The $50 Billion Bloodbath: Why Airline Operations Control Centers Will Be Ghost Towns by 2030

I fly cargo so everyone assumes it’s coming to our side of the world first, seems logical. The one thing I can’t figure out is if everyone loses their job, who is going to be able to afford airline tickets to fill seats on airplanes???? Further if we start replacing white collar jobs who is going to buy shares of AI company of the week so they have the funny money to develop all of this. At a certain point the hamster dies and the wheel stops spinning.

At the end of the day AI can’t replace a sailboat and that’s where you’ll find me when it’s all said and done and we have no point to continue running on the wheel for other people’s benefit.
So....

I mean, "yes" but also... the economics will decouple from reality. Google just announced yesterday (or maybe Friday, don't remember) their tooling to allow bots to access money. You'll have bots buying and selling virtual assets first, then bots buying and selling physical assets, and the hamster wheel will keep spinning. There'll be robots hiring people to do stuff the robots can't do, etc. In the short run I expect inflation to continue, in the long run I expect a massive deflationary pressure on most "stuff" that people can buy because to paraphrase what you say, "who is going to be able to afford [literally anything]?" In the short run, continued inflation, but beyond some crazy inflection point I expect competition and political instability to undercut the whole thing. But the robots themselves will prop up the economy. It'll be jarring, because you'll see massive reductions in well being for people you personally know, but the S&P500 will never be higher. We're kind of already there where the material conditions of the average person don't match the value of the market in any meaningful way.

After a certain point, that will become untenable and politicos will start pandering to the increasingly dispossessed masses and to hold on to power (and avoid a Nepal situation) they'll cut us "unwashed masses" in. Also, competition to sell to increasingly poorer people will drive some deflation at this point? I don't know - I'm not an economist and my crystal ball is broken. End stage though? I suspect the hamster wheel will continue automatically but most of the humans will have fallen off of it. But, "yeah" there is not a point to any of this. It's all made up. Money isn't real. And basically we're all playing this crazy game where we try to optimize and gamble correctly so we can stop working eventually when we could be working to intentionally try to automate and remove all drudgery and terribleness from our working lives and stop working now. The game has been broken for a long time.

I mean, there will always be stuff to do, but if we could get rid of some of the boring stuff that would be a nice start. It will likely be the end of the current economic paradigm though and a lot of people are not going to like that. It won't be some sharp Marxian transition - it'll be just like when democracy and the republic began to spread around the world (along with Capitalism) - there will be places where it's more or less prevalent until the "normal" way of things is some sort of post-scarcity UBI thing instead of "every man for himself." But all this is going to take time - probably a decade? And in the mean time I don't know about you but I have the strong urge to do better than simply avoid privation.

But yeah! Go spend time on your sailboat because that probably is the answer. That's literally my plan to all of this. Set things up the way I want them more or less now, then ride the wave into the future. If I'm not fed feet first into the paperclip machine, I'll be set even if the automation utopia never arises because I don't have any debt, I own my land, and barring massive increases in property taxes (mine are low), I'm pretty much able to sustain.
 
It’s not just the safety and regulatory aspect, it’s the common sense and experience aspect.

Just the other day I prevented a diversion by sending a flight to another corner post than filed. Years of experience told me that if they continued on their filed route with the way the radar was shaping up they would be stuck in holding for 30+ minutes and then get a significant reroute around the weather, which is exact what happened to every flight that continued on with that arrival.

I highly doubt AI will ever be able to replicate that. I am even more confident after seeing Flightkeys in action and the absolutely insane things it comes up with.
If every common sense decision you make is "logged" then it becomes part of the next model's training dataset. After a few years of this, a similar situation arises, the AI will know that the fuel burns will be higher in this situation and that the airplane won't get to the intended destination... so... yeah... it will be able to replicate that.

Hell, I bet if we had the data me and you could build out a very simple "fuel prediction" model that could do this where we would "predict" the fuel that the airplane should land with. We could probably do it in a weekend. Or, "will-divert" as a predictable value? Like, could we pull together 5,000 flights? All the historical weather data, too, and total fuel burn? I bet your shop has that data... Will it be as good as you? Probably not. But let's suppose it's 80% as good as you. So, now you can be the "supervisor" and make sure it's predictions are "good" - but now (because you're not doing all the work yourself) you can do twice as many flights. How many coworkers do you need under that paradigm?
 
I'm really curious about this.

View: https://youtu.be/IvkSabIRrKc?si=1BY-2knZPIM6p26j


An example of the attention to potential conflict analysis that we are heavily exploring on the DOD side as we try to model out the Air Littoral space between seen and unseen objects operating in integrated space.

DARPA has an entire AI system called Astarte that data mines across multiple platforms and message format types to recognize and recommend reroute of assets.
 
Ah yes, I can see it now. AI not listing an alternate for MCO in the summer because the TAF doesn’t require it. Or adding TPA as an alternate mid air when thunderstorms are approaching TPA as well. I’m starting to think those who believe AI will replace dispatchers in the US are not dispatchers at all……
 
Ah yes, I can see it now. AI not listing an alternate for MCO in the summer because the TAF doesn’t require it. Or adding TPA as an alternate mid air when thunderstorms are approaching TPA as well. I’m starting to think those who believe AI will replace dispatchers in the US are not dispatchers at all……

100%. They’re all data driven that fall into the AI hyperbole. Not saying AI doesn’t have exponential potential (hah), but sit with a dispatcher managing 30 flights during an irop and realize how many decisions they make NOT based on data alone, but experience and common sense.
 
