I'm sure an automated system would do great in 95% of conditions but it's that 5% where you really, really need a human in the loop to make judgement calls that would be incredibly difficult to program.
Here's a classic example of a controller noticing a pilot is hypoxic, indirectly testing his competency and ability, and likely being careful with how he routed traffic around the unreliable pilot until he got better. This alone seems pretty hard to imagine automated with current technology without some overkill prone-to-failure solution.
Sorry for being that person, but wouldn't situations like this be better resolved by replacing the pilot with full automation too?
I mean, I've seen quite a few cases where inordinate amounts of effort are expended to create automation to fit the needs of humans and legacy human-oriented interfaces, whereas redesigning the entire flow for the capabilities of modern tech can be significantly cheaper and more reliable.
I'm sure that there are a million various complications in this area that I have no idea of, but still would hope that the people in charge are looking to redesign the whole thing rather than just each piece individually.
It's not - it's much simpler. Drivers move in a 2D plane densely populated with cars and things that are not cars. Planes sparsely populate the sky by themselves. (There are birds too, but not that high up, and they're largely unavoidable in real-time anyway)
There's already semi-automated collision avoidance. If two planes are on a collision course, TCAS tells one plane to climb and another to descend. This works because outside of tightly managed scenarios like a holding stack, and planes shouldn't be anywhere near each other horizontally. There are no other planes above or below you (or anywhere nearby) and by fixing the directions as up and down, TCAS doesn't even have to calculate compass bearings that would work for avoidance.
If you were to design an entire ATC system from scratch (pilot interfaces, sensors everywhere in the airport and planes etc) it can be automated. But with pilots having to actually talk to ATC (and sometimes talk over each other with no feedback) instead of observing their status on a screen and pressing buttons on what they want to do or change their status it seems like it will be quite hopeless for quite some time.
What you can probably do is create software which observes traffic and simulates it into the future and notifies the human ATCs about risks. It might even be a good idea to try and digitize it for the ATCs so they talk less and press buttons more (which will feed into the simulation) and use TTS for the legacy transmissions to pilots that don't have an updated interface. Given the regulation on that industry it seems unlikely anyone competent enough to do it will have an interest to even try.
> If you were to design an entire ATC system from scratch (pilot interfaces, sensors everywhere in the airport and planes etc) it can be automated.
Even then you'll probably run into the long-tail distribution issues, similar to self-driving cars. 99.9% of all situations are pretty standard, but once in a while something so abstruse happens that it's not pre-programmed and requires some creativity to solve.
> What you can probably do is create software which observes traffic and simulates it into the future and notifies the human ATCs about risks.
Fully agree. Some of the recent close calls really were "obvious" much earlier, meaning they were not caused by late course changes.
Have we just forgotten that there is more to computing than LLMs?
Most big planes can land themselves now - in calm weather, at least. It's done regularly when fog is worse than a certain point, because the necessary radio signals pass through the fog. Both the airport and the plane need certain equipment. The same navigation equipment is used on many manual approaches, but with a human in the loop.
> hallucinate 10% of the time in charge of human lives
Out of curiosity, about a year ago I queried a few models about how to fly a particular instrument approach. It was an ILS approach using a DME arc transition. Other the basic concept of lateral and vertical guidance, most of the models got literally everything wrong. Wrong headings, wrong NAVAID frequencies. Wrong procedures. Maybe they’re better now in this domain, but they were confident in their claims of the ability to read an approach plate. But it was terrible.
