Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
A cargo plane flew 50 miles with no pilot onboard using a semi-automated system (businessinsider.com)
26 points by elorant on Dec 25, 2023 | hide | past | favorite | 45 comments


> "If you tell it to do nothing else, or if you lose communications with it, it's going to do the last thing you told it to do, which is the definition of autonomy. It has no direct human control."

That’s not at all what I thought autonomy means.


I agree, and I'm a researcher in the field. "Autonomy" implies decision making in the absence of external control/input. To me, that's very nearly the opposite of the quote.


I guess it depends on the complexity of "tell" and what meaning it could decode from the telling. If "deliver the cargo safely over there" meant that it could, in inclement weather, decide "safely" was impossible and then negotiate with ground control to land at a new airport, that's fairly autonomous.


It makes more sense if you think about it as, "it's in the last state that you put it in, and from here its state is up to it.".


This implies that a brick on a gas pedal is autonomous driving.

It might be technically correct, but hardly true to the intent.


The article is about "semi-autonomous", not "autonomous" driving. And looking up the history of driving automation, where "semi-autonomous" was a car driving on a track -- yeah, sounds about right and fully within intent's limits.


Yeah, I guess my automatic wrist watch has gotten autonomous.


Words are weird, aren't they.


Does autonomy imply sentience? Because a trained hawk can act autonomously but isn’t going to do anything very interesting without instructions.


Planes have been able to take off, fly to a destination on a flight path, land automatically, since I was in High School. I'm 64.


I was about to post a snarky reply, but after actually reading TFA, it's not clear that this plane does anything different from a modern-ish commercial aviation autopilot except for being able to set it remotely. Still waiting for "real" autonomous flight meaning path planning.


The Garmin Autonomi system can do a limited form of real autonomous flight including path planning. It's mainly intended to automatically land the aircraft at the nearest safe airfield in case the human pilot becomes disabled.

https://www.garmin.com/en-US/blog/aviation/garmin-autoland-a...


Hahaha so true. Flying things aren't more abundant and prevalent because there must be insane lobby machinery and regulatory capture around the aviation industry. Plus economics of course.

For example, Zipline[1] is innovating in a part of the world, which wouldn't be possible elsewhere.

[1] https://www.flyzipline.com/


The factor limiting prevalence of flying things is that the biggest cost tends to be energy, and flight is pretty energy intensive. So it only makes sense when the travel time is so much more valuable than the energy cost of the transportation.

Plus, autonomously taking off/landing a large plane at designated locations where all other activity is carefully controlled, is much easier than having them fly and land in unpredictable environments such as at people's doors.


I mean theres probably not a worse case scenario for Ai vehicles than a plane crashing in a city. Car crashes are nothing in comparison. I don't think the regulation here is misplaced.


Definitely then, go automatic. Better record than human pilots?


Pilots' union I suppose.


While this sounds like a cool idea, my first thought is: What happens when someone is able to remotely gain control?. Does the plane not accept commands other than "Please fly to point X. If you can't contact me again, land at Y." or can you send more complex instructions like "Fly at this altitude and heading"? If it's the latter, that is basically "joystick level" control but with more steps.

There will always be vulnerabilities in hardware and software and when they used the term "encrypted satellite", I worry that there is vulnerability in there somewhere. Even something as simple as a phishing attack of pilots could lead to dangerous outcomes.

Nevertheless, I'm sure there will be a Defcon village for it in the next few years.


> What happens when someone is able to remotely gain control?

Commercial jets are already flown by wire and networked. There is an added risk going autonomous. But it's a difference of degree more than kind. (The next difference in kind is fully-autonomous flight, i.e. no pilot on board.)


This is really interesting, as it could greatly decrease the human risk of using lighter, cheaper aircraft for cargo. Systems always optimize to their incentives, so we may see a fairly rapid shift to this kind of thing for certain routes.


Yes, but safety will always a concern.

The only way to get this to work is first that government needs to signal that this is approved use of technology and airspace.

If that happens then capital markets will follow by investing in cargo-only airports to take civilian danger out of the equation.


A Boeing B-17 has landed on its own before, without a pilot, in 1944.

https://news.ycombinator.com/item?id=28657071


Wow, that's an incredible story. What are the odds?


> address the pilot shortage.

There is no pilot shortage (just like there is no stem shortage). Just a shortage of people willing to work for poverty wages.


This could be drawn out to almost career. There isn’t an X shortage, just a shortage of people willing to work for Y that the market will bear.

Automation really is the only way to increased human productivity. The more machines we have doing things, the less things the humans need to do, simple. The only struggle will be ensuring that all the profit for robot labor doesn’t wind up in just the pockets of a few.


Commercial pilots are actually paid incredibly well. The reason for the shortage seems complex and has to do at least partly with regulations passed that mandate 1500 hours of flight time before one can even apply for a job at an airline.


And also that militaries are training far fewer pilots with the advent/expectation of future wars being primary fought with drones and missiles rather than fighters.


Average entry level pay for pilots in the US is $50K. That’s not only not incredibly well, it’s at best median for most jobs that require formal training and certification.


Also, given the automation already present, it’s a stressful job: “pay attention to these computers for 8 hours just in case something happens.”


ngl pilots are going to be automated soon for commercial people/cargo for domestic origin/destination. Unlike a truck that has to drive on roads, planes have a large room to make errors. In the future there won't be any pilots or even flight attendants and there would be a designated person on there to land the plane if needed at a much lower salary.


It can be said that no one survived the landing.


That's pretty cool, but I can see some ways this could go wrong. A hostile nation or well-funded terrorist group (sometimes those are the same thing) could hijack a plane without even being on board.

I hope they work hard to make it secure. It'll be a while before humans are willing to get on board one of these planes.


That's a bit of a strange thing to focus on, I don't see how having a pilot on board would change any of that.


Having a pilot on board means someone can fight back.


I think GP meant they could hack it remotely. If an attacker hacks a fly-by-wire plane remotely, I'd rather not be onboard.


Plenty of planes today are fly-by-wire, it's just that this one also happens to be remotely controllable.


Or, the whole fleet by damaging or interfering with the satellite links.


You may not want to research how many airliners are fly by wire...


Given that the ratio of passenger to cargo flights is roughly 10:1, it's not clear to me how this could ever really address a pilot shortage in a big way?


It’s actually 6:1. You would also have some more flexibility with autonomous cargo planes, possibly some better suited designs for the aircraft if it didn’t have to support people.

https://www.quora.com/What-is-the-ratio-of-passenger-to-carg....


Start with cargo, move to passenger flights once the automation has accumulated thousands of successful flights and millions of hours.


Also install in commercial passenger aircraft and enable to be pilot of last resort if onboard human pilots become incapacitated and a cryptographically verified command is sent to the aircraft (airline? ATC? TBD who should have this authority).

Very similar to Cirrus Safe Return Autoland, but with additional safeguards (remote command from an approved authority) vs passenger command.

> Once activated by the touch of a button, Safe Return assumes control of the aircraft and transforms the Vision Jet into an autonomous vehicle that navigates to the nearest suitable airport for landing, communicates with air traffic control, lands and brings the aircraft safely to a complete stop, allowing passengers to exit the aircraft.

https://cirrusaircraft.com/story/faa-certifies-cirrus-vision...

https://youtu.be/PiGkzgfR_c0


> move to passenger flights once the automation has accumulated thousands of successful flights

In the interim, you can reduce the number of pilots to one on non-ETOPS flights as well as their required minimum flight hours.


The fewer pilots flying cargo means the more pilots to fly passengers.


Your license include a flight test on Flight Simulator 2000




Consider applying for YC's Fall 2025 batch! Applications are open till Aug 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: