Crash at Reagan National Airport, DC. Helicopter crash into airliner.

AKiss20 said:
And that’s different for when an AI ATC goes out how?
Click to expand...
The controller could walk outside with a handheld. The option for AI? Regardless, the context of my original post was to AI replacing human ATC. Having experienced the failure and fixed a number of aviation safety critical systems to include their redundancy systems, we are still far from replacing certain aviation positions with AI for various reasons to include the loss of power scenario.
 
Bell206 said:
The controller could walk outside with a handheld. The option for AI? Regardless, the context of my original post was to AI replacing human ATC. Having experienced the failure and fixed a number of aviation safety critical systems to include their redundancy systems, we are still far from replacing certain aviation positions with AI for various reasons to include the loss of power scenario.
Click to expand...
Why couldn't AI switch to a battery operated backup? And in 2025, data center is designed to operate indefinitely without connection to the external grid or commonplace. Not to mention, that AI in geographically diverse data centers could much more easily back up ATC facilities all across the country. The logistics aren't really the roadblocks here.
 
Warmi said:
How do you know that you know anything and what does that even mean knowing something ?
Our brains work using the same basic binary primitive switches numbering in billions as do computers so how do you know that whatever you consider knowledge is not just some sort of multidimensional vector using basic vector maths and dot product like comparisons producing an illusion of what we refer to as intelligence … it has to be something along these lines given that there is nothing rally magically more complicated inside of our heads beyond a network of binary switches.

Btw , given huge advancements in AI perception and scoring abilities , I fully expect AI to start using almost exclusively self reinforcing learning where it simply produces effectively unlimited number of randomly generated scenarios and relies on what is basically an insanely accelerated , brute force , natural selection like process to accumulate knowledge …
We shall see …
Click to expand...
You just went totally Merleau Ponty on us there
 
Lindberg said:
Why couldn't AI switch to a battery operated backup?
Click to expand...
That would only be an option if the failure was not infrastructure related. Unfortunately, most ATC issues are. For example, the ANC earthquake 7-8 years ago required tower personal to evacuate and once they runways were inspected, they resumed work out of a pickup parked next to the runway with portable radios. I guess if you include AI robots into the mix that may work if they were programmed for such an event? To me the biggest hurdle for AI to jump, just as computers were or still are in some cases, is the interface into the existing "analog" side of the aviation industry.
 
I don't think we'd go from 100% human to 100% AI overnight. We'd start with AI tools that make the humans more productive and more informed to make better decisions.

Right now, the traffic management programs could really use some AI. Same for creating a plan that will work over multiple sectors and Centers. Today, we'll be slowed down, and turned off course, by one sectors then given a shortcut and sped up by the next. Each controller is making his own plan with very generic guidance from downline sectors.
 
NextGen seems like a long-range plan to eliminate flight crews. Then airlines could put dummy controls up front like that hotel simulator in Colorado Springs and charge extra for passengers to "fly" the plane. :)
 
Bell206 said:
That would only be an option if the failure was not infrastructure related. Unfortunately, most ATC issues are. For example, the ANC earthquake 7-8 years ago required tower personal to evacuate and once they runways were inspected, they resumed work out of a pickup parked next to the runway with portable radios. I guess if you include AI robots into the mix that may work if they were programmed for such an event? To me the biggest hurdle for AI to jump, just as computers were or still are in some cases, is the interface into the existing "analog" side of the aviation industry.
Click to expand...
AI doesn't need a handheld. AI doesn't need radar screens. AL doesn't need a tower. AI doesn't need to be at the airfield. If AI gets beyond augmenting humans in ATC, the infrastructure will be entirely different.
 
Lindberg said:
AI doesn't need a handheld. AI doesn't need radar screens. AL doesn't need a tower. AI doesn't need to be at the airfield. If AI gets beyond augmenting humans in ATC, the infrastructure will be entirely different.
Click to expand...

NextGen has taken over 20 years and was nothing more than upgraded screens, computers and radars. How long do you think a substantial change in infrastructure to support AI will take the government? Elon could probably do it in 6 months if unleashed, the Feds maybe by next century.
 
I'll trust AI the day that AI controls birds. And don't get me started on the whole "birds aren't real" nonsense.
 
midwestpa24 said:
NextGen has taken over 20 years and was nothing more than upgraded screens, computers and radars. How long do you think a substantial change in infrastructure to support AI will take the government? Elon could probably do it in 6 months if unleashed, the Feds maybe by next century.
Click to expand...
Elon could do it in six months and then the mid air collision and CFIT rate would 100x within a month.
 
Lindberg said:
If AI gets beyond augmenting humans in ATC, the infrastructure will be entirely different.
Click to expand...
Exactly. And the FAA, EASA, ICAO have been looking into those requirements for several years. But as I mentioned, until an infrastructure is built from the ground up based on AI, and to an extent aircraft as well, there will still be issues adapting to the existing systems. Plus any AI system certifications will also require participation and agreement on an international level. Regardless, AI still needs hardware to perform its job so for now its still stuck with handhelds, radar, and towers for the foreseeable future.

midwestpa24 said:
NextGen has taken over 20 years and was nothing more than upgraded screens, computers and radars.
Click to expand...
FYI: NextGen is a long-term, opened-end program to modernize the NAS as new technologies and equipment are developed. Having listened in on the NextGen work for AAM/UAM incorporation into the NAS, it’s a bit more than new screens, computers, and radars.
 
AKiss20 said:
Elon could do it in six months and then the mid air collision and CFIT rate would 100x within a month.
Click to expand...
...and he'd spin it with a cute euphemism while fanboys raved about 'progress.'

Nauga,
and that cartoon about the birth of advertising
 
AKiss20 said:
Elon could do it in six months and then the mid air collision and CFIT rate would 100x within a month.
Click to expand...

I suppose that might be possible, if his team completely ignored the controller union and the various ground empires
 
Bob Noel said:
I suppose that might be possible, if his team completely ignored the controller union and the various ground empires
Click to expand...
Those are probably related to the regulations about issuing an NPRM and getting user/industry input. Hasn't had to do it for changes at other agencies, so expect he's got a hall pass.

From the limited detail available on the makeup of DOGE, it sounds like none of the Dogies have any aviation experience.

Ron Wanttaja
 
I’ve seen how well auto generated subtitles work.

AI is going to have fun with all those accents and very few following the Pilot/Controller glossary exactly.
 
wanttaja said:
Those are probably related to the regulations about issuing an NPRM and getting user/industry input. Hasn't had to do it for changes at other agencies, so expect he's got a hall pass.

From the limited detail available on the makeup of DOGE, it sounds like none of the Dogies have any aviation experience.

Ron Wanttaja
Click to expand...
They seem pretty decently informed about space travel, though.
 
Lawson_Stone said:
They seem pretty decently informed about space travel, though.
Click to expand...
Odd to consider the comparison...no ATC in space, no overall coordination, no way to control over a particular "airspace." No birds, but plenty of little leftover bits that you can just barely track but can still kill you.

And just like many airports, no touch-and-goes......

Ron "Losing an argument of perigee" Wanttaja
 
Elon is running a company that basically commoditized space launch business … while not suggesting that he has any demigod like abilities ( as seem to be the case recently with some news outlets ) but collectively his company is well equipped to handle tough engineering problems so I would be reluctant to dismiss him with “ the mid air collision and CFIT rate would 100x within a month” type of silly remarks.
 
I'm not sure turning ATC modernization over to a guy whose technical development plan includes "rapid unscheduled disassembly" events is a good idea...
 
