Are pilots going to be replaced by AI?

[hais]

10 hours ago, 1980Mooney said:

Headline: Pilot Unions Form Coalition to Oppose Single-Pilot Operations, Citing Safety Concerns and Profit-Driven Motives

The only reason that 3 competing pilot unions would form a coalition to try to stop single pilot operations is because they know it is a imminent real threat to their employment.

Notice that it is 3 international pilot unions (not the APLA) banding together - no wonder the concept that airlines have "profit-driven motives" is foreign and confusing to them.

I am sure their strategy will be brilliant - go on strike while being made redundant.

The same fight is going on with rail - in North America and Europe.  The industry is pushing for single man rail crews in the cab.  The UK rail unions continue to strike for 2- man crews.  However apparently the European unions have already given up on freight trains and are operating those with single man crews.

Maybe the logical progression in aviation will be that freight flights, FedEx, UPS, etc will be the first to go to single pilot ops. 

Perhaps regionals and commuter will benefit the most. Reading this report, I suspect no one would complain much if the single pilot is PNF :

https://asrs.arc.nasa.gov/publications/directline/dl5_one.htm

[1980Mooney]

1 hour ago, Andy95W said:

Could one pilot fly the advanced airliners of today?  Of course.  But that does raise a few questions:

1.) how do you prevent a suicide/murder like GermanWings 9525?

2.) how do you prevent the single pilot from making a mistake such as a runway incursion or flying into a mountain at night, or even just falling asleep?

3.) if the answer is to have a means of control from the ground, how do you prevent a hostile actor (terrorist?) from intentionally crashing the airplane from afar?

I think the basic problem with question #1 and #2 is that I believe you are assuming that the pilot will actually be flying the plane.  A more sophisticated "AI" autopilot will be flying the plane in the future.  The "pilot" will be monitoring.  I hope you don't think it is actually pilots today that keep planes from flying into mountains at night these days - the route is set in the FMS.  Same goes with a pilot falling asleep.

As far as runway incursions go, one set of eyes should be adequate.  

On the GermanWIngs 9525 disaster, there were two (2) pilots on board - That didn't stop a determined pilot.  Are you suggesting that planes need to go to 3 pilots so that one single pilot is never alone?

Lastly what today keeps a hostile actor from hacking our spy satellites and drones and redirecting or crashing them? Encryption and security.

[Will.iam]

50 minutes ago, 1980Mooney said:

I think the basic problem with question #1 and #2 is that I believe you are assuming that the pilot will actually be flying the plane.  A more sophisticated "AI" autopilot will be flying the plane in the future.  The "pilot" will be monitoring.  I hope you don't think it is actually pilots today that keep planes from flying into mountains at night these days - the route is set in the FMS.  Same goes with a pilot falling asleep.

As far as runway incursions go, one set of eyes should be adequate.  

On the GermanWIngs 9525 disaster, there were two (2) pilots on board - That didn't stop a determined pilot.  Are you suggesting that planes need to go to 3 pilots so that one single pilot is never alone?

Lastly what today keeps a hostile actor from hacking our spy satellites and drones and redirecting or crashing them? Encryption and security.

If the pilot monitoring can not override the AI flying, then why have the pilot monitoring? What good is a pilot monitoring when he can not stop the AI. Just like the airbus tree eater or boeing’s mcas. There is only one captain. And that person or AI has final say as to the final outcome. When there is a disagreement and one is correct and the other is wrong but both think they are correct who gets to choose? The other pilot can try to influence the pilot flying or fight / force him to a different action but those things will not work for an AI if it had the final say. The AI is going to do what it was programmed to do. You either let the AI be the captain and thus you don’t need a pilot monitoring as they can’t take over anyways or you let the pilot be able to take over and thus lose the protection of the AI when humans make mistakes and who knows if  it was a mistake to override the AI in this situation?

[1980Mooney]

39 minutes ago, Will.iam said:

If the pilot monitoring can not override the AI flying, then why have the pilot monitoring? What good is a pilot monitoring when he can not stop the AI. Just like the airbus tree eater or boeing’s mcas. There is only one captain. And that person or AI has final say as to the final outcome. When there is a disagreement and one is correct and the other is wrong but both think they are correct who gets to choose? The other pilot can try to influence the pilot flying or fight / force him to a different action but those things will not work for an AI if it had the final say. The AI is going to do what it was programmed to do. You either let the AI be the captain and thus you don’t need a pilot monitoring as they can’t take over anyways or you let the pilot be able to take over and thus lose the protection of the AI when humans make mistakes and who knows if  it was a mistake to override the AI in this situation?

