Originally Posted by
Ian W
all automation systems currently have a predicate that in the 'otherwise case' or if things get difficult, the automation can drop out and give the aircraft to the pilot.
As soon as pilots start to say 'we cannot cope in manual flight; we cannot switch off systems that are in error - even though the switches have been there for decades as have the NNC for the failure and we were specifically told of the issues..... THEN money will be spent on automation that does not hand back to a pilot but manages itself even in cases of unknown error.
I’m not sure if you are misinterpreting my point. I am concerned about any automation system that might not do what it is supposed to, for example, it might refuse to drop out and hand over when it should! Or, as an extreme example, a system that has gone berserk because it is “stuck in a loop” and continually drives a flight surface to its extreme position, or (as happened on one of the “faster, cheaper, better” Mars missions) turns the engine off because it decided the lander leg had touched the surface when in fact it was still way up high – I bet that was Level C software!
I think pilots will always be needed, real AI is still many decades away. And pilots should have the option to turn a system off, BUT the aircraft safety case (with Human Factor considerations) must show in detail how they will then cope. Perhaps, extra, specific to emergency, information needs to appear to assist them, also what warnings/alarms will be muted – who needs an 11 minute stick shaker!!