Will AI replace GA pilots?

I don't think the average person understands what AI is. (They often think it is some form of Gemini or ChatGPT that will magically perform complex tasks ab initio.) There are large language models, machine learning, and robotics. GA is already populated by decent robotics. Even my dumbass autopilot and GPS can fly enroute and approaches without much input from me, if I so choose. Large language models are somewhat useful for basic information retrieval tasks and some limited creative stuff, but is laughably stupid for more complex things. Machine learning has to be trained on information from a curated data set, usually for fairly well-constrained problems. ML has proven to be revolutionary for narrow knowledge areas like games (chess and go), sequence-based protein folding prediction, medical imaging analysis, etc. Flying an airplane in a milieu of weather and traffic, especially for GA, which can normally fly above much weather, is a board and complex task, much more so than for current ML successes. We haven't even succeeded in building truly capable ML-guided automobiles, another complex task, albeit simpler than flying, and one with actual training data. One part of ML training that is unacceptable for things like planes and trains and automobiles with people in or around them is that we can't allow them to fail. And the way ML works is it has fail and note when that happens.

This is not to say that automation hasn't increased safety in GA operations. It definitely has. But I think the general public has an unrealistic notion of what artificial intelligence is and what it can do right now.
 
MackMooney said:
hmm, I don't think flying is really more difficult than driving.
Click to expand...
Hard to explain that if flying is as easy as driving a car why is the accident rate by mile 5-10 times as high with a higher barrier to entry, more training required for the vehicle and the pilot, an above average segment of society in productivity, more regulation and monitoring and with all that people augering in at too high a rate? Reality is that controlling a vehicle in 3 dimensions at much higher speeds in much more varied conditions is far harder than controlling a car at slow speeds only having to control 2 dimensions.
 
Ed Haywood said:
Could AI make GA flying simple enough that real pilots are no longer necessary?
Click to expand...

I'm not sure that AI specifically changes this much, but technology broadly? Yes, I think so, though not in my flying lifetime (I'm in my mid-50s). I think we will get there eventually. We haven't completely solved this problem for driving, though it's getting closer, and flying is obviously that much more complicated, so it'll be a little while. The first automation-based flight that kills someone (there will be one, I'm sure) will set the development back years, as it should. But it'll get there.

I don't think this is a problem that AI specifically has a lot to bring to though.
 
Unlikely that AI will replace GA pilots as piloting is what it is all about. No different than AI replacing recreational skiers on a mountain, surfers in CA or HI, etc., etc., etc.,

It is going to replace the commercial positions. One example will be that all the VTOL aircraft being used as air taxis, etc., will all be controlled/guided (e.g. "piloted") by AI. Eventually I think they will also control all the commercial flights, but I think that is a LONG way off. Both because AI isn't capable of doing it right now and because it will be a LONG time before the majority of people would buy a ticket on and AI "piloted" aircraft.
 
Rockymountain said:
Hard to explain that if flying is as easy as driving a car why is the accident rate by mile 5-10 times as high with a higher barrier to entry, more training required for the vehicle and the pilot, an above average segment of society in productivity, more regulation and monitoring and with all that people augering in at too high a rate? Reality is that controlling a vehicle in 3 dimensions at much higher speeds in much more varied conditions is far harder than controlling a car at slow speeds only having to control 2 dimensions.
Click to expand...
hmm, no clue but i would estimate the average driver has more experience and in worse conditions than anything we face flying. heck, even the guy with 16000 hours flying probably has 100X as much experience driving.

personally, i just don't find it that difficult.
 
Last edited:
"Skiiing" (the mountain, not the 3am craigslist ad circa 2015, and I plead the 5th) is a pluralistic enough activity with a wide base of monetary/revenue support, flying recreationally is not. The threat of obsolecence isn't the recreational use case directly, it's the commercial one by proxy. IOW, once they decree it isn't in the commercial interest to allow you to handfly in the same NAS the commercial space has already gone full HAL in, your hobby goes bye bye. A form of emminent domain if you may, minus the payout in a gun/real estate confiscation scheme, to add insult to injury.

Europe is functionally there already, and they didn't even need the automation piece. They just used rationalizations around distances and population density to decree it's not in their cultural interest for the masses to have access to recreational flying. I enjoy collectivist societies, but protections for minority activities is admitedly not their strong suit.

The US has a lot more critical density to go before we get there. Culturally we're too individualistic to get directly cordoned off, but by commercial interest absolutely we'll eventually get corralled away from the Bravos, and that will be that.

The good news is that US airlines are cheap bastids, so they won't eat the costs of ab initio/MPL American "pilot 2.0" unless they absolutely have to. As such, our horse and buggy affair has a few decades left in the tank. What has posterity done for me lately, amma right boomers? Don't answer that! :biggrin:
 
chemgeek said:
I don't think the average person understands what AI is most things on the internet they read or comment on
Click to expand...
FIFY :cool:

But seriously........
No, Skynet is not (probably) going to take over. AI and automation will help, but IMHO we'll always have 2 pilots in a commercial airliner, we'll have a pilot in GA aircraft, etc. We will have a human in the control loop.

Why? The world offers unpredictable issues that a human needs to be there to make decisions. Intuition, hunches, etc. need to come into play at times. As well, somethings just can't be solved, such as the Salesman problem. Somethings require a moral decision - do I land my disabled aircraft on a road and possible hit a car or try to stretch the glide to land on the golf course? ( which I guess you could decide that golfers are expendable??)
 
You must log in or register to reply here.
Back
Top