver on what was purchased I promise you price would no longer be the only factor. This is not a truism about Americans you've found, it's a truism about distorted and manipulated markets lacking sufficient regulation.
> airlines were one of the first of many industries to use fees, line by line cost breakdown, obscure deals and a baffling purchase process to strand customers on an island of price comparison and nothing else
I'm one of those Americans who doesn't buy the cheapest ticket. I buy comfortable seats at convenient times with airlines I'm loyal to. Guess what I see none of? Any of those things! My bags are free, my drinks are free, I get to go to the front of the check-in line and, many times, security line. On occasions where I've been charged a change fee, a quick phone call gets it reversed. I board a 6:30AM flight tomorrow and am honestly looking forward to a few hours of cool views, relaxing reading and maybe a touch of alcohol.
Airlines tried competing on pitch and leg room. Their behinds got handed to them by travellers switching to budget airlines. If you want better service, there are business class tickets available at prices cheaper than pre-deregulation fares.
Note that this effect isn't constrained to Americans. European non-business passengers' demand for air travel is also extremely elastic with respect to price [1].
If I was a company looking to break into the automated passenger/cargo carrying drone business, I would focus on ferrying people and supplies to offshore drilling sites and remote mining locations. That way you are flying in uncontrolled airspace over water or unpopulated land. You can get more regulatory clearance that way and thus prove out your system, and those industries are much more likely to promise you a recurring service contract.
Technically, it seems like aircraft are closer than cars. Probably in the vehicles themselves and definitely with respect to external control systems.
In terms of public acceptance, I wouldn't be surprised if you have the psychology of it right, especially when you consider that for commercial flights you have to get buy-in from nearly all of your potential passengers or you will lose them to competitors. Maybe there will be a period of automated and non-automated flights with the latter more expensive.
For other uses like cargo and business jets, adoption could be much faster.
Interesting they see the pattern as going -> one pilot -> full automation. I would prefer to go from 2 pilots to no pilots if it happened. I don't like a single person being in complete control over the aircraft due to mental health etc. That said, I like full automation even less.
I don't think commercial jets are an especially favourable target for complete automation, compared to cars.
- Slow iteration, each mistake that you could learn from potentially costs a hundred lives
- No graceful shutdown possible most of the time – if automation fails in a car, you just stop or pull over, you probably won't die. A plane can't just land itself if automation fails unless the comm link works
- Comm link might not always work, it can be damaged from light mid air collisions, bird strikes, drone strikes, other damage, improper maintenance, possibly bad weather, hacking, jamming, over-saturation due to mass automation failures etc.
To me it seems that it's easier to deal with road chaos than acquire an extra couple 9s of uptime in this problem.
"Slow iteration, each mistake that you could learn from potentially costs a hundred lives"
The potential for disaster is absolutely there, but aircraft crashes rarely do cause that many deaths, they are just the ones media loves to cover. The survival rate of aircraft crashes are at about 90%.
Every year millions of people die on the road, amounting to several thousand each day. That means that just a bump of 10% in the death rate of automobile accidents would mean hundreds of thousands more victims every year, they're just distributed across many accidents that would only be reported on during a slow news day. A bump up in aviation death rates would mean a couple thousand more.
No one has any doubt that the planes can fly themselves in 99% cases even with todays technology, but bringing this up to a higher uptime while still being fully automated will be very hard I think.
I am only about 20 hrs into my flight training, so take what I say with a grain of salt.
I still have yet to hear anyone offer up an explanation how AI will integrate with our current air traffic control system and other aircraft. Right now if I want to fly through certain kinds of airspace (basically any airspace around a medium sized airport, civilian or military) I have to talk to air traffic control, and they tell me what do to. As a pilot in command, I have final authority to deviate from these clearances for safety reasons, mainly because it is my ass on the line.