Airline management will want AI to build routes and add alternates that are friendly to the bottom line.

Destination is forecast at 1/2 sm and fog. The alternate 10 miles away is forecast 200 and 1/2 with fog but the airports, plane and crew are CAT III qualified. Its legal by the books so management will want the AI to select it for the fuel savings.

Already with flightkeys we see an issue where the weather forecast from WSI puts in a cloud ceiling in their FPGs for the bases of thunderstorms. Flightkeys software thinks you can fly under a thunderstorm and tries to build a route under instead of around. Will AI know to route around storms or will it try to plan under storms to achieve cost savings?

Sometimes on longer haul flights, pilots request routings that are even more conservative than what a dispatcher builds to avoid weather. Does AI get programmed to just build the pilot requested route or does management want a human to try to talk the captain into taking the routing thats less costly to the bottom line?

Airline management likes to spead out alternates during diversion events. When landing weight or max takeoff weight becomes an issue, the dispatcher and captain can often find workarounds to get everyone on. Does AI work with the captain on weight issues or will AI look at weight issues and insist all the freight gets loaded instead of the non revs or the long alternate must stay to ensure alternate balancing for diversions?

How will AI deal with MELs that are by the book legal but not a good situation and pilot refusals of MELs and logbook items that maintenance says are legal to fly with? Will AI try to find a solution that works for everyone by getting everyone involved to talk it through first or will AI simply go straight to the chief pilots to try to force pilot into taking the plane?

AI will do what its programmed to do. Airlines arent going to spend billions on AI flight planning just for the savings to be eaten up by pilots having an independent thought on the best way to operate a flight.
 
If every common sense decision you make is "logged" then it becomes part of the next model's training dataset. After a few years of this, a similar situation arises, the AI will know that the fuel burns will be higher in this situation and that the airplane won't get to the intended destination... so... yeah... it will be able to replicate that.

Hell, I bet if we had the data me and you could build out a very simple "fuel prediction" model that could do this where we would "predict" the fuel that the airplane should land with. We could probably do it in a weekend. Or, "will-divert" as a predictable value? Like, could we pull together 5,000 flights? All the historical weather data, too, and total fuel burn? I bet your shop has that data... Will it be as good as you? Probably not. But let's suppose it's 80% as good as you. So, now you can be the "supervisor" and make sure it's predictions are "good" - but now (because you're not doing all the work yourself) you can do twice as many flights. How many coworkers do you need under that paradigm?
You're acting like training a model is cheap, easy, and fast. It's none of those things.
 
Ah yes, I can see it now. AI not listing an alternate for MCO in the summer because the TAF doesn’t require it. Or adding TPA as an alternate mid air when thunderstorms are approaching TPA as well. I’m starting to think those who believe AI will replace dispatchers in the US are not dispatchers at all……
That's not how AI works. What you described is already a stupid situation something like FPM could cook up. AI learns and can consider multiple sources and analytics to make a decision that fits within the regs.
 
How will AI deal with MELs that are by the book legal but not a good situation and pilot refusals of MELs and logbook items that maintenance says are legal to fly with? Will AI try to find a solution that works for everyone by getting everyone involved to talk it through first or will AI simply go straight to the chief pilots to try to force pilot into taking the plane?
Captain: “I’m not comfortable taking this no ice bird.”
AI DX: “Birds can fly through icing.”
Captain: “Ok, I want to refuse take this plane due to the MEL.”
AI DX: “I’m sorry Captain, I can’t let you do that.”
 
Sounds like you're in the wrong forum then....
I mean, I'm a pilot by trade (can't do that anymore), but I literally know how this stuff works, I studied it in grad school, it's currently my profession. I've been here on JC for like almost 2 decades now? I don't know, mostly I was responding to this:

You're acting like training a model is cheap, easy, and fast. It's none of those things.

And... legit, it's all of those things now lol. I guess how cheap and fast "depends" but... easy? Yeah, • is pretty easy. You can feed in data and fit a model to anything. Including flight planning / following / or whatever.

I'm not trying to piss in anyone's cheerios, I'm just saying, this is absolutely something that can be done, will be done, etc. The incentives are too strong. Is that going to happen overnight? No. But I'd be damn surprised if this didn't drastically change the way flights are dispatched over the next 5 years. You don't have to believe me if you don't want to, you do you, but you can model basically anything with these types of models.

Their applications are literally universal. For weirdly shaped functions there are workarounds. Whatever. But... yeah, this is a thing. Like I said, I'm not a dispatcher, I was a pilot, but I promise you that basically anything you can do with KVM (keyboard-video-mouse) will be able to be automated faster than you'd suspect. I literally automate stuff for a living right now in a different field.
 

View: https://youtu.be/IvkSabIRrKc?si=1BY-2knZPIM6p26j


An example of the attention to potential conflict analysis that we are heavily exploring on the DOD side as we try to model out the Air Littoral space between seen and unseen objects operating in integrated space.

DARPA has an entire AI system called Astarte that data mines across multiple platforms and message format types to recognize and recommend reroute of assets.

Why the hell am I not doing this lol?
 
Back
Top