American Airlines is facing higher wage costs, and laying off customer service staff. Fast food restaurants move to kiosk ordering and automated kitchens. UPS agreed to unprecedented driver pay, while limiting the number of employees making that pay and now pursuing layoffs.
The Air Line Pilots Association pursued barriers to entry to becoming a pilot, getting Congress to impose rules that made it more expensive and time-consuming. This limited the number of commercial pilots, and made it possible to earn extraordinary wages. It’s now possible for a pilot, without college degree, to earn half a million dollars.
SWAPA Pilots supporting @WestJet @ALPAPilots as they hold informational picketing in Toronto, Canada. pic.twitter.com/WgtGeRy9EL
— Southwest Airlines Pilots Association (@swapapilots) May 9, 2023
They’ve created the economic conditions where developing technology to replace them makes sense. And that technology is going to be safer than having a human in the cockpit before it’s deployed.
Single pilot cargo operations are coming, and the major pilot union is fighting this. In an event this week that made my blood boil when I saw it promoted through social media, but managed to ignore until Enilria highlighted it, the President of ALPA called reduced pilot operations “the single greatest existential threat to the piloting profession.” And, as Enilria points out, he added ‘oh yeah and for safety too.’
“I believe that this issue is the single greatest threat to aviation safety. This is about removing pilots—all of us—from the flight deck over time,” said Ambrosi during his opening remarks. “Reduced-crew operations is a tipping point—for our profession and the safety of our skies. And it’s not happening a decade from now, it’s happening right now. The scale of this threat should make everyone in this room sit up and take notice—and want to fight back, which is exactly what ALPA is doing.”
Two important principles:
- Safety rules shouldn’t be the same for all flights. ALPA pushes ‘one level of safety’ but they’re really pushing ‘the same rules for all situations’ which isn’t how you promote safety. Pilot fatigue is a major issue, and it’s not just about the number of hours a pilot works (or whether they’re actually resting on layovers). Time zones can be a big challenge, too. You need different experience and duty limits flying a widebody jet from the U.S. to Asia than to do takeoffs and landings on one hour flights in a 30 seat regional jet. Rest needs are different and risks are different.
- When an AI co-pilot is safer than a human, we should move to AI. This will be demonstrable. The standard shouldn’t be “two pilots in the cockpit” it should be “safety.” When two humans is less safe than one human and a machine, we should opt for the safer alternative.
Last January an American Airlines Boeing 777 headed from New York JFK to London Heathrow had a near-collision with a Delta Air Lines Boeing 737 headed to Santo Domingo. It crossed in front of the Delta jet that was in its takeoff roll. Air traffic control saw this happening and at almost the last possible moment called out to Delta to abort takeoff (“Shit! Delta 1943 cancel takeoff clearance!”).
The Delta flight stopped less than 1000 feet from where it would have intersected with American’s plane. Runway 4L was being used for takeoffs. The American Airlines aircraft did not follow air traffic control instructions. ATC audio shows they were told to “”cross runway 31 Left at Kilo” and instead crossed runway 4 Left at Juliet, in front of the accelerating Delta Boeing 737.
The pilot union defended pilots refusing to speak to NTSB investigators. A report on the incident found that “the captain became distracted and confused about takeoff instructions and the co-pilot lost track of their plane’s location.”
The disoriented pilots chose to still fly to London, erasing cockpit voice recordings and hindering the investigation. Maybe an AI co-pilot would have been better? Or at least DEI-hired pilots?
There will be a time when AI is safer than a human co-pilot. It’s not clear whether that’s in 2025, 2030, or later but achieving this is inevitable. Airbus has made great strides with autonomous flight. Their Autonomous Taxi, Take-off, and Landing (ATTOL) program demonstrated full cockpit automation with an Airbus A350. Their Dragonfly project aims to extend this under emergency conditions. Vertical takeoff and landing developers plan for operations without any pilots, in some cases from the start and for others as a longer-range plan.