So if a drone flight is flying an instrument flight plan from point A to point B, how does an air traffic controller command them? Part of the advantage of verbal commands being given to aircraft over a common frequency is that other pilots know what aircraft are doing around them. If a drone is being silently told to avoid for traffic, that doesnt help me if I am the aircraft being avoided. If a drone is told to hold but doesnt have enough fuel, or to divert due to weather, can it recognize a command that would create a state of emergency and over rule the air traffic controllers directions? Since air traffic controllers have need to communicate with aircraft in their airspace, how will these comms be authenticated so someone doesnt spoof ATC to make a drone do something stupid?
I know NASA is working on these problems with the FAA, but I havent heard anything yet. However, aviation isnt a "move fast and break things" area. Regulations are written in blood, and I dont really want to be the guy who gets sacrificed so that the FAA can retroactively fix an issue that someone dismissed as "psychological".
>As a pilot in command, I have final authority to deviate from these clearances for safety reasons, mainly because it is my ass on the line.
I think this is the most crucial aspect. Machines are excellent at executing patterns and adhering to constraints. Improvisation, not so much. And really, in emergency situations when things go awry and off script, it's that "clutch" factor that makes all the difference.
And that, in a nutshell, is my concern about fully autonomous cars, and the enthusiasm some express for removing all manual control because "computers are better than people".
A computer will have notified ATC and all local traffic of intent, and received acknowledgement of the distress call from all affected parties, while the human is still half way through the first "Mayday, …"
One brand's autonomous pilot will only have to be taught about handling unusual conditions once, while humans have to be trained from scratch each iteration. 20 human pilots, about 22 iterations of the same training program.
Come up with a new failure scenario and a strategy for mitigating it? Train the computer once. Train every human pilot independently. The lead time is atrocious for getting software updates out to wetware, so good luck getting every commercial human pilot to be familiar with a new procedure inside ten years, by which time you have new humans to train because some of the old ones died and taken all their training and experience with them.
As commercial aircraft incidents are well-documented, it might be possible to make some rough estimate of the change in accident rate if current airliner control systems were used as the basis of full automation. Off the top of my head, I would suggest relatively recent incidents that might well have ended in fatal crashes without pilot intervention include the Qantas flight 32 engine disintegration and fire, Qantas QF72 uncommanded pitch-down incident, and the British Airways flight 38 fuel icing non-fatal crash. On the other hand, crashes that might have been avoided include those of Asiana flight 214 at San Francisco (power mismanagement on approach) and Coglan Air flight 3507 at Buffalo (mishandling of icing and the resulting stall.)
The reason why humans are needed in the cockpit is not about handling the perfectly automatable day-to-day routine. It's about that one-in-a-million case when several things go wrong at once in a way nobody thought of before and a pilot's judgement call could save lives while an elaborately programmed AI would get totally and completely stuck.
I don't want to pay $17 (high fake number from a poster above) for a one in a million chance of saving my life. Might as well play the lottery at those odds.
Beyond the marketing there's the issue of who to blame when things go bad. These days the airline will always say "pilot error" is to blame. Once the pilot is gone that excuse obviously won't fly.
That's a problem for the insurance companies not the airlines.
The reputation damage for the airline is more or less the same.
Also note that while pilot error is the conclusion of about 60% of all aircraft accidents it's not the case for major airlines flying big passenger jets.
That statistic is inflated quite a bit by smaller chartered flights since it counts all accidents with 5 or more passengers onboard.
Also note that that statistic also shows that the majority of accidents happen on taxing, landing or take off.
You'll be surprised by how many taxing accidents there are that cause fatalities which are deemed as pilot error despite that no passenger have been killed but rather ground crew.
There are more taxi accidents than there are accidents at cruising altitude.
So yes pilot error is a big factor, it's smaller for hull loss accidents for large commercial jets but it's still the case.
I should also note that a fatal accident can also still be pilot error despite it having a different root cause like a bird strike or a mechanical problem if the pilot did not perform correctly.
That's low to you? That's actually amazingly high, unless most of their flights are really long international ones. But I suspect most flights are 2 or 3 hours.
I was musing on that too. I did some (likely flawed) calculations and got a much higher number. But I was figuring training costs, hotels, per diems, medicals. All that stuff stacks up with a pilot.
