Is this an LLM style AI, or a machine learning AI.
If it's a LLM, it's not worth a penny
If it's a true machine learning AI, then I can see the worth of it.
Is this an LLM style AI, or a machine learning AI.
If it's a LLM, it's not worth a penny
If it's a true machine learning AI, then I can see the worth of it.
You're absolutely right to push back, I should have landed that plane before it ran out of fuel. Would you like me to explore options for funerals?
So a quick search suggests air traffic controllers make around $75,000 annually. $875 million would pay the salaries of 11,666 air traffic controllers for a decade at that wage.
Even considering inflation, training, medical, etc that $875,000,000 could pay for thousands of hired Amemricans for a decade. Republicans are literally the worst.
$75,000 annually
The Bureau of Labor Statistics reports about double that as the median pay ($148,080).
I still agree with the rest of your math and think the investment is worth it, but just clarifying the number.
the other issue is aparently it's hard to get people to do that job.
I do believe its considered top 2 most stressful jobs, not hard to see why
I know someone who’s atc. He described it as “very very slow Tetris, but when you misplace a piece someone dies.”
I was listening to live ATC broadcast this evening for O'Hare ground control.
Jesus tapdancing Christ, no idea how they do it.
All you hear is a steady stream of instructions like "j1234, approach 10L and hold, then sierra sierra golf tango tango alpha" which are "directions" for pilots to taxi to a gate.
They are ripping them off at hyperspeed, no mistakes. The audio quality of some pilots is just terrible but somehow they know wtf they said. They have dozens of planes on the ground at a time, organizing the chaos of them all without blinking. In the hour I listened, I heard one dude doing it all, though I think there are two different ground controllers.
Planes blocking taxiways, pilots misunderstanding instructions, someone is taking too long to push back from the gate and they're making changes and adapting on the fly.
Holy shit y'all.
Listen for yourself and try to figure it out. Audio: https://www.liveatc.net/play/kord1n2_gnd_in.pls Ground mapd: https://www.flightaware.com/resources/airport/ORD/APD/AIRPORT+DIAGRAM/pdf
Good luck.
This touches on an interesting point: relying solely on radio for route instructions is archaic and we desperately need to overhaul it with a digital alternative. Why can't we take some of that budget to modernize ATC?
The audio quality of some pilots is just terrible but somehow they know wtf they said.
That is in my experience highly dependent on what kind of radio/antenna is used to record the live broadcast, it's orientation and it's location.
I have an Icom IC-R6 radio scanner, even at the airport with line of sight of the infrastructure you can still get a terrible reception if the angle of the radio is wrong, the antenna is wrong or you have parked in the wrong spot.
People might reconsider if we give them a 30hr work week, but still pay the full salary. Some jobs are just harder and time worked doesn't equal effort. I've had 2 day stints that should have paid me more than the entire preceding month!
What I'm trying to say is that the people who run America are unwilling to pay to do it right.
I bet they could get more to do it if they paid more. You could give out a lot of raises with $875 million.
Seems better yes, in EASA airspace ATC is around 100k when starting your career and 220k near retirement.
I bet FAA is within those numbers.
The problem is finding qualified people to be controllers. IiRC, something like 3000 apply for certification per year and only 150 make it.
This position should really pay alot more than it does
$875 million would pay the salaries of 11,666 air traffic controllers for a decade at that wage.
You're off by a factor of 10 here. It's either a year of wages for 11666 controllers or 10 years of wages for 1166 controllers.
That should go swimmingly
Only if there’s a body of water near the terminal.
I think they should test it out on the private jets first. Ya know, to make sure it all works great.
Absolute worst possible use of an LLM: something safety-critical.
I'm sorry, you're right. I should not have directed that plane to the runway with the two planes that have already crashed into each other.
It's fine the first two have obviously canceled each other out leaving the runway clear.
No, no, it's sorting the crashes. If it had directed the planes to other runways, then there would be more crashes to clean up.
I guess the next time I want to travel internationally I'll take a train to Canada and fly out from there.
I don't see ANY comments in here where it's clear that people read the fucking article.
"Har har, AI and ATC, I'll make a funny joke about flying will be unsafe". Yes, it's expected, but if you actually read what this is - it's going to issue predictions about how busy particular times are, basically.
ZERO use of the AI in the tower by ATC to control aircraft in any way.
Sure, they might try to introduce such a thing - at which time your jokes will actually apply!
it’s going to issue predictions about how busy particular times are, basically.
and for that it's somehow worth $875M.
That's only step one. Just wait.
Welp, I'm never flying again. And fuck my house is in a main flight path.
If you're in a main flight path you will be safe.
"It looks like I may have confused the available freeway lanes for a designated runway. Did you know I405 contains 523,000 times more asphalt than LAX? I will work on securing an additional freeway for future landings and I have prepared a list of new flight paths."
FAA describes the SMART system
The media and old boomers are going to love talking about this system the moment a plane crashes.
8,750 salaries worth of money betting AI can do this with no actual evidence it can.
900 million could just hire a lot of skilled engineers and air traffic operators that can be well paid but that doesn’t sound as cool and futuristic as AI
Ah yes, the same technology that was used to kill 123 Iranian school children
In theory, an automated system should be able to outperform humans at a rote task like this under the right conditions. However, I don’t imagine throwing LLMs in the mix without redesigning the whole system for automation is going to end well.
top 50 comments