Warmi said:
Elon is running a company that basically commoditized space launch business … while not suggesting that he has any demigod like abilities ( as seem to be the case recently with some news outlets ) but collectively his company is well equipped to handle tough engineering problems so I would be reluctant to dismiss him with “ the mid air collision and CFIT rate would 100x within a month” type of silly remarks.
Click to expand...
Spacex (not Elon) is an evolution of established technologies. Nothing they are doing involves massive fundamental shifts in human knowledge. They have done some impressive things but fundamentally it’s aimed at operating in a different tradeoff space than those before them as well as developing evolutionary, not revolutionary, technologies.

ATC AI would represent a complete revolution in technology. We know so little about the bounds of AI as it is such a new space. Everyone is throwing it at every problem because it is the shiny new thing that no one has seen before and we are quickly seeing that there are problem spaces where our fledgling versions of it work surprisingly well and problem spaces where it falls squarely on its face. We need to have a much better understanding of this space before we are going to routinely be relying on it for safety critical decision making and operations without HIL. That type of understanding does not get developed in 6 months. When Tesla introduced autopilot 10 years ago, there was much speculation that we would have SAE level 5 within the scope of years or a decade. It’s now 10 years later and we barely have SAE level 4 (waymo) in some of the least challenging environments and SAE level 2-3 in everything else.
 
Last edited:
AKiss20 said:
Spacex (not Elon) is an evolution of established technologies. Nothing they are doing involves massive fundamental shifts in human knowledge. They have done some impressive things but fundamentally it’s aimed at operating in a different tradeoff space than those before them as well as developing evolutionary, not revolutionary, technologies.

ATC AI would represent a complete revolution in technology. We know so little about the bounds of AI as it is such a new space. Everyone is throwing it at every problem because it is the shiny new thing that no one has seen before and we are quickly seeing that there are problem spaces where our fledgling versions of it work surprisingly well and problem spaces where it falls squarely on its face. We need to have a much better understanding of this space before we are going to routinely be relying on it for safety critical decision making and operations without HIL. That type of understanding does not get developed in 6 months. When Tesla introduced autopilot 10 years ago, there was much speculation that we would have SAE level 5 within the scope of years or a decade. It’s now 10 years later and we barely have SAE level 4 (waymo) in some of the least challenging environments and SAE level 2-3 in everything else.
Click to expand...
I think something got crossed in this thread - AI for ATC and Elon driving air space modernizations are not necessary directly related , I was referring more to this development https://www.aerotime.aero/articles/elon-musk-notam-us-air-traffic-system when it comes to Elon Musk , rather than him replacing everything with AI 6 months from now.
 
Dana said:
I'm not sure turning ATC modernization over to a guy whose technical development plan includes "rapid unscheduled disassembly" events is a good idea...
Click to expand...

"cute" sound bite, but the flip side is that wrt engineering testing is that if you don't have any test failures, then your tests aren't properly designed.
 
Warmi said:
Elon is running a company that basically commoditized space launch business … while not suggesting that he has any demigod like abilities ( as seem to be the case recently with some news outlets ) but collectively his company is well equipped to handle tough engineering problems so I would be reluctant to dismiss him with “ the mid air collision and CFIT rate would 100x within a month” type of silly remarks.
Click to expand...
I tend to agree with you, but: I don't think Musk or any of his Dogies realize the magnitude of the problem he so flippantly announced he could solve.

45,000 known flights per day. That's not that much, to a computer person. Should be able to work an algorithm that merges the flight plan data, ADS-B out, Mode C out, and radar data, right?

The trouble is, that's KNOWN flights per day...basically VFR and IFR flight plans. It doesn't include all the VFR traffic that doesn't file flight plans or talk to ATC...and many of them have neither Mode C or ADS-B out. The Dogie programmers are going to freak when they find out. The solution (to them) will be obvious: EVERYONE on a flight plan, EVERYONE with a transponder and ADS-B, EVERYONE calls and gets permission for any action. Just clone the Russian ATC system.

