Plane crash

[AKKuya]

We need MORE human air traffic controllers WITH less responsibilities.  DELEGATION is key to the system.  Each controller is assigned a specific task and or smaller geographical area.

AI is simply a tool for the humans in DECISION making.

[LilMak]

True but an autopilot can already takeoff, fly to a destination, and land a plane. With more brains it can do that and handle emergencies too. It could also link to an AI controller communicating thousands of times faster than humans and error free.

I understand what you’re getting at but I can say (as an insider) that, even with the technology in currently place, there are hard limits to what we’re capable of at this point. It’s not so much about training AI so much as it is about the infrastructure that still needs to catch up and the ability to integrate all the systems. Air traffic is an unbelievably complex set of systems that have been built in for over a hundred years. It’s not some mainframe somewhere you can just swap out for a new one.

[Busher]

I understand what you’re getting at but I can say (as an insider) that, even with the technology in currently place, there are hard limits to what we’re capable of at this point. It’s not so much about training AI so much as it is about the infrastructure that still needs to catch up and the ability to integrate all the systems. Air traffic is an unbelievably complex set of systems that have been built in for over a hundred years. It’s not some mainframe somewhere you can just swap out for a new one.

You are exactly on point. I am a retired airline pilot with a little over 30,000 hours in various airplanes.

The poster that claims AI can do these miraculous feats with no need for pilots clearly has no knowledge of aviation beyond AH. While it likely will come in the future the dollars and technology needed to create an autonomous aviation system are a long way off.

Just to clarify one of the false statements only one major transport-category jet is currently certified for fully automatic (autonomous) takeoff:

It's the Airbus A350‑1000 with the Airbus Autonomous Takeoff  system.
This capability was demonstrated and validated in 2020, but it is not yet in routine airline service.

Autoland is nothing new. It was developed in the 1960s and as most pilots with an instrument rating know that the runways used for Cat II and Cat III landings have advanced equipment only installed at major airports.

In closing I would love to know how the eggheads are going to teach a computer how to deviate around convective weather as well as an experienced human.

[AKIron]

I understand what you’re getting at but I can say (as an insider) that, even with the technology in currently place, there are hard limits to what we’re capable of at this point. It’s not so much about training AI so much as it is about the infrastructure that still needs to catch up and the ability to integrate all the systems. Air traffic is an unbelievably complex set of systems that have been built in for over a hundred years. It’s not some mainframe somewhere you can just swap out for a new one.

I was an ATC radar tech for 20 years in the Air Force. I know how complex the systems are. It will take some time but it is coming. There will still be mechanical failures but human error will become a welcome thing of the past.

[Busher]

I was an ATC radar tech for 20 years in the Air Force. I know how complex the systems are. It will take some time but it is coming. There will still be mechanical failures but human error will become a welcome thing of the past.

I think you might consider re-focusing your desire to implement AI in the interest of safety.

General aviation deaths in airplanes since 2020 (and this includes many operations beyond scheduled airlines) is 118.

Highway Traffic fatalities only in 2025 was over 37,000 (which was happily, a 12% improvement over 2024).

I don't think the money needs be spent in aviation.

[AKIron]

My desire, and yours, are irrelevant. People demand air travel to be the safest it can be made to be. AI will prove to be the best solution.

[icepac]

AI is still naive and wrong often. 

I remember the earliest days of attempting AI where we used the texas instruments MVP chipset and later nestor Ni1000 neural network chipset but, even then, the data that makes up a knowledge base was massive.

[Eagler]

Hopefully they do what is safest not most profitable..

Then again lawsuits are very expensive..

Eagler

[AKIron]

There's really no excuse for what happened in this accident. There are strict procedures in place to prevent this sort of thing. Airports are not the wild west. Even if the emergency vehicle was given clearance to enter the active they should have looked. If you watched the video you can see how they should have seen the aircraft.

[LilMak]

There's really no excuse for what happened in this accident. There are strict procedures in place to prevent this sort of thing. Airports are not the wild west. Even if the emergency vehicle was given clearance to enter the active they should have looked. If you watched the video you can see how they should have seen the aircraft.

Which critically points out the fallacy of just plugging in AI into ATC as the solution that will magically fix everything. There was still a human in the truck. I doubt people will want to simply give over everything to our AI overlords in the name of safety. There WILL be pushback and there are cost prohibitive limits. For example, maybe the fire truck needs special autonomous equipment. That would be expensive. What about lawn mowers? (that grass between the taxiway and runway doesn't mow itself) Aircraft tugs? Do baggage handlers need shock collars to prevent them from walking in front of aircraft? What about personal aircraft? Would I have to retrofit that equipment a 1948 Cessna 170 in order to fly it? This is just the stuff that immediately pops into my head without much thought.

[AKIron]

The video I watched showed the truck was cleared to cross. Then the controller realized his error and called for the truck to stop. AI would not have cleared the truck to cross. The most responsible for this was the controller.

[AKIron]

People have always embraced new technology reluctantly then once accepted can't live without it. I can think of no better use of AI than in air traffic. It's easy enough to simulate an airport environment to train and evaluate AI. I'd be surprised if that's not already being done.

[icepac]

Ground control has deteriorated ever since kennedy steve retired.

[Randy1]

When I was involved in power plant accident reviews for the root cause analysis, the first thing you had to do was push human error out of the review process.  You have to this to avoid what you might call a rubber stamp decision.  You have to identify all things that influenced the accident occurring.   That list would be quite long.

[hazmatt]

What happens if everything is controlled by ai, and the hardware or power that's running the ai fails?