There’s nuance to the language that pilot unions use around the number of pilots in the cockpit. They don’t say there needs to be two pilots. They say at least two. There’s still institutional ennui over the reduction in cockpit crew that brought us down to three, and they can’t fully give up the idea that there should be more. At one time you might have a captain, first officer, flight engineer, navigator and radio operator on a flight, though cockpits were downsized to three by the early 1960s. Eastern Airlines flight engineers went on strike in response to new aircraft which no longer supported a need for their employment.
Pilot unions will fight against continued loss in jobs, and very few would push the button today to go from two to one human in the cockpit. But the technology will arrive that will make travel much safer replacing a human at least as co-pilot. And at that point pilot unions will be clearly lobbying against safety.
Gary PLEASE STOP POSTING THIS – you’re embarrassing yourself.
NUMEROUS PILOTS have come on your site to tell you you are wrong, you are out of your element, you are opining on matters on which you lack training and experience.
WHY DO YOU CONTINUE TO SPEW NONSENSE? I’ll let the pilots, like SMR and 2808 Heavy, come back here and tell you and everyone why you’re wrong to say that pilots just want to preserve their jobs.
I’m not a pilot or aerospace engineer, but I am an expert in AI — not just language models that are all the hype, but all the mathematical, statistical, computational, and social scientific theory, down to brass tacks. You’ve described yourself as a master debater, Gary. What do you know about consciousness, from its cognitive psychological and philosophical roots? What do you know about Markov chains (the extremely oversimplified building blocks of today’s popular generative AI models)? Do you even know anything other than pretending to be an economist because you once “Coasean bargianed” your way into paying a kid 5 bucks not to recline in economy class?
Your spam filter will catch my comment if it’s too long so let me say this
NO, THE COPILOT OF YOUR NEXT FLIGHT WILL NOT BE AN AI because human-level general intelligence is necessary to pilot a plane, and AGI (artificial general intelligence) is impossible. It’s not a matter of getting more GPUs or TPUs out on the market. It’s not a matter of training on larger datasets. It’s a flat out impossibility.
Please just focus this blog on whether we should go to the Cathay or the Qantas lounge at any given airport, and know the limitations of your own expertise.
Well, this would solve the problem of them taking the upgrade space!
I insist on normal human intelligence and experience . Comes in handy when solving unforeseen problems . Pilots are completely correct . EVs have learned that the hard way with their fool self-driving vehicles .
@Dignity … +1 .
What I worry about is the situations that have never happened before and how would AI handle it. Imagine Cactus 1549 in the hands of AI. An airliner has never successfully ditched and everyone survive. Yet there is nowhere to land in NYC. If AI learns from historical incidents, and is only as good as the programmer who wrote it…what would AI want to do?
There are many instances in commercial aviation that would perplex AI…BA 747 that flies through volcanic ash and all engines quit (not erupting at departure time and flying at night, nobody knew it had erupted).
Air France A330 can’t stall until its pitot tubes froze up. Computer basically can’t figure out what was wrong and basically quit, and went into a backup flying mode because the inputs did not compute. Did the pilots screw up..,yes, but if we’re talking about AI being safer, the fact that the A330 could not figure out what was going on and basically quit, scares me on if you think AI is the end all be all in aviation safety.
@Dignity – We’ve had autoland for years. Airbus is already close. And pilots come here, engage in motivated reasoning and complain and insult but never offer actual reasoned analysis.
Click bait
@Safety Guy – an AI would process non-normal checklists faster than a human, it would calculate all possible landing scenarios from diversion points and ability to make each, faster than a human … buying more time than the pilots of 1549 had. Why would you think an AI can only replicate past incidents?
I want to move to AI pilots full stop now. I think it’s ridiculous that an aircraft doesn’t entirely fly itself or that there are overpaid pilots just kicking back and collecting their union dues. Get a grip. This is the best opportunity for cost cutting I’ve seen in a long time.
Talk about manufactured controversy. One pilot will *NEVER* be an option.
That’s because the question isn’t whether one pilot + AI is safer than two pilots. The question is whether one pilot + AI is safer than two pilots + AI.
Right now two pilots plus current automation/AI is as safe as we can get (or at least, a 3rd pilot isn’t worth the weight… and aviation has never had more than two PILOTS.) With additional AI, we’ll get more safety. If AI improves to the point that one pilot+AI is safer than we CURRENTLY are with two pilots, that doesn’t mean we get rid of the 2nd pilot and regress safety back to where we are now.
Only if we get to the point where the 2nd pilot doesn’t add to safety should we consider one pilot, and that will happen right after pilots don’t need to eat or pee and won’t have the fish, or AI is so good that you trust the plane to fly itself with zero pilots regardless of conditions or failures.
So if you don’t trust a flight with ZERO pilots, then you need two, because no human can operate 100% of the time.
Also, “less than 1000 feet” is a bit of sensationalism. That’s 3 football fields, or FIVE runway widths. Planes on the tarmac are within 1000 feet of each other ALL THE TIME. Parallel taxiways are only about 300′ from runways. The stop line on INTERSECTING taxiways at JFK is only ~200′ from the runway (maybe less).
@SafetyGuy … +1 , especially for your rational common sense . Those were real-life examples you provided . Thank you .
@Gary. Why do I think AI can only replicate past incidents? Because AI is only as smart as the programmer. I believe (maybe wrong) that AI would have thought trying to land on the Hudson would not be an option because history had shown that to be a terrible option.
1549 did not follow the checklist…there was no checklist of dual engine failure at low altitude (I’m sure you watched the movie). They turned on the APU to keep all instrumentation on (violated the existing checklist). Would AI violate checklist procedures? Or do things out of order because the pilot’s experience and judgement told him/her that doing something different would result in the best outcome?
Another hypothetical scenario…person needs medical assistance ASAP. Closest major city has a thunderstorm over the airport. There is another town nearby that has a regional hospital but the runway is too short (according to manufacturer limits). The closest airport that has a hospital that is the safest option is 45 min away…what does AI do? Does AI land in a thunderstorm? Does AI risk going off the runway? Or does AI do the most legal option and drive 45 min away…patient is going into cardiac arrest.
In a way MCAS was AI…it would override the pilots if it sensed a stall because it “knew better” yet the aircraft were not in a stall.
Bottom line is AI not its programmers can program every possible situation. What are the constraints of system logic?
Could AI be beneficial in some areas…absolutely. I see cars with heads up displays that overlay driving directions ontop of 3D models of the city streets. That would do numbers to reduce runway incursions.
What could possibly go wrong? Even though AI security thinks I’m a bot and won’t let me access certain sites. Not to mention other technology failures and complex conditions that can occur.
There is no substitute for the human mind, never will be.
I do see it happening at some point because pilots are so expensive and only going to get more expensive…
Heck with the unions, I’ll fight to keep this from happening!! I will NEVER EVER board a flight or put my family on a flight under the control of AI. To think that experiential intelligence can be programmed is a pipe dream – just far too many scenarios of things that could go wrong that it’d take decades to code and would still be far from perfect or what many would consider “safe enough” to risk it. Even though airlines would probably love this, they’ve also recognized the huge liability associated with it – one incident would upend years of research and billions invested- likely sinking the entire industry for a significant period of time similar to 9/11. Public perception matters and it is very fragile.
Another topic of concern, our own FBI came out this week and raised the red flag concerning Chinese hackers potentially disrupting our infrastructure and somehow we’re supposed to believe that airplanes flown by AI will be impenetrable to something similar? They’ll certainly have to be connected to a ground station as a backup “just in case” so the risk of being hacked will most definitely be present. Who’s willing to gamble with that? Certainly not me or my family.
Some ideas may sound good in theory, but that’s where it abruptly ends. This is one of those ideas.
I think Gary has taken his loathing for pilot’s cries of “safety” when they really mean job/wage protection (there is, indeed, simply no reason to require X hours of flying to be a commercial pilot, as hundreds of hours coasting through good weather don’t teach anything, it just forces people to spend more money to be pilots) to ANY issue where safety is sighted.