I never meant to imply that the pilot monitoring could not take over. Who gets to choose with 2 pilots up front?  The PF at the controls is flying.  The other pilot is monitoring.  Fight /force between 2 pilots? - that is a sure way to crash.  That CRM went out years ago.  AF447 - the 2 pilots were both acting PF fighting each other on the controls with opposite inputs. One has to hand off to the other.  If the AI Autopilot goes down the pilot takes over.  How does 2 pilots help that situation?

[Will.iam]

18 minutes ago, 1980Mooney said:

I never meant to imply that the pilot monitoring could not take over. Who gets to choose with 2 pilots up front?  The PF at the controls is flying.  The other pilot is monitoring.  Fight /force between 2 pilots? - that is a sure way to crash.  That CRM went out years ago.  AF447 - the 2 pilots were both acting PF fighting each other on the controls with opposite inputs. One has to hand off to the other.  If the AI Autopilot goes down the pilot takes over.  How does 2 pilots help that situation?

Well for one captain always has final authority. So ultimately one decision but pilots do get into fughting for control when lives are about to die. fedex L-10ll comes to mind or when one pilot is trying to commit suicide and the other is fighting to save the plane. AF447 was a lot of lessons learned and highlighted pilot misunderstandings. The 2 pilots were more confused than fighting. They didn’t understand or realize the computer had gone out of normal law and thus they had no computer to protect them. They assumed the computer would not let them stall. Airbus not coding the software to keep the stall warning to continue in a deep stall only confused the pilots more for when they tried lowering the nose the stall horn would come back on but pull back on the stick it would turn off. We instinctively want to do things that silence warnings and sounds but that was wrong for this situation. Bottom line their training was inadequate for the situation they got in. Thus why we now have training for these situations you can get into. Airbus originally claimed the airbus can not stall. They have had to eat those words multiple times. Just like the titanic cannot sink, we humans find a way to prove to the engineers the impossible is possible. 

[1980Mooney]

4 hours ago, Will.iam said:

Well for one captain always has final authority. So ultimately one decision but pilots do get into fughting for control when lives are about to die. fedex L-10ll comes to mind or when one pilot is trying to commit suicide and the other is fighting to save the plane. AF447 was a lot of lessons learned and highlighted pilot misunderstandings. The 2 pilots were more confused than fighting. They didn’t understand or realize the computer had gone out of normal law and thus they had no computer to protect them. They assumed the computer would not let them stall. Airbus not coding the software to keep the stall warning to continue in a deep stall only confused the pilots more for when they tried lowering the nose the stall horn would come back on but pull back on the stick it would turn off. We instinctively want to do things that silence warnings and sounds but that was wrong for this situation. Bottom line their training was inadequate for the situation they got in. Thus why we now have training for these situations you can get into. Airbus originally claimed the airbus can not stall. They have had to eat those words multiple times. Just like the titanic cannot sink, we humans find a way to prove to the engineers the impossible is possible. 

As said many times the airbus stalled because the human at the controls manually pulled 1.6 G's upward at 35,000 ft. without full power on the engines.  If the pilot would have maintained attitude and power while the pitot tube thawed and reengaged the autopilot all would be well.  It is no different than if you are at altitude in cruise and your pitot tube freezes.  Do you panic and pull the yoke back? Or do you keep it straight and level while you figure out what is going on?

[T. Peterson]

9 hours ago, 1980Mooney said:

I think the basic problem with question #1 and #2 is that I believe you are assuming that the pilot will actually be flying the plane.  A more sophisticated "AI" autopilot will be flying the plane in the future.  The "pilot" will be monitoring.  I hope you don't think it is actually pilots today that keep planes from flying into mountains at night these days - the route is set in the FMS.  Same goes with a pilot falling asleep.

As far as runway incursions go, one set of eyes should be adequate.  

On the GermanWIngs 9525 disaster, there were two (2) pilots on board - That didn't stop a determined pilot.  Are you suggesting that planes need to go to 3 pilots so that one single pilot is never alone?

Lastly what today keeps a hostile actor from hacking our spy satellites and drones and redirecting or crashing them? Encryption and security.

I don’t think Andy was addressing AI. He was addressing single pilot ops. The technology of today is wonderful thus making the single pilot idea at least something to consider. However, the wonderful technology is still managed by a pilot and Andy’s concerns are valid.

[Will.iam]

3 hours ago, 1980Mooney said:

As said many times the airbus stalled because the human at the controls manually pulled 1.6 G's upward at 35,000 ft. without full power on the engines.  If the pilot would have maintained attitude and power while the pitot tube thawed and reengaged the autopilot all would be well.  It is no different than if you are at altitude in cruise and your pitot tube freezes.  Do you panic and pull the yoke back? Or do you keep it straight and level while you figure out what is going on?

And we don't have an Airbus protections system on our aircraft so we know if we pull back on the controls we will cause a stall, there is no computer system there to save us so we do not do that.  Compare that to the training airbus pilots received especially for a terrain warning that you pull full back on the stick  or a TCAS command to climb and let the computer handle the airplane.  never in the sim that I am aware of do they teach a situation that makes you react to climb at max performance but also kick the computer out of direct law to see if you notice you no longer have stall protection to see if you will recognize you are in fact going to stall and manually avoid the stall. An analogy is having a 4X4 truck thinking you can not get stuck going mudding through a field.  The 2x4 truck driver knows he could get stuck so he doesn't even try it but the 4x4 driver still gets stuck just he gets stuck in deeper mud.  There was another crew that died in an acceptance flight they were testing all the airbus systems and the airplane was working as designed engaging each protection system as needed until they went to slow the airplane down to try to stall it and waited for the system to stop the plane from stalling.  Unfortunately because it had worked for every other test correctly they were lax when the airplane got too slow thinking it would stop the stall until it was too late and they actually stalled. Why would they be so lulled into this? I blame part of that because of the mantra an airbus can't be stalled the computer will not allow it if it's in normal law.  Well their computer was in normal law there was no indication that it wasn't.  But the problem was the AOA sensor was frozen from water that had turned to ice at altitude and so was feeding incorrect data to the computer when the airplane slowed down to stall speed the AOA wasn't reflecting that.  Had the software had a check and balance to check that airspeed and g-meter disagreed with the AOA and at least give a warning something was not right, the pilots might have been alerted in time to be able to stop the stall from happening.   

[T. Peterson]

8 hours ago, Will.iam said:

Well for one captain always has final authority. So ultimately one decision but pilots do get into fughting for control when lives are about to die. fedex L-10ll comes to mind or when one pilot is trying to commit suicide and the other is fighting to save the plane. AF447 was a lot of lessons learned and highlighted pilot misunderstandings. The 2 pilots were more confused than fighting. They didn’t understand or realize the computer had gone out of normal law and thus they had no computer to protect them. They assumed the computer would not let them stall. Airbus not coding the software to keep the stall warning to continue in a deep stall only confused the pilots more for when they tried lowering the nose the stall horn would come back on but pull back on the stick it would turn off. We instinctively want to do things that silence warnings and sounds but that was wrong for this situation. Bottom line their training was inadequate for the situation they got in. Thus why we now have training for these situations you can get into. Airbus originally claimed the airbus can not stall. They have had to eat those words multiple times. Just like the titanic cannot sink, we humans find a way to prove to the engineers the impossible is possible. 

All excellent points, Will, plus the fact that it was a failure of technology which put AF 447 in extremis. There is no doubt the pilots responded incorrectly, but is it a wonder? For years the airlines had pushed for almost total reliance on automation and this was especially true for many foreign airlines. I distinctly remember during our Airbus transition training a chief pilot walking in and saying that the new test of airmanship is how well we were able to program and manipulate the boxes. I can only imagine the culture at AF. I wonder if the AF pilots had ever flown with either the A/P, FD, or A/T off? How about all 3 off at one time? So in this culture of technological over dependence, the pilots are at 37000 feet in turbulence, IMC and unreliable airspeed and they screw it up. Now isn’t that surprising.

What AF 447 proves is the fallibility not of pilots, but of technology, and the abject stupidity of failing to train for when all the whizmo ghizmo fails. Proof? Why did all the airlines race into upset jet training after this accident?

We don’t even have the present technology perfected. I routinely have to help FeFe make crossing restrictions and yet we are postulating AI? Utterly premature.

[carusoam]

Holy cow!

I think you guys have nailed it… three ways to Sunday!  

