Airbus + Cathay working on Single Pilot during Cruise with A350

[ex-Dispatcher]

Automated systems have a long way to go

I'm not a pilot - and so can't comment on the issues you see on the flight deck... I am a PNT technologist (was an FAA flight dispatcher in Tristar era). Personally I still think automation has a lot of issues to overcome - My interest is in navigation systems for autonomous systems - and just that aspect is not straightforward to implement safely. The trouble that most developers forget about comes from a complex environment that tends to throw "surprises" or extremes in condition at the system. Much testing of autonomous vehicles for instance is carried out in the benign environment of California - many millions of miles of road-testing is carried out without a hitch.... but as soon as the European weather environment is introduced, as soon as heavy traffic is encountered (or areas of high multi-path in an urban envrionment that compromise GNSS and hence integrity..... well you get the picture. And whilst many technologists believe that its easier to fully automate flight because of what they perceive as the more controlled environment, this thinking often precludes weather situations en-route (always used to exercise me as a dispatcher, no two days in Europe were ever the same).
One of my observations from ground view is that many automated systems fail to behave properly when an unusual situation is encountered - mundane and "simple" systems such as automated taps, supermarket checkouts...
And from a complex system point of view, once bugs in software are documented, it is becoming increasingly difficult to locate and solve them. Perhaps I'm an old fashioned dispatcher, but I like flying with skilled and knowledgeable pilots who have experience of many of those "boundary" conditions that have the potential to confuse an automated system.

edit: MIssed out an important point I wanted to make (my own opinion based on above) - often there is an assumption that a fully automated set of systems can become fully autonomous. That isn't true -we have issues with even fully automating systems today - but even if you get there, how do you bring all of that automation together so that it can perform a complex mission. Some intelligence needs to be in control of all the systems, interpret all of the system outputs and then act on them - in aviation to modify the flight - so that it arrives at destination safely. That is one big jump from where we are today - And how do you programme that intelligent module to carry out a commercial flight from start to finish safely - what would be the mission description (think of the consequences of a poorly scoped wish on a simple genie)? In aviation today we have those "intelligent" modules- the pilots. When we have developed a robot that can fly an aircraft like a pilot and who understands all the mission complexities and how to react with all the flight stakeholders - that is the say we may be closer to fully autonomous aircraft - personally I think there is a long journey to go to get there..

For issues surrounding the development of AI - I would recommend listening to this year's BBC Reith lectures on AI given by Professor Stuart Russell, a British computer scientist at University of California, Berkeley - 4 excellent and interesting lectures on some of the most fundamental issues and challenges with AI - very relevant to this discussion I'd suggest...