You can probably also factor in a large component of overhead just managing 12,712 people. Scheduling, logistics, HR, etc, etc.
Even if you said it's in the range of $17-30 per passenger, that's pretty significant on many routes - particular high-volume short-haul.
You can also probably fit a dozen more seats on the plane.
12712 pilots * 200K average salary w/ overhead = $2.5B
That is not a small number.
$354 average domestic fair x 143M passengers = $50.6B in total fairs.
So $2.5B pilot cost is roughly 5% of their total costs. You can not lower gas costs nor airport costs nor plane costs much, but if you could eliminate $2.5B in salary, and add more flexibility, then yeah, I'd say there is massive incentive if the marketing and safety issues can be overcome.
Pilots also need (I would guess pretty frequent) (re-)training, hotels, lounges, long term parking, time off, health insurance, etc. They're not just $200K salary, they're probably more like $500K each/year in total cost once you get to 747 level, though they may only see $200k in salary themselves.
Most pilots do not make nearly that much money... entry level is often below $30k. Maybe you'll break 6 figures at a top-tier airline after a decade of experience.
If I had to try and solve the marketing problem, I would try and focus on the security angle. "This plane can't be hijacked by a terrorist", "this plane would have stopped 9/11". Although, flying a plane remotely like a drone would present a huge hacking risk and would be hard to trust in a major cyber war. I'd prefer an entirely AI based solution with a remote piloting fallback.
A remote pilot on call for emergencies? Interesting. I'm picturing a room full of people who are on alert, as a service for near-instant, on-demand remote vehicle control; avoid car/plane crashes, land planes, sort of navigating the edge cases where human contol is still needed to bridge the gap
Before we get too excited about self-flying planes, watch the following video, which is the NTSB's animation of US Airways 1549 landing in the Hudson River after having both engines knocked out by geese:
If this happened to a self-flying plane, would someone on the ground be able to remotely take the controls in a matter of seconds and land it safely (or at least avoid crashing it into a densely populated area)?
This is a good example of why anecdotal evidence is a bad way to make decisions (also - why the entire world is making the humongous mistakes of ditching nuclear right now because of a couple examples of things going wrong; despite nuclear having saved millions of lives vs. the few it has cost).
For every example like this, there are many more you can point to where NOT having self-flying planes cost everybody on board their life. From hijackings like 9/11, to rogue pilots like MH370 or the Germanwings crash, to simply human error like the SFO Asiana crash, the SFO Air Canada almost-crash, or the Tenerife collision, or the Air France Atlantic crash, or many many more…
Planes are already pretty much fully automated. Humans make mistakes far more often than computers. Pretty much every airplane crash in the history was caused by simple human error. The more we automate, the better.
As someone who also flies >150k miles a year, I'm happy about this.
The most interesting change that self-flying planes could enable is making small planes & helicopter transportation viable for everyone.
Assuming automation progresses to the point where safety is solved (i.e. at least as safe as in a car), and that noise is also solved, you could imagine a large quadcopter style drone that will pick you up and drop you off anywhere. Once it converts to electricity, the cost of operation would be very low.
Generally speaking, I would say that the difference in safety between a commercial aircraft and a common car is essentially the quality of who is driving.
The first is driven by a professional pilot, someone who belongs to an elite, that has spent many long hours in studying and training, that has passed several (I believe not at all easy) licensing tests, subject to continuous checks (medical, drugs, alcohol, etc.) periodical exams and ability tests, continuous update and that is never or almost never left alone in the cockpit, with a colleague ready to intervene or to assist him/her if/when needed.
The second can be driven by anyone, including almost inexperienced kids, elderly people, drivers under the effect of this or that drug, of alcohol, having distractions (think about phones, pets or children in the car), the licensing and medical tests are either extremely easy to pass or very rare, like one every 5-10 years, etc.
Still, there are not as many car crashes as there could be (i.e., if you prefer, even sub-standard car drivers can often drive good enough as to not cause too many accidents).
We don't have right now the techology to fully replace a single average car driver, and we plan to replace two much higher standard aircraft pilots?