Sometimes it really IS about safety. And a commercial aircraft with one pilot will never be safe.
I write clearly in this post that today’s AI isn’t something to push that button on.
But it is going to get there in some number of years. And when it does, two pilot cockpits in lieu of an AI will be standing in the way of safety.
Cited*… we really need an edit feature, like every virtually other comment section on the internet.
Yes, at some point in the future, AI+zero pilots will be better than any pilots. But we’ll never have one pilot+AI.
Will that be before or after 2 pilots on the ground are as good as 2 pilots in the cockpit?
Before or after we have transporters or some other mode of transportation that makes air travel seem silly?
@Gary … Come on , now . You must be pulling our leg . EV self-driving vehicles show the limits of AI ; and any aircraft programmed by a mouse-holding tech wonk would be “dumb” , not “intelligent”.
Maybe AI in the future could stop incidents like the Germanwings crash. Or even if the mushroom muncher on Alaska Airlines had happened to be piloting a flight when he decided to try to crash it.
Two pilots vs one pilot + AI = False dichotomy.
If safety is truly the primary concern then why not two pilots + AI?
Didn’t a ‘confused’ AI fight pilots attempting to save 2 Boeing 737 MAXs? The dead 737MAX passengers certainly are not cheering AI in the cockpit.
@AlohaDaveKennedy – No
Gary is a pretty bad person posting this stuff. Spreading very false information…its nowhere near going to happen. If it does someday it should be long after technology is developed. How is recent Aviation tech doing? Engines blowing up….doors falling off..electric trim crashing airplanes. Let’s remove pilots….
Gary is a flyer who absolutely HATES pilots.
I
Remember folks…
2008…16 years..That is the LAST time anyone died on a Part 121 Commercial Jet here in the USA due to pilot error, Gary does not like the rules that changed after that crash in 2008..that worked!
Gary and others also do not realize something. Although data shows many accidents have happened due to human error (and I agree)…Where is the data showing how many lives have been saved by human skill? God help us all as we try to find AI to endanger our lives and take our jobs… then we have numb nuts like this guy posting this crap
Also..2 pilots in the cockpit instead of AI will be standing in the way of safety…What data do you have support that?
1995…29 years….this is where the technology of our ATC system lies…29 years in the past. I still have to wait on the ground for take off for 20-30 miles of in-trail spacing when it is not that busy. (then the next flight will be delayed and the airline will be blamed)
Gary…how do you figure..2025..2030 or even anywhere i the near future we will have the technology to support this full autonomous system when we cannot get into the 21st century
These AIs gonna be CPR capable?
Do Heimlich’s?
@SMR … +1 . Human Skill is a product of Only Humans . The tech mouse-holders designing tech couldn’t are so impractical they couldn’t help change a flat tire .
@Gary AI is great…Tesla autopilot has it. How’s that going? All those accidents it is preventing from driver error. Except they are all recalled for wait for it….too many accidents caused by it. That’s only responsible for maneuvering in 2 dimensions
@SafetyGuy … +1 … and the EV autos don’t even need to deal with Gravity and Drag , and still cannot hack it . Gress what … Our leg is being pulled with this nonsense .
Ooh, you really stirred the hornet’s nest this time, Gary, lol.
AI is still in its infancy, so anyone claiming that AI will *never* be to fly an airliner safely is either ignorant or not being honest.
There is no such thing as 100.0% safety in any endeavor, whether it’s self-driving cars or a human driving a car. The same holds (will be) true for airplanes. The Pentagon already has many contracts out for autonomous vehicles that fly in the air, drive on land, move on the water and under the water. AI is already able to defeat the best fighter pilots in simulated aerial dogfights, and companies like Tesla are getting fairly close to having self-driving cars. But the Pentagon’s mission and standards of safety are obviously much different than that of the FAA and EASA.