1) We can all benefit by good automation…

2) Good automation requires decent sensors…

3) Decent sensors have the ability to let things know when they aren’t working properly anymore… and not include the erroneous information… (post it note over the bad data)

4) For Automation to replace a two person crew…. That is a technical possibility… but requires the remaining crew member to not have health issues while at work… no heart attacks allowed.

5) did anyone mention a back-up plan for a single pilot casualty?  
The plane gets landed remotely, by a pilot controlling from corporate headquarters…?

6) Speaking of normal law….  How much computer, sensor, and logic technology is covered in flight school lately?  (Probably a lot more post y2k, than before)

Looks like we can be discussing the steps required to get to the full automation that is in the future…

R.I.P.   Gordon Moore, started at Fairchild electronics, before building Intel…

Moore’s Law…. The number of transistors on a chip will double every year…. (Stated in the mid 60s…) and was still holding when I learned about it in the mid 80s… before slowing to doubling every 1.6 years or so…  286, 386, 486, Pentium…. Then multiple cores and parallel processing…

Where Intel has run into trouble lately… they haven’t been able to make the transistors smaller like AMD and NVDA have been doing…

NVDA has announced their latest standard… a 4-nano standard… if this pans out… Moore’s law gets Re-invigorated…

Making the transistors smaller…

1) Saves energy

2) more efficient to keep cool

3) operates faster, does more work…. The amount of work required for real AI to occur.

4) fits in a smaller package

Proof of this… have you seen how well the faster GPS and faster autopilots actually control things…?  Until they have difficulty with things that stick or slip….  

The AI guys are happy with this chip, as are the Crypto miners….

https://www.pcmag.com/news/nvidias-latest-gpu-boasts-a-4-nanometer-process-from-tsmc-pcie-gen-5

Chip designs made in the US and Europe are very powerful….

Chip manufacturing for these designs are often in Taiwan, an island country off the coast of China… TSMc

There are a lot of trade restrictions for these technologies to keep chips, designs, and their manufacturing equipment… from making their way to China and Russia… capitalism, communism, and war…

 

Keep teaching STEM topics to everyone that can take it…

See how many computers are running in your favorite restaurant, barbershop, car mechanic’s shop…

 

For the person that asked… where is the engine monitor on that fancy instrument panel…. (Bar graph ends at 85mph, but indicated speed goes well past 100)

The engine monitor is behind the instrument panel with a dual output… yellow service engine soon, and red service engine now….

The monitor, when it senses certain things like engine knock… it can cut engine timing and reduce throttle input on its own… not ideal for driving…  not very workable for flying….

And for those that ignore the lights… the engine keeps limping along at about 80 percent power max, until the catalytic converter packs up…. Then you have 0% power until you can find a drill to poke holes in some sheet metal….    

 

On the other hand…

I just asked Alexa… who was Wilbur’s brother?  And she gave me both brother’s full names… 

When I typed ‘wilbur’ into Google… a few Wilburs showed up and Wright was third from the top….

 

I still like the real MSers better than the AI MSers…  the fake MSers have no personality or other machines they have….  

I think it took me about three hours to read this entire thread…

Go MS!

Best regards,

-a-

 

 

[T. Peterson]

9 hours ago, carusoam said:

Holy cow!

I think you guys have nailed it… three ways to Sunday!  

1) We can all benefit by good automation…

2) Good automation requires decent sensors…

3) Decent sensors have the ability to let things know when they aren’t working properly anymore… and not include the erroneous information… (post it note over the bad data)

4) For Automation to replace a two person crew…. That is a technical possibility… but requires the remaining crew member to not have health issues while at work… no heart attacks allowed.

5) did anyone mention a back-up plan for a single pilot casualty?  
The plane gets landed remotely, by a pilot controlling from corporate headquarters…?

6) Speaking of normal law….  How much computer, sensor, and logic technology is covered in flight school lately?  (Probably a lot more post y2k, than before)

Looks like we can be discussing the steps required to get to the full automation that is in the future…

R.I.P.   Gordon Moore, started at Fairchild electronics, before building Intel…

Moore’s Law…. The number of transistors on a chip will double every year…. (Stated in the mid 60s…) and was still holding when I learned about it in the mid 80s… before slowing to doubling every 1.6 years or so…  286, 386, 486, Pentium…. Then multiple cores and parallel processing…

Where Intel has run into trouble lately… they haven’t been able to make the transistors smaller like AMD and NVDA have been doing…

NVDA has announced their latest standard… a 4-nano standard… if this pans out… Moore’s law gets Re-invigorated…

Making the transistors smaller…

1) Saves energy

2) more efficient to keep cool

3) operates faster, does more work…. The amount of work required for real AI to occur.

