IEEE Spectrum Article: "When Will We Have Unmanned Commercial Airliners?"
Join Date: Jan 2001
Location: Dallas, TX USA
Posts: 739
Likes: 0
Received 0 Likes
on
0 Posts
Mathfox and Vapilot2004 both make good points. Computers are good at the boring and repetitious (via programming), while humans are good at the adaptive and unpredictable (via insight and intuition).
The reason A320s don't have auto throttle for ground maneuvering is because use of throttles on the ground in too unpredictable to program. The reason we don't have autos that can drive themselves from one location to another, is because the journey from source to destination is too unpredictable to program, with far too many variables. Maintaining a speed, a heading or a climb rate while airborne, is child's play compared to an auto journey.
The human mind can deal successfully with the unpredictable and the unexpected, but the computer can only deal with the unpredictable and the unexpected, to the degree that the programmers predicted and expected the unpredictable and unexpected. which will never fully happen.
It's not about hardware, it's about software and its limited decision making. No form of AI ever developed now or in the foreseeable future, can accurately handle the unpredictable and unexpected the way a human being can (when the human is properly trained).
The reason A320s don't have auto throttle for ground maneuvering is because use of throttles on the ground in too unpredictable to program. The reason we don't have autos that can drive themselves from one location to another, is because the journey from source to destination is too unpredictable to program, with far too many variables. Maintaining a speed, a heading or a climb rate while airborne, is child's play compared to an auto journey.
The human mind can deal successfully with the unpredictable and the unexpected, but the computer can only deal with the unpredictable and the unexpected, to the degree that the programmers predicted and expected the unpredictable and unexpected. which will never fully happen.
It's not about hardware, it's about software and its limited decision making. No form of AI ever developed now or in the foreseeable future, can accurately handle the unpredictable and unexpected the way a human being can (when the human is properly trained).
Join Date: Sep 2010
Location: earth
Posts: 1,341
Likes: 0
Received 0 Likes
on
0 Posts
The reason A320s don't have auto throttle for ground maneuvering is because use of throttles on the ground in too unpredictable to program. The reason we don't have autos that can drive themselves from one location to another, is because the journey from source to destination is too unpredictable to program, with far too many variables. Maintaining a speed, a heading or a climb rate while airborne, is child's play compared to an auto journey.
I don't doubt that eventually we will get to the point that AI-operated aircraft are safer and more efficient than human-piloted ones. At that point in history, though, you will be able to say the same about virtually every other human endeavour. Anything we could do would be better done by machine intelligences.
In the cockpit automation is on the rise but the human part of it is still necessary for unforeseen situations and interacting with other humans. I don't see this changing a lot in the short to medium term, very much like fusion power and strong AI were always 20 years in the future, whether the predictions were made in the 70s, 80s, 90s, etc.
This is without adding in the cost and the problem of consumer acceptance: we are still only a few steps down the road of making a reliable self driving car Google Car. Who wants to be first up in the Google Plane?
In the cockpit automation is on the rise but the human part of it is still necessary for unforeseen situations and interacting with other humans. I don't see this changing a lot in the short to medium term, very much like fusion power and strong AI were always 20 years in the future, whether the predictions were made in the 70s, 80s, 90s, etc.
This is without adding in the cost and the problem of consumer acceptance: we are still only a few steps down the road of making a reliable self driving car Google Car. Who wants to be first up in the Google Plane?
Join Date: Feb 2005
Location: flyover country USA
Age: 82
Posts: 4,579
Likes: 0
Received 0 Likes
on
0 Posts
Join Date: Oct 2009
Location: new zealand
Posts: 42
Likes: 0
Received 0 Likes
on
0 Posts
^^^ LOL
Those who think that eliminating the pilots will save space/weight, think of all the equipment that they would have to put in so the aircraft can be remotely controlled.
Those who think that eliminating the pilots will save space/weight, think of all the equipment that they would have to put in so the aircraft can be remotely controlled.
Join Date: Oct 2009
Location: The Isle Dordt
Posts: 290
Likes: 0
Received 0 Likes
on
0 Posts
Those who think that eliminating the pilots will save space/weight, think of all the equipment that they would have to put in so the aircraft can be remotely controlled.
In my experience computers can perfectly repeat boring tasks... It has been shown that computers are better at counting cells in microscopic samples than human analysts. But don't ask a computer to do anything outside its area of programming.
Worse yet, examples of engineering breakthroughs (like self driving cars) depend heavily on machine learning systems. These are effectively trainable computers that 'make up' their algorithms on the fly (a dramatic over-simplification of the technology) based on training cases which need to be selected and validated by those same engineers (often a degenerate case, in which oversights in the training models propagate into the trained agents' behaviors). Or use actual operational data which can lead to some big holes in the algorithms' state space decision paths given that nobody is going to let the computer 'play around' during a revenue flight just to fill in some parameters*.
Back in my day, software certification processes did not allow for applications that, over time, can generate new decision paