The AI factor

I was discussing the topic of pilot career paths on LinkedIn recently with a former aircraft technician who was “yelling” at me: using loads and loads of exclamation points. His take was that there would be no need for pilots in 20 years, and that I and the industry should not be encouraging people into a flying career.

It didn’t matter how moderate were my responses, along with considered facts around Boeing and Airbus estimates suggesting 790,000 more pilots will be required globally over the next 20 years – around one third in the Asia Pacific. It didn’t matter that I cited driverless cars making an entrance in 2009 yet in 2019 they are still not a commodity for general public road use, and my view that “the general public is being slow to take on driverless anything. ” He railed aggressively against any form of ‘discussion’ and then, well then, he brought Trump into it, so he lost me.

It did, however, spark more research into burgeoning AI technology and how it will change the pilot career and passenger experience.

So I went to those at the forefront and asked them: design engineers who could offer a hands-on perspective.

I went into the research with the assumption that pilotless aircraft will become the norm for some forms of flight, that the human factor will be required for at least another 20 years, and that buy-in to pilotless aircraft by passengers may well take even longer.

We already have single crew cockpits for freight transport. We already have remotely piloted military drones and military ‘zero-man/oneman’ hybrids in planning. These have a cockpit but can fly manned or unmanned depending on the mission or task required.

You will recall that the need for an engineer on every flight was made redundant as technology improved and risk factors were further improved. So this is the way we are heading. Of that there is no doubt.

To that end, my sources offered me this wonderful quotation that provided an excellent starting point for this discussion: There are seven stages of Human – Robot/AI Replacement

1. A robot cannot do my job.

2. A robot can do a lot of my job but not everything.

3. A robot can do everything except when it breaks down, which is often.

4. A robot can do everything flawlessly for now but cannot adapt to new stuff.

5. A robot can have my old boring job because it’s not a job humans would really want to do.

6. A robot can do my old job, but my new job is much better and interesting.

7. I am so glad . . . a robot cannot do my Job. (reference: Kevin Kelly, page 59, The Inevitable) In 2019 we appear to be fluctuating somewhere between points 3 and 5, depending upon the industry to which it is applied.

This is also not just about the technological advances taking over pilot careers, even if the technology is ready and tested and faultless.

Technology in and of itself is not the only deciding factor. There are multiple considerations:

Commercial operations and profitability

There is no more global an industry than running an airline. Which means management’s fiscal responsibility is to ensure that aircraft fly for maximum profit. This will require global methods of regulation to ensure AI advances in aircraft operate safely and play by the rules. On the other hand, profitability means businesses must take advantage of any area where they can reduce costs. Employing fewer flightcrew is one of these cost factors. This is a constant in an open commerce society.

Training, retraining and restructuring

With the 737 MAX we have seen, sadly (among other things), that pilot training was not in sync with new technology. There was a gap between what the manufacturers/ engineers knew and built and what the pilots knew. Our current training for engineers will decide the engineering managers of 20 years’ time. Engineering training regimes have been watered down to be more efficient, but less flexible. That implies an impact on our ability, long-term, to find suitably qualified people.

People can’t evolve faster than technology so technology will also have to adapt to us. Virtual training is here now and nowhere more so than in simulator training for pilots and cabin crew. But can simulators fully replicate extreme weather conditions or frightened passengers and their reactions? The pilot experience is close to the real thing but cannot replicate the full human emotional response to events. Truth is that in a sim no matter what goes wrong, there will be no fatal crash.

To paraphrase Chesley ‘Sully’ Sullenberger, captain of the ‘Miracle on the Hudson’ Flight 1549: While humans are the least predictable (element) of the safety system, they are also the most resilient and adaptable.

“We must ensure that humans are always in complete control of the aircraft and its flightpath. ”

Politics

Nothing provokes a government leader to change his/her dinner plans like a disaster, and a plane crash is among the most influential. Crashes are rare but as soon as one happens, no matter where in the world, every government gets asked “can it happen here”? If the answer is Yes, then a change is inevitable – either more regulation or a new design specification, or both. more efficient aircraft and is pushing for alternative fuels. An electric widebody jet is not really feasible, but what about a liquid hydrogen gas turbine that only emits water vapour.

Or a scramjet that does seven times the speed of sound with, again, water vapour as the waste product.

Passenger conceptual buy-in

More emphasis will be placed on deciding how and when the human-AI interface boundary will be pushed by profit and safety. If people don’t want to be flown by a computer alone, there is no profit in it for the airlines. This will require a long-term educationbased marketing strategy to ensure the passengers feel (not understand, but feel) that safety is not compromised by changing the division of workload between human and machine. Public opinion takes time to change.

Safety/risk management and testing

The speed of decisions, reliability and hence availability, will be pushed up by AI as it collects and collates data that will drive improvements in both reliability and risk management.

Planes will be able to fly more and be maintained less.

Aircraft already have the ability to communicate complex faults from the aircraft through the levels of aircraft management and to the design organisation. A fully realised global network of information and resource management with AI/machine-led decision-making is not far off.

AI analysis of data can improve outcomes for our emotion-based human judgements, so data will go beyond just technical events. How many times does a pilot misjudge a landing? Why do they misjudge it? And how do we prevent it from happening again? How do we design the cockpit and instruments to stop the misjudgement? Can we apply additional tech/AI to support the pilot to prevent it? We have just scratched the surface of the much larger, complex safety/ risk/human-factors/AI discussion. In fact, I did not realise the enormity of this topic until now. There is a lot to take in. All we can do is look at the way in which our world traditionally progresses, take the lessons of the past around how human beings evolve and interact with technology, apply them to our current state and make educated projections.

It will take many years to reeducate the general public to accept less human interaction (single-pilot cockpits). Cultural change is hard to predict. However, the aircraft being built today are also restrained by current laws, so evidence needs to be developed before changes are made.

If this type of change is too rapid it faces rejection as incomprehensible by the user, be that the provider of the service (airline), or user of the service (passenger). The opposite, too slow to change, is not really a problem since in a free market the competitive edge is always pushing new ideas forward.

Pilots and aircraft engineers will naturally see their roles change and they need to prepare for more regular retraining and further education as technology advances. In the end, it rests with us to accept that we must evolve along with technology to remain current and relevant in our ever-changing careers.

It is close to the real thing but cannot replicate the full human emotional response to events’