62 comments
[ 4.4 ms ] story [ 115 ms ] thread> In the meantime, it seems unlikely that we'll see mass unemployment of pilots. Mandatory retirement will see 42% of US pilots retire in the next ten years. If anything, this shortfall will push the automation case forward faster than ever.
This is a misnomer. There is no pilot shortage. Next time you fly a commuter flight, ask your pilot how much they make (it's about $20k-50k/year). There are an overwhelming number of available pilots in the training pipeline.
Just as with truck drivers, its not a labor cost problem. Its a corporate disdain for cutting into profit margins.
https://www.bloomberg.com/news/articles/2017-08-17/-smokey-a...
"The annualized driver turnover rate at large truckload fleets was 74 percent in the first quarter and the industry was short about 48,000 drivers at the end of 2015. That shortage is expected to balloon to almost 175,000 by 2024, according to the American Trucking Associations."
"Raising salaries is another option, but it would cut into profit margins. The University of Pennsylvania’s Viscelli said companies would have to double driver pay -- now about $41,000 -- to reduce turnover. About 30 percent of new drivers quit in the first three months, according to Stay Metrics."
EDIT: (crew is the lowest component of per hour flight time)
B757-200 FOC
Costs per block-hour of operations (avg. 186 seats):
CREW $ 489
FUEL $ 548
MAINTENANCE $ 590
OWNERSHIP $ 923
TOTAL FOC $ 2550 per block-hr
Ref: https://www.icao.int/MID/Documents/2017/Aviation%20Data%20an...
Also, how many years of lead time and training do you think is necessary before piloting a commercial aircraft? Based on what I've read it is 4 years if can devote most of your time to it. You can't immediately ramp up pilot numbers no matter how much you increase salaries.
Only after spending 10-15 years gaining seniority.
> Also, how many years of lead time and training do you think is necessary before piloting a commercial aircraft?
About 1500 hours, and close to $100k (I have a private pilot, IFR, and multi engine certs). Only within the last few years did Congress increase the mandatory number of hours from a few hundred.
I know because I flew a few hundred hours and spoke to too many pilots before deciding to stay in tech and continue making a tech wage, instead of having to wait another decade to get back to that pay grade (with no guarantee you make it, based on macro air carrier finances; air travel slows and the public can't afford to fly? you get furloughed indefinitely with no pay).
Many people have issues flying due the feeling of not being in control; I expect we will first need the automation of cars to acclimatise people to the concept of trusting in an AI driver/pilot, before automated flights will be accepted.
Cargo flights strike me as the much clearer entry point for the technology.
You lack control whether it's a pilot or an AI in the cockpit, so rationally this shouldn't make much difference from this point of view. Of course, people are anything but rational. :)
I think it would be amusing to let an AI fly some commercial routes without telling people -- after all, I bet there's nothing in the carriage contract about having humans at the controls. I'm sure the FAA wouldn't approve, though.
If there's one consistent view Americans have expressed about flying, it's that they'll choose the cheapest flight over almost anything else. Bring a cheap self-flying route to market. People will get over their qualms.
> Cargo flights strike me as the much clearer entry point for the technology
Agree. That said, a helicopter seat from Manhattan to the Hamptons costs about $700. If you cut that to $200 or 300 the demand would be massive, pilot or not.
Almost anything else. I agree people will eventually get over their qualms; I just think this is a big qualm to get over.
Though what you've said about cheap helicopter flights has me thinking - perhaps the way in will be via the scaling up of quadcopters and drones. Once smaller drones are a familiar thing, larger passenger drones would be less of a stretch.
I'd argue it's easier psychologically to hop in a drone based on an ancestry that's never had an on-board pilot, than to hop in a plane that's had the pilot removed and replaced.
I'm sick to death of this meme that Americans are only interested in low prices. The fact is airlines were one of the first of many industries to use fees, line by line cost breakdown, obscure deals and a baffling purchase process to strand customers on an island of price comparison and nothing else. See also: cell phones plans.