4) fits in a smaller package

Proof of this… have you seen how well the faster GPS and faster autopilots actually control things…?  Until they have difficulty with things that stick or slip….  

The AI guys are happy with this chip, as are the Crypto miners….

https://www.pcmag.com/news/nvidias-latest-gpu-boasts-a-4-nanometer-process-from-tsmc-pcie-gen-5

Chip designs made in the US and Europe are very powerful….

Chip manufacturing for these designs are often in Taiwan, an island country off the coast of China… TSMc

There are a lot of trade restrictions for these technologies to keep chips, designs, and their manufacturing equipment… from making their way to China and Russia… capitalism, communism, and war…

 

Keep teaching STEM topics to everyone that can take it…

See how many computers are running in your favorite restaurant, barbershop, car mechanic’s shop…

 

For the person that asked… where is the engine monitor on that fancy instrument panel…. (Bar graph ends at 85mph, but indicated speed goes well past 100)

The engine monitor is behind the instrument panel with a dual output… yellow service engine soon, and red service engine now….

The monitor, when it senses certain things like engine knock… it can cut engine timing and reduce throttle input on its own… not ideal for driving…  not very workable for flying….

And for those that ignore the lights… the engine keeps limping along at about 80 percent power max, until the catalytic converter packs up…. Then you have 0% power until you can find a drill to poke holes in some sheet metal….    

 

On the other hand…

I just asked Alexa… who was Wilbur’s brother?  And she gave me both brother’s full names… 

When I typed ‘wilbur’ into Google… a few Wilburs showed up and Wright was third from the top….

 

I still like the real MSers better than the AI MSers…  the fake MSers have no personality or other machines they have….  

I think it took me about three hours to read this entire thread…

Go MS!

Best regards,

-a-

 

 

If we nailed it 3 ways to Sunday, I think you added another 13!

Well done!

[alexz]

On 4/2/2023 at 4:22 AM, carusoam said:

 

Making the Marking smaller…

1) Saves energy

2) more efficient to keep cool

3) operates faster, does more work…. The amount of work required for real AI to occur.

4) fits in a smaller package
 

You are right about almost everything except that currently making transistors smaller only leads to smaller package and faster operations. There is such a thing as leakage current and it’s gets larger as transistors get smaller so saving energy and keeping cool becomes a real problem  (just look at massive heatsink on you intel cpu or GPU). There is also quantum effects starts to play where you transistors can switch without a command  as well  cosmic ray particles with lower energy can cause switch, which leads to more circuitry to guard agains it which defeats making transistors smaller…. Limits of current technology is very near and getting closer and closer to them is more and more expensive. 

 

[N201MKTurbo]

1 hour ago, alexz said:

You are right about almost everything except that currently making transistors smaller only leads to smaller package and faster operations. There is such a thing as leakage current and it’s gets larger as transistors get smaller so saving energy and keeping cool becomes a real problem  (just look at massive heatsink on you intel cpu or GPU). There is also quantum effects starts to play where you transistors can switch without a command  as well  cosmic ray particles with lower energy can cause switch, which leads to more circuitry to guard agains it which defeats making transistors smaller…. Limits of current technology is very near and getting closer and closer to them is more and more expensive. 

 

Doing my side gig over at Intel, I get to see things maybe I’m not supposed to, but they keep marching on and it amazes me what they can do.

[Will.iam]

On 4/1/2023 at 12:22 PM, T. Peterson said:

All excellent points, Will, plus the fact that it was a failure of technology which put AF 447 in extremis. There is no doubt the pilots responded incorrectly, but is it a wonder? For years the airlines had pushed for almost total reliance on automation and this was especially true for many foreign airlines. I distinctly remember during our Airbus transition training a chief pilot walking in and saying that the new test of airmanship is how well we were able to program and manipulate the boxes. I can only imagine the culture at AF. I wonder if the AF pilots had ever flown with either the A/P, FD, or A/T off? How about all 3 off at one time? So in this culture of technological over dependence, the pilots are at 37000 feet in turbulence, IMC and unreliable airspeed and they screw it up. Now isn’t that surprising.

