🤖 A century of autopilot
Where the copilot disappears, the last second remains
Autopilot started in 1912. Today autopilot flies 95% of major-airline operations. But the remaining 5% — the abnormal 5%, the most dangerous 5% — that's where humans stay.
📡 100 years of cockpit automation
AI didn't appear suddenly. Automation is as old as flight itself.
| Year | Tech | Meaning |
|---|---|---|
| 1912 | Sperry gyro autopilot | First attitude hold — pilots could let go. |
| 1947 | Auto Atlantic crossing (C-54) | Takeoff, cruise, landing — first demo without human input. |
| 1965 | CAT IIIa autoland (Trident) | Landing in 200m visibility. Foggy airports stay open. |
| 1988 | A320 fly-by-wire | Yoke → computer → control surface. Pilots input "intent." |
| 2020 | Garmin Autoland (single-engine) | One-button autoland. Passenger presses if pilot incapacitated. |
| 2023 | FAA RCO (Reduced Crew Ops) study begins | Two pilots → one. Evaluation underway. Possible deployment after 2030. |
| 2025+ | eVTOL autonomous urban mobility | Joby/Archer/Wisk certifying. Five-seat air taxis without pilots. |
🤖 Where pilots have already been replaced
The most effective automation is the kind no one notices.
Navigator (gone by 1980)
INS and GPS replaced star sights. An entire crew position vanished. Pilots protested; time decided.
Flight engineer (gone by 1990s)
FADEC and EICAS absorbed the third seat. Boeing 757/767 onward — two-pilot crews.
CAT III foggy landing
Down to 0m visibility, automatic. Better than humans. Humans monitor and decide ABORT.
🤝 Where pilots remain — the abnormal 5%
Where automation is weakest. 99% of the time the autopilot is safer — but when 1% breaks, no one but a human can answer.
2009 US Airways 1549 (Sully) — 208 seconds
Bird strike, both engines out. Autopilot off. The computer recommended LaGuardia and Teterboro — Sully refused both and chose the Hudson. 208 seconds. 155 souls. All survived. No algorithm had simulated that water landing.
2010 Qantas 32 (A380 engine explosion)
Engine 2 disintegrated. Over 50 systems failed. EICAS spit out 120 messages. Five pilots sorted, judged, recovered for two hours — 469 souls survived. Sorting priority under alarm bombardment is human work.
2018 Southwest 1380 (engine debris pierced cabin)
Captain Tammie Jo Shults — Navy background. One engine, decompression, one passenger killed. A combination automation cannot handle. Emergency landing in Philadelphia in nine minutes.
⚖️ The Single-Pilot Operations (SPO) debate
2024 EASA and FAA both evaluating. Industry argues cost; pilot unions argue safety.
For (Airbus · Cathay trial)
Cost: 30% airline crew cost reduction
Tech: Ground-based remote backup pilot proposed
History: Navigator and engineer disappeared on the same logic
Against (ALPA · ECA · IFALPA)
Incapacitation: One pilot loses consciousness = unmanned aircraft
Ground link: Comm loss or jamming = no backup
Psychology: 13 hours of solo responsibility = exceeds human limits
Germanwings 9525: Solo invites intentional risk
FAA's current position: "Two-crew until equivalent or better safety is proven." But cargo, sectors, and new airframes may see staged deployment. 2032–2035 is the inflection point.
💼 What this means for your career
📈 Next 25 years: Shortage + age-65 retirement + post-pandemic early retirements = major-airline hiring boom. Automation displaces seats slower than retirements clear them.
⚠️ At-risk segments: Regional single-pilot freight → autonomous singles → multi-engine SPO. In a 30-year career your seat is redefined at least once.
🎯 Hedge: Multi-engine IFR PIC + simulator type ratings + (where possible) test-pilot or instructor credentials. Go deeper into what automation can't do.