AI will almost certainly cut its teeth on autonomous drone deliveries and then freight operations. Only after these are both established will the idea of AI replacing human pilots gain more traction. A possible future scenario might involve AI with full capability to fly the airliner combined with one person as a backup pilot in the cockpit plus a remote human pilot as an additional backup on the ground (much like USAF pilots flew Predator drones over Iraq and Afghanistan while sitting on the ground in Nevada).
One thing is certain, AI will look increasingly attractive as the capabilities of AI increase more and more and their cost keeps dropping, while at the same time the costs/wages of human pilots continue to climb.
I see a future for a remote copilot before AI. That technology is already here as remotely flying drones has proven. I wonder if several crashes would have not happened if a remote pilot had been able to figure out the solution that the onboard pilots were not able to figure out.
@jns… what crashes are those?
Auto-flying is an order of magnitude easier to program than auto-drive as there are an order of magnitude less number of parameters. It’s not AI, per se, as there’s no need for self-learning / deep learning, just straight programming. It’s a highly controlled environment. There will undoubtedly be outlier scenarios where a creative and quick thinking individual human pilot would make a better judgment call. However, auto-fly systems will become available in the very near future that will outperform and be safer than the average human pilot. And when it’s available, it would be a shame not to utilize it.
@Jake-1.. Looking forward to the future where none of us have jobs. It is great that the media is helping to spread this and help give false senses of security to the general public.
@SMR: Nobody wants jobs. People want stuff. It just so happens jobs are the main vehicle for people getting stuff.
The problem if everyone loses their jobs isn’t that no one has jobs. That’s fine. The problem is that the people who own the robots/ai will keep all the stuff and let everyone else starve.
What about AI for legitimate journalism? Maybe more truths could be told rather than fantasy statements. FAA has said for years that will not certify any FAR 121 jet single seat until there is technology that has been proven that can land a jet remotely. Additionally, they need protections that software can not be hacked, if you look at the multiple software breach’s just recently that tells us we are not even close. This article is nothing but a rouse.
Hey Gary, Just wondering if you have a prediction on when AI is going to replace you as a blogger? At least pilots have a useful skill.
They could always repurpose the pilots to serve PDBs since the flight attendants aren’t interested in doing it.
@June – It potentially replaces some content, like news, because (at least OpenAI’s GPT-4) has trained on this blog. It wouldn’t replace first hand reporting or insider leaks, however.
The regulatory agencies (and not just the FAA) take many, many years to even initially consider these sorts of things and anyone involved in aviation know how many decades we are away from this even reaching prototype stage. Zero chance they change their protocol because some airline exec wants a bigger bonus. The military has had autonomous aircraft for years and have spent hundreds of billions developing them. There’s a reason why its useage is still not widespread….
If anyone on here thinks any less than two pilots that are rested, and properly trained is a safer than any other automated option, is drinking their own bathwater. And you probably shouldn’t be part of this conversation in the future.
And let me add experienced as well. Not people that have no experience on airplanes that are high-performance, that demand some level of skill to fly and understand what’s going on in that environment.
I will never get on an airplane controlled solely by AI and computer automation. I am an engineer who works in automation and the things I have seen AI screw up are mind boggling, things that a human could have delt with instantly. Ever try to take an automated taxi from a place with a large parking lot and lots of humans and cars go every which direction? Waymo is a joke in that situation. Tesla fsd doesn’t work and it is proven and witnessed in many videos. What airlines have figured is that they will save enough money that a few ai incidents won’t deter the public while they rake in record profits feom not having to pay humans. I have a feeling that the majority of humans feel the same. I am actively afraid of the idea.
Looks like the bus driving gravy train is over.
Wow, a lot of luddites in the comments section. Yes Gary, pilotless airplanes will be the wave of the future. Not just yet but it will happen. Single pilot operations will become the norm only when no pilots are necessary. It’s closer than most people think. One pilot will most likely be on board to give some kind of assurance to the pax but he or she will not be necessary for the conduct of the flight.