If you knew what you were getting and could hold airlines accountable when they didn't deliver on what was purchased I promise you price would no longer be the only factor. This is not a truism about Americans you've found, it's a truism about distorted and manipulated markets lacking sufficient regulation.
I'm one of those Americans who doesn't buy the cheapest ticket. I buy comfortable seats at convenient times with airlines I'm loyal to. Guess what I see none of? Any of those things! My bags are free, my drinks are free, I get to go to the front of the check-in line and, many times, security line. On occasions where I've been charged a change fee, a quick phone call gets it reversed. I board a 6:30AM flight tomorrow and am honestly looking forward to a few hours of cool views, relaxing reading and maybe a touch of alcohol.
Airlines tried competing on pitch and leg room. Their behinds got handed to them by travellers switching to budget airlines. If you want better service, there are business class tickets available at prices cheaper than pre-deregulation fares.
Note that this effect isn't constrained to Americans. European non-business passengers' demand for air travel is also extremely elastic with respect to price [1].
[1] http://dare.ubvu.vu.nl/bitstream/handle/1871/48303/140307.pd...
Most of the flight is spent on Autopilot these days so you would not know if there's someone at the commands or not anyway.
In terms of public acceptance, I wouldn't be surprised if you have the psychology of it right, especially when you consider that for commercial flights you have to get buy-in from nearly all of your potential passengers or you will lose them to competitors. Maybe there will be a period of automated and non-automated flights with the latter more expensive.
For other uses like cargo and business jets, adoption could be much faster.
- Slow iteration, each mistake that you could learn from potentially costs a hundred lives
- No graceful shutdown possible most of the time – if automation fails in a car, you just stop or pull over, you probably won't die. A plane can't just land itself if automation fails unless the comm link works
- Comm link might not always work, it can be damaged from light mid air collisions, bird strikes, drone strikes, other damage, improper maintenance, possibly bad weather, hacking, jamming, over-saturation due to mass automation failures etc.
To me it seems that it's easier to deal with road chaos than acquire an extra couple 9s of uptime in this problem.
The potential for disaster is absolutely there, but aircraft crashes rarely do cause that many deaths, they are just the ones media loves to cover. The survival rate of aircraft crashes are at about 90%.
Every year millions of people die on the road, amounting to several thousand each day. That means that just a bump of 10% in the death rate of automobile accidents would mean hundreds of thousands more victims every year, they're just distributed across many accidents that would only be reported on during a slow news day. A bump up in aviation death rates would mean a couple thousand more.
http://abcnews.go.com/International/odds-surviving-plane-cra... http://www.who.int/gho/road_safety/mortality/traffic_deaths_...
Meanwhile flights have been almost completely automated for years already
I still have yet to hear anyone offer up an explanation how AI will integrate with our current air traffic control system and other aircraft. Right now if I want to fly through certain kinds of airspace (basically any airspace around a medium sized airport, civilian or military) I have to talk to air traffic control, and they tell me what do to. As a pilot in command, I have final authority to deviate from these clearances for safety reasons, mainly because it is my ass on the line.
So if a drone flight is flying an instrument flight plan from point A to point B, how does an air traffic controller command them? Part of the advantage of verbal commands being given to aircraft over a common frequency is that other pilots know what aircraft are doing around them. If a drone is being silently told to avoid for traffic, that doesnt help me if I am the aircraft being avoided. If a drone is told to hold but doesnt have enough fuel, or to divert due to weather, can it recognize a command that would create a state of emergency and over rule the air traffic controllers directions? Since air traffic controllers have need to communicate with aircraft in their airspace, how will these comms be authenticated so someone doesnt spoof ATC to make a drone do something stupid?
I know NASA is working on these problems with the FAA, but I havent heard anything yet. However, aviation isnt a "move fast and break things" area. Regulations are written in blood, and I dont really want to be the guy who gets sacrificed so that the FAA can retroactively fix an issue that someone dismissed as "psychological".
I think this is the most crucial aspect. Machines are excellent at executing patterns and adhering to constraints. Improvisation, not so much. And really, in emergency situations when things go awry and off script, it's that "clutch" factor that makes all the difference.
