Originally Posted by
DaveUnwin
If it had been flying autonomously, would it have tried to 'save' itself by trying to reach a runway?
It would have, very quickly, calculated what was possible and executed the least dangerous course of action with the overriding goal of passenger safety, it is safer to land on a river than landing in a built up city, it's all relative and that's how the 'thinking' part works. Of course it can make that calculation in milliseconds, and in the hudson incident that 'speed of thought' may well have resulted in an airfield landing . . . as was proven.
Autonomous aircraft control systems will only happen if they are (long term) cheaper than using humans, part of that cost is the cost of killing people.
Already the reality of the implications of creating truly autonomous cars is losing its gloss . . . VW, for example, recently stated that full autonomy may never happen . . . the primary reason is that there is almost no way of recouping the return on investment required. R&D prototypes are one thing, sellable product is something else.
Also ref aircraft having to be completely reliable - that's simply not the case, a machine learning based system (almost certinaly NOT learning on the job) is probably ideally suited to dealnig with all manner of aircraft failures, because it can be systematically trained to deal with everything that can fail, and if that one in a million failure happens, it's seen it before and learned how to deal with it as safely as possible.
Would I get on an autonomous aircraft today, probably not, much like I would not get in an autonomous car . . . why ? . . . the people creating these systems are not good enough . . . yet . . .
I suspect for a while the human solution, although flawed, is far more palatable than the increased costs of full autonomy. How long that remains the case will be interesting to see.