Part1 : AI in AVIATION

“How AI will transform Aviation?

 

My answer should scare any pilot or DGCA bureaucrat with intellectual honesty.

Why? Because my definition of “experienced pilot” does not match the current industry definition.

Our over-reliance on AI is our worst enemy.

Using speed as one example out of many:

For example: knowing the speed, relative to stall speed, by the sound of the airflow, by the sensitivity of the controls, by how bank responds to a rudder input, by what pilot inputs to controls produced what frequency of harmonic outputs. Pilots learned these, and perhaps others of which they were consciously unaware, because of hours of manual (not automated) flying. They could close eyes, and could sense whether they were too slow, about right, or too fast. These are cortex (thinking brain) processes.

 

Much more importantly, if speed should decrease toward a stall, or increase to an overspeed, that can be a “startle event”. In a startle, the human mind does not use its cortex. In a startle, the mind is taken over by the amygdala: the infamous “amygdala hijack”.

 

What does the startled pilot do, to save his airplane? NOT what the cortex has been trained to do. Nope, sorry. The startled pilot reacts with a thought, automatic startle reflex from the amygdala.

 

If the pilot’s training includes lots of startle training, then his reaction is good.

 

If the pilot’s training does not include lots of startle training, his reaction is bad. Incorrect. Unpredictable. Irrational. Unsafe. Deadly.

 

How might a pilot obtain lots of startle training? Two methods:

 

Years of flying and surviving manual (no automation) flying like that in defense industry. Experiencing danger, and saving himself … unintentionally teaching his amygdala how to survive a startle event. Pilots used to get years of that hard-earned experience. Used to; not so much any more.

Or, repeated subjection to startle events in a modern simulator. No, not mild surprise. Totally unexpected startles which are bad enough to elicit the startle response.

Newly hired airline pilots have neither. And, they don’t obtain good experience once hired. They use (often by airline policy) AutoPilot, AutoThrottles, and Flight Director 99% of the time — almost zero fully manual flight. Current and future pilots’ skills never get as good as previous generations’. They don’t know what they don’t know. Their incidents and accidents seem inexplicable.

Over-reliance on automation has given us not so skilled aviators, but systems managers. When the automation fails, the pilots will not have true “experience”, and they will too often fail.

All is not lost. Until perfect automation arrives (it never will), we should change current pilot training so as to have A SAFER PILOT — HUMAN FACTORS STARTLE TRAINING. It’s a lengthy iBook in the Apple Book Store, for iPads. Until we train that way, we need to worry.