One brand's autonomous pilot will only have to be taught about handling unusual conditions once, while humans have to be trained from scratch each iteration. 20 human pilots, about 22 iterations of the same training program.
Come up with a new failure scenario and a strategy for mitigating it? Train the computer once. Train every human pilot independently. The lead time is atrocious for getting software updates out to wetware, so good luck getting every commercial human pilot to be familiar with a new procedure inside ten years, by which time you have new humans to train because some of the old ones died and taken all their training and experience with them.
All the emergency features in modern world are either mechanical or electronic
The reputation damage for the airline is more or less the same.
Also note that while pilot error is the conclusion of about 60% of all aircraft accidents it's not the case for major airlines flying big passenger jets.
That statistic is inflated quite a bit by smaller chartered flights since it counts all accidents with 5 or more passengers onboard.
Also note that that statistic also shows that the majority of accidents happen on taxing, landing or take off. You'll be surprised by how many taxing accidents there are that cause fatalities which are deemed as pilot error despite that no passenger have been killed but rather ground crew.
There are more taxi accidents than there are accidents at cruising altitude.
So yes pilot error is a big factor, it's smaller for hull loss accidents for large commercial jets but it's still the case.
I should also note that a fatal accident can also still be pilot error despite it having a different root cause like a bird strike or a mechanical problem if the pilot did not perform correctly.
United Airlines employs 12712 pilots. [http://www.airlinepilotcentral.com/airlines/legacy/united_ai...]
United has 143 million passengers. [http://newsroom.united.com/corporate-fact-sheet]
If we say that United has a cost per pilot of $200,000 (likely high) then the cost per pilot per passenger is $17.77
You can probably also factor in a large component of overhead just managing 12,712 people. Scheduling, logistics, HR, etc, etc.
Even if you said it's in the range of $17-30 per passenger, that's pretty significant on many routes - particular high-volume short-haul.
You can also probably fit a dozen more seats on the plane.
That is not a small number.
$354 average domestic fair x 143M passengers = $50.6B in total fairs.
So $2.5B pilot cost is roughly 5% of their total costs. You can not lower gas costs nor airport costs nor plane costs much, but if you could eliminate $2.5B in salary, and add more flexibility, then yeah, I'd say there is massive incentive if the marketing and safety issues can be overcome.
https://www.youtube.com/watch?v=5S5hRRio-E8
If this happened to a self-flying plane, would someone on the ground be able to remotely take the controls in a matter of seconds and land it safely (or at least avoid crashing it into a densely populated area)?
For every example like this, there are many more you can point to where NOT having self-flying planes cost everybody on board their life. From hijackings like 9/11, to rogue pilots like MH370 or the Germanwings crash, to simply human error like the SFO Asiana crash, the SFO Air Canada almost-crash, or the Tenerife collision, or the Air France Atlantic crash, or many many more…
As someone who also flies >150k miles a year, I'm happy about this.
Assuming automation progresses to the point where safety is solved (i.e. at least as safe as in a car), and that noise is also solved, you could imagine a large quadcopter style drone that will pick you up and drop you off anywhere. Once it converts to electricity, the cost of operation would be very low.
The first is driven by a professional pilot, someone who belongs to an elite, that has spent many long hours in studying and training, that has passed several (I believe not at all easy) licensing tests, subject to continuous checks (medical, drugs, alcohol, etc.) periodical exams and ability tests, continuous update and that is never or almost never left alone in the cockpit, with a colleague ready to intervene or to assist him/her if/when needed.
The second can be driven by anyone, including almost inexperienced kids, elderly people, drivers under the effect of this or that drug, of alcohol, having distractions (think about phones, pets or children in the car), the licensing and medical tests are either extremely easy to pass or very rare, like one every 5-10 years, etc.
Still, there are not as many car crashes as there could be (i.e., if you prefer, even sub-standard car drivers can often drive good enough as to not cause too many accidents).
We don't have right now the techology to fully replace a single average car driver, and we plan to replace two much higher standard aircraft pilots?
Let's hope no one learns to hack that system.