What AF 447 proves is the fallibility not of pilots, but of technology, and the abject stupidity of failing to train for when all the whizmo ghizmo fails. Proof? Why did all the airlines race into upset jet training after this accident?

We don’t even have the present technology perfected. I routinely have to help FeFe make crossing restrictions and yet we are postulating AI? Utterly premature.

Just got briefed on the A320 that went through the hail storm that destroyed the radome and cracked the windows.  The crew was turning to go through a split between cells of weather at 37,000ft when they directed FeFe to turn the autopilot only gave them a 10 degree turn rate instead of the 25 degrees the pilots were normally expecting to have, this enlarged their turn radius so that instead of turning before the cell they flew through it.  Now in hind sight they said they could have disconnected the autopilot and turned at a higher rate as the computer would give maximum turn rate up to 67 degrees or buffer to stall which ever was higher which at that gross weight at altitude would have been the buffer to stall.  My argument to that was if the autopilot was already limiting the turn rate to 10 degrees what makes them think that turning the autopilot off would have allowed them to turn any tighter? and if there was more buffer there to turn tighter why didn't the autopilot turn up to it's normal rate?  I would think if the autopilot was only giving me 10 degrees, I would have stopped the turn and turn the other direction to make a 270 degree turn thus flying away from the storm and giving me the space I need to align up with the gap they were trying to fly through.  besides flying in RVSM airspace you are not allowed to hand fly even though PTS standards holds me to +-100 ft and the autopilot was certified with +-200 ft but pilots seem to be reluctant to declare an emergency to disconnect the autopilot in RVSM airspace.  I guess they also assumed it would not be that bad and just accepted the flight path through the edge of the red part of the radar as I'm sure if they had known what was about to happen they would have done everything possible to avoid that experience.

[T. Peterson]

41 minutes ago, Will.iam said:

Just got briefed on the A320 that went through the hail storm that destroyed the radome and cracked the windows.  The crew was turning to go through a split between cells of weather at 37,000ft when they directed FeFe to turn the autopilot only gave them a 10 degree turn rate instead of the 25 degrees the pilots were normally expecting to have, this enlarged their turn radius so that instead of turning before the cell they flew through it.  Now in hind sight they said they could have disconnected the autopilot and turned at a higher rate as the computer would give maximum turn rate up to 67 degrees or buffer to stall which ever was higher which at that gross weight at altitude would have been the buffer to stall.  My argument to that was if the autopilot was already limiting the turn rate to 10 degrees what makes them think that turning the autopilot off would have allowed them to turn any tighter? and if there was more buffer there to turn tighter why didn't the autopilot turn up to it's normal rate?  I would think if the autopilot was only giving me 10 degrees, I would have stopped the turn and turn the other direction to make a 270 degree turn thus flying away from the storm and giving me the space I need to align up with the gap they were trying to fly through.  besides flying in RVSM airspace you are not allowed to hand fly even though PTS standards holds me to +-100 ft and the autopilot was certified with +-200 ft but pilots seem to be reluctant to declare an emergency to disconnect the autopilot in RVSM airspace.  I guess they also assumed it would not be that bad and just accepted the flight path through the edge of the red part of the radar as I'm sure if they had known what was about to happen they would have done everything possible to avoid that experience.

I turn the auto pilot off any time it irritates me by not doing or doing too slowly what I want it to do. I don’t concern myself one iota with RVSM airspace. The rules are made to be servants not masters.

[ilovecornfields]

2 hours ago, T. Peterson said:

I turn the auto pilot off any time it irritates me by not doing or doing too slowly what I want it to do. I don’t concern myself one iota with RVSM airspace. The rules are made to be servants not masters.

Makes sense to me. I assume there’s a Part 121 equivalent to Part 91.3 that allows you to do so.

Seems that if the autopilot is about to fly me into a dangerous situation then it would be negligent of me NOT to disengage it, regardless of which airspace I’m in. I understand it’s more complicated with the fly-by-wire planes but the few airline unusual attitude videos I’ve seen (I think they were produced by American) were very clear on disengaging the automation when it’s not doing what you want it to do.

Is this any different then the Cirrus pilot who used the AP to make a perfect standard rate turn right into the Metroliner?

[ilovecornfields]

https://airfactsjournal.com/2023/07/what-pilots-can-teach-the-world-about-ai/

Thought this was an interesting commentary on the future of AI in aviation.