Planes have had “Ai” for decades, it was called autopilot. This is for air traffic controllers. They sit in the tower and basically coordinate all the traffic on around the airport. Why this is important, look at the crash at laguadia(airport in NYC I’m likely spelling wrong). Basically fire crew wanted to cross run towards an unrelated incident, plane wanted to land. Atc told fire crew they were clear and told the plane they were clear to land… On same runway. Plane crashed into water truck of fire crew causing severe damage to both vehicles. “luckily” only the 2 pilots of the plane lost their lives, it could of been so much worse.
MAYBE I could see a use of Ai. Currently their are sensors to detect where planes and land vehicles are but imagine if an llm could interpret the talk on the comms and detect if two statements could result in a collision. Especially if there are multiple Atc officers active. Like Tom tells one plane to land on runway 22 and Lucy tells another plane to land on runway 22 and Tom and Lucy don’t hear eachother. Or if a single officer tells a plane to land on runway 30 but they were 10 minutes out. A truck to refill the food/drinks asks to cross runway 30 when the plane is only 20 seconds out. I could see the Atc officer forgetting about the plane and granting clearance.
Ai logs all these orders, reading the sensors for current situation and activates visual and audio alarms that two orders may conflict. Officer could tell the plane to go around, maybe cost some fuel and time for a false alarm but potentially save hundreds of lives. Maybe Ai doesn’t detect conflicting orders and it’s up to the human officers to notice. We are now in the same situation we are currently. Now, if you try to eliminate the human element and just rely on Ai people will die in many accidents.
It’s insane how many people confuse automation with AI. Autopilot is not AI. Computers are not AI. Even LLMs are not a full general AI. They’re language models and very badly suited to handling novel live situations like ATC.
Autopilot is not AI. It’s basically a system that makes sure that some gyroscopes are aligned with whatever destination is typed in. It makes no artificial intelligent choices of any kind.
Air traffic controlling requires a lot more data processing. A lot of it could potentially be made or is already made with algorithms, provided that the data is based on trustworthy sensors and radars that are always up to date.
They’re probably not, so humans need to sort out which data to trust and make decisions that sometimes conflict with the erroneous data.
It makes no sense to throw a language model at the task.
Planes have had “Ai” for decades, it was called autopilot. This is for air traffic controllers. They sit in the tower and basically coordinate all the traffic on around the airport. Why this is important, look at the crash at laguadia(airport in NYC I’m likely spelling wrong). Basically fire crew wanted to cross run towards an unrelated incident, plane wanted to land. Atc told fire crew they were clear and told the plane they were clear to land… On same runway. Plane crashed into water truck of fire crew causing severe damage to both vehicles. “luckily” only the 2 pilots of the plane lost their lives, it could of been so much worse.
MAYBE I could see a use of Ai. Currently their are sensors to detect where planes and land vehicles are but imagine if an llm could interpret the talk on the comms and detect if two statements could result in a collision. Especially if there are multiple Atc officers active. Like Tom tells one plane to land on runway 22 and Lucy tells another plane to land on runway 22 and Tom and Lucy don’t hear eachother. Or if a single officer tells a plane to land on runway 30 but they were 10 minutes out. A truck to refill the food/drinks asks to cross runway 30 when the plane is only 20 seconds out. I could see the Atc officer forgetting about the plane and granting clearance.
Ai logs all these orders, reading the sensors for current situation and activates visual and audio alarms that two orders may conflict. Officer could tell the plane to go around, maybe cost some fuel and time for a false alarm but potentially save hundreds of lives. Maybe Ai doesn’t detect conflicting orders and it’s up to the human officers to notice. We are now in the same situation we are currently. Now, if you try to eliminate the human element and just rely on Ai people will die in many accidents.
It’s insane how many people confuse automation with AI. Autopilot is not AI. Computers are not AI. Even LLMs are not a full general AI. They’re language models and very badly suited to handling novel live situations like ATC.
Autopilot is not AI. It’s basically a system that makes sure that some gyroscopes are aligned with whatever destination is typed in. It makes no artificial intelligent choices of any kind.
Air traffic controlling requires a lot more data processing. A lot of it could potentially be made or is already made with algorithms, provided that the data is based on trustworthy sensors and radars that are always up to date. They’re probably not, so humans need to sort out which data to trust and make decisions that sometimes conflict with the erroneous data.
It makes no sense to throw a language model at the task.
But no one has said language model, they said Ai of which there are many types.