Bob Noel said:
.... the flip side is that wrt engineering testing is that if you don't have any test failures, then your tests aren't properly designed.
Click to expand...
As "traditional" space development guy, I really get a kick out of the SpaceX testing philosophy. They don't have Congress breathing down their neck, they instrument the vehicles to the point where they learn more from a failure than a success, and they have management that accepts losses. Kudos to them, definitely.

However, it is NOT a philosophy that carries into an ATC replacement. All of SpaceX testing is done with unmanned vehicles. Maybe you could set up a simulation with ~90,000 flights, but it'll be hard to factor in randomness of the weather, the stupidity pilots get into, pop-up airspace restrictions, etc. This isn't a six-month program; this is probably a six YEAR program with one HECK of a simulation needed. With no budget allocated for it.

Ron Wanttaja
 
The SpaceX test flights are testing hardware and the conventional software that operates them. They are designed by engineers and programmed by coders.

An ATC AI will not be a chatbot. It won't replace all of the controllers, at least not for a very long time. It will collect and process data to produce solutions that have a much wider reach than anything individual controllers could produce. They will predict conflicts long before they happen and provide deconfliction earlier.
 
Warmi said:
Elon is running a company that basically commoditized space launch business … while not suggesting that he has any demigod like abilities ( as seem to be the case recently with some news outlets ) but collectively his company is well equipped to handle tough engineering problems so I would be reluctant to dismiss him with “ the mid air collision and CFIT rate would 100x within a month” type of silly remarks.
Click to expand...
Elon's rocket program isn't interfacing with anyone or anything. Here's the launch window, the permits are filed, and here's the orbit. Push the button and hope it goes. Operationally, there is very little coordination. That's great from an engineering perspective because he doesn't have to deal with all of the crap that exists outside of his bubble.

ATC will be a little different with birds, weather, runway closures, failed com's, emergencies, pop-up IFR, VFR, drones, kites, nuts in lawn chairs supported by balloons, Chinese balloons, non-ADSB helicopters, and a thousand other things I'm not gonna take time to list.
 
kyleb said:
Elon's rocket program isn't interfacing with anyone or anything. Here's the launch window, the permits are filed, and here's the orbit. Push the button and hope it goes. Operationally, there is very little coordination. That's great from an engineering perspective because he doesn't have to deal with all of the crap that exists outside of his bubble.
Click to expand...
That's an impressive minimization of the quite remarkable accomplishments of SpaceX.
 
kyleb said:
Elon's rocket program isn't interfacing with anyone or anything. Here's the launch window, the permits are filed, and here's the orbit. Push the button and hope it goes. Operationally, there is very little coordination. That's great from an engineering perspective because he doesn't have to deal with all of the crap that exists outside of his bubble.

ATC will be a little different with birds, weather, runway closures, failed com's, emergencies, pop-up IFR, VFR, drones, kites, nuts in lawn chairs supported by balloons, Chinese balloons, non-ADSB helicopters, and a thousand other things I'm not gonna take time to list.
Click to expand...
What people seem to miss is the fact that Musk is not modern Albert Einstein but rather another Henry Ford type of figure. His magic is not so much about brilliantly solving tough engineering problems but rather solving them in ways that are both , efficient in terms of getting them to the market quickly as well as making them long term sustainable ( aka profitable .)
Space X with be remembered not for getting stuff into space but for getting it up there at scale and at 1/10th of the cost of everyone who came before them.

Whatever the complications of running ATC are ( as compared to running a space company ) the complain seems to be that it takes forever and costs crapload of money to get things moving in this domain so , I guess , another Henry Ford like approach to the problem may not be the worst idea …
 
3393RP said:
That's an impressive minimization of the quite remarkable accomplishments of SpaceX.
Click to expand...
Not minimizing anything. Just relating the differences between solving an engineering problem where you control most of the variables to solving a systems problem where there are an incredible number of dynamic outside influences where you have little/no control. Think about driverless cars and all of the wonky edge cases which have made that so challenging. Then do it at 600 MPH and take away the option of pulling to the side of the road if things get out of hand.
 
Bell206 said:
The controller could walk outside with a handheld. The option for AI? Regardless, the context of my original post was to AI replacing human ATC. Having experienced the failure and fixed a number of aviation safety critical systems to include their redundancy systems, we are still far from replacing certain aviation positions with AI for various reasons to include the loss of power scenario.
Click to expand...
It wouldn't even require AI; if all planes in a given airspace were required to have ADS-B or equivalent, the coding for collision avoidance would be trivial. In that it already exists in several forms, and for greatly complicated scenarios, compared to airspace around a busy metro area.
So essentially replacing TCAS with something else; the "logic" should likely be homed on the ground, but it could be in individual aircraft. Regardless, a new piece of equipment would be needed for aircraft. When ADS-B was required many considered it to be already out-of-date.
 
Kenny Phillips said:
It wouldn't even require AI; if all planes in a given airspace were required to have ADS-B or equivalent, the coding for collision avoidance would be trivial. In that it already exists in several forms, and for greatly complicated scenarios, compared to airspace around a busy metro area.
So essentially replacing TCAS with something else; the "logic" should likely be homed on the ground, but it could be in individual aircraft. Regardless, a new piece of equipment would be needed for aircraft. When ADS-B was required many considered it to be already out-of-date.
Click to expand...

I don't think the collision avoidance algorithm is quite as trivial as you might think. It's not difficult to come up with realistic scenarios where escape maneuvers create even more collision hazards.

(then there is the whole problem of dealing with spoofed ADS-B messages).
 
Some of the following may be sarcasm, so.....

The airlines launch into the air on a schedule, always with a near full load of people, nearly regardless of the weather, and recover to the ground environment with an assumed 100% regularity. With slightly less regularity, the passengers are at an unplanned location, and forwarded at airline expense to their original destination.

Space X launches in carefully analyzed weather with no high winds or gusts, no thunder in the area, and a precise time. Any undesirable anomaly, the departure is rescheduled, with no passenger inconvenience, and the main liftoff device is only most of the time recovered.

SpaceX scheduling and most of the time safely returning to reuse of the launch system would not play well with ordinary travelers.



"99% of the air travelers today arrived at a suitable landing site safely today. The remaining 1 % ran out of fuel and landed at unplanned locations! Condition of the passengers on the 1% will be reported by local authorities as the information is discovered."
 
Lindberg said:
Did anyone say that SpaceX and ATC solve the same problem?

I believe the suggestion was more like, people who are good at solving complex problems are good at solving complex problems.
Click to expand...
Engineering i.e. hardware design is a vastly different undertaking than designing a system that has to operate with a ton of other systems that are impacted by things not necessarily in their direct control.

Again, the example of self driving cars comes to mind.
 
Dana said:
I'm not sure turning ATC modernization over to a guy whose technical development plan includes "rapid unscheduled disassembly" events is a good idea...
Click to expand...
that's not entirely fair. He has spoken openly about those events as expected and a part of the research and development part of any space craft development. Go back and watch all the US and Soviet early attempts at space launches and you will see the same thing. Presumably ATC modernization would proceed with a lot more awareness of what has already been done, what works, what doesn't, etc. Musk seems to get that.
 
kyleb said:
Engineering i.e. hardware design is a vastly different undertaking than designing a system that has to operate with a ton of other systems that are impacted by things not necessarily in their direct control.

Again, the example of self driving cars comes to mind.
Click to expand...
Ok, I don't think you're contradicting the quoted post. Certainly you recognize that SpaceX hardware operates with a ton of systems that are impacted by things but in their direct control. I don't think anyone would claim that everything between firing a rocket into space and catching it back at the launch pad or landing it on a floating barge, not to mention docking and recovering a spacecraft with the space station, is within SpaceX's control.
 
You must log in or register to reply here.
Back
Top