Machines or Pilots?

I don’t know how genuine the featured Ryanair advert is – it could well be “fake news” – but it does fit the cost-cutting image of Michael O’Leary! He has long advocated that he wanted single-pilot airliners by 2020, as well as advocating “stand-up” cabins (no seats) and passengers paying to use the toilets. A lot of this is usually dismissed as publicity stunts, but his recent stand-off with pilot unions clearly indicates that a no-pilot airline would be his ideal. It is quite enlightening to hear him contradicting himself about the importance of pilots: https://youtu.be/fSmnHbGMMok

He is not alone.

Research into autonomous flight is well underway. Boeing’s research has been underway since 2013 and Airbus has been working with French aerospace research company Onera and avionics manufacturer Thales for a number of years now.

The financial impetus to move to pilotless aircraft is immense – analysts see a potential profit opportunity worth about $35 billion for the aviation and aerospace manufacturing sectors. Read more here: http://aviationweek.com/commercial-aviation/pilotless-commercial-aircraft-follow-money

Artificial Intelligence has already surpassed human ability – just research Google’s AlphaZero programme, which defeated Stockfish 8, a programme capable of calculating 70 million chess positions per second. AlphaZero, though, was not programmed or taught any chess at all. It taught itself to play chess in just four hours and then beat the 2016 World Champion (Stockfish 8) on the 7th December 2017. Yes – machines have long surpassed humans at playing chess, a game which for centuries was considered one of the greatest achievements of human intelligence.

We as pilots are inevitably quick to point out that no machine can match our intuition achieved through training and experience – forgetting that “intuition” is only pattern matching, a skill which algorithms have proved to be much faster and more consistent at.

We’re also very aware of passengers preferring two (or more) pilots in the cockpit – but younger people and new generations will be quicker on the uptake and may accept a no-pilot environment more easily. Just think of the many cockpit jobs which have become redundant over the past few decades – radio operators, navigators, flight engineers – all once essential. And now? Could pilots be next?

While I submit that we are inevitably on the road to autonomous aircraft, it will in all probability not happen very soon.

For instance, Thales is working on “genetic algorithms,” the fittest of which survive. Several variants of an algorithm are combined. The resulting second-generation algorithms are compared by having them solve a problem. Those most effective are kept and combined to create a third generation and so on. So the ninth or tenth generation will be significantly better than the first. The problem, the way I understand it, is that the algorithms’ solutions become unpredictable – not wrong, but unpredictable. And in aviation we need predictability – you cannot certify a component with unpredictable characteristics. We need to understand “why” the algorithm arrived at a particular solution.

Boeing has just learnt this lesson with the MCAS system on the B737-Max. While the designing engineers thought they understood the “hidden” system and its implications, the human-machine interface failed because of the failure to include it in pilots’ conversion training. The pilots were simply not expecting the excessive nose-down trim demanded by the system, due to the FMC being confused by a faulty AOA sensor. And here we’re not even in autonomous territory yet, because the pilot can still disengage the electric trim and revert to good old manual flight.

So while the MCAS of the Max can still be understood and be “predictable” – once we hand over complete control to the aircraft, it has to be completely predictable.

That “predictability” appears to be some way off – “If we cannot explain what the system does, we will have a hard time obtaining a certification,” said Virginie Wiels, Onera’s director of information processing and systems. Significant progress can be expected by 2021, according to Marko Erman, Thales’ chief technical officer. But Wiels does not foresee any application on the flight deck in the next 10 years.”

That would imply that the possibility exists to have pilotless aircraft airborne by 2030!

So O’Leary may have to wait a few years to see his dreams come true – until then, I’m afraid we need to understand the immense pressures pilots experience in a very demanding environment – one where their crucial role is progressively being denigrated and misunderstood.

 

 

 

 

 

Advertisements