Artificial intelligence (AI) has succeeded in one thing if nothing else: dominating the tech headlines. If what you want is a regurgitation of factual-sounding information not so much strained through critical review as aggregated from every source (accurate as well as questionable), then AI may suffice for basic information gathering. But it’s also amazingly dumb in that it can’t seem to differentiate the published works of an actual expert from late-night internet flatulence of a prolific crank.
No, I’m not wearing a tinfoil hat at the moment but, yes, I do think that AI will come for flesh-and-blood writers like me. What I do—research from credible sources things I don’t know, combine that with what I do know and mix it up with an ability to look forward based on my experience—hasn’t yet appeared in AI. Though I have no doubt this column will get absorbed, Borg-like, and countered by someone saying AI is the best thing to happen to aviation. Zero sum me, AI, I triple-dog-dare you.
Examples?
Just for fun, we ran a few prompts through ChatGPT asking for illustrations of various kinds of airplanes, from fast 300-mph homebuilts to utility airplanes.
Some were rather amusing, including a bubble-topped twinjet with massive turbofans and wings that would hold about enough fuel to get the engines started and the airplane driven off the chocks. To say nothing of how you’d keep it from turning into one of those spinning fireworks should one of those Airbus-sized engines call it quits before the other one. Hey, at least the tail is stubby and the rudder small.
I especially like the four-seat twin turboprop with two seats out in the wind behind the main canopy. That’ll keep the mother-in-law quiet, eh? How about the piston single with a couple of small turbojets under the wing? Reminds me a little of the B-36 or the C-123, which had booster jets to help the big piston radials. Proof again there’s nothing new. Oh, there’s a two-seater in the bunch that could be a trainer—interesting mainly for its mid-wing, strut-braced architecture with one of the struts, um, going nowhere.
Then you have the utility airplanes, one of which looks like a steampunk nightmare after a particularly saucy who-can-eat-the-hottest-chiles competition. It has a fourth landing gear hanging off the side—perhaps there’s another outdoor seat we can’t see in the illustration? Or just junk in the data?
Yes, yes—I know. These are just fanciful renderings created by simple prompts, a crazy quilt of airplane-like shapes brought together for our amusement. No serious airplane designer would go to ChatGPT to start work.
What amuses me is the chatter around such illustrations. You’ll see comments to the effect of “normal airplanes are boring. Engine in the front, tail at the back, people in the middle. Blah!”
Well, friends, because that’s what works. It’s fair to say general aviation moves on settled science. We know what works efficiently and safely when the human is the only thing in the control loop. We know that every design is a compromise and that the engine-wing-people-tail configuration manages the compromises most effectively. Do the bots think we’re boring? Who cares?
What Data Are You Looking At?
I’ll give you another example of where AI will likely fall down, at least in the admittedly fast-moving state of the art. That’s our annual buyer’s guide. I have been asked off and on over the years why we don’t just do an “easy” scrape of website data, condense and collate the results and then drop them into our database. Bob’s your uncle.
We don’t because it doesn’t work. The leading companies are pretty good about updating their websites and keeping the models and prices current. Most others are not as good and there are many companies for whom updating their website is among their least favorite tasks. If you go purely by what’s on the web, you’ll get inaccurate, incomplete results presented as gospel.
We are different because we make actual contact with each manufacturer and in the process learn a lot—who is looking to add models but isn’t ready to announce, who is likely to drop a model or two but not yet and which companies are actively for sale. These are things I’d want to know as a potential builder—things you’ll never learn by reading the majority of the company websites as gathered by AI.
This hands-on expertise extends to all of our regular contributors who have built airplanes, designed airplanes, repaired multiple aircraft and have been able to build parts and tooling from scratch. These are not skills you get by reading every article ever written on the subject then regurgitating the results.
Humans learn by doing and the best teachers understand when to inject their own experience into the lesson plan. They know what it’s like not to know. They have gone down that road of discovery. They can synthesize a future outlook based on cycles and trends they’ve seen over time. Most important of all, the people whose bylines you see on these pages have well-calibrated BS meters.
Why is that important? Because you’re building something that could hurt you. Gravity doesn’t care where you got your information, so you better make sure it’s right. The humans here care.
Marc Cook I’m sure is a great Aviation Writer but on the subject, he could not be more wrong of the contention that AI cannot replace Aviation Experts in the design of aircraft. AI has the Internet world at it’s fingertips and the ability to become an Aeronautical Engineer in weeks. A recent design session with Ara., an Assistant to Grok3 had it “spec-ing” parts out of the Aircraft Spruce catalog. It is early in AI development so there is no recording capability, so I used my own recorder to capture the entire narrative. An aircraft could easily be built from that recording after first having verified everything with an Aeronautical Engineer.
It won’t be long until the big aircraft manufaqcturers turn over their Katia systems to Ai. No more employees lifting their legs up in bathroom stalls to avoid scrutiny, while reading a newspaper.
In Letters to the Editor in 1979, I wrote a piecece Kitplanes titled “Future Schlock” where I lambasted general aviation small airplanr manufacturers for having no VISION and (essentialy) leadership as we headed into the future.
The manufacturers never did “get with the program” and as a consequence, Homebuilding with Burt Rutan and to a smaller degree, Jim Bede led the manufacturers all the way to today.
Blackfly, by Opener was one of the really new innovations (a forehead slapper for Aeronautical Engineers in fact) besides the model airplane jet engine which is indeed incredible. Also, let’s not forget Michel Colomban’s CRICRI which needs more emulation for its performance and efficiency. None of this progress would be possible without Aircraft Spruce and its competitors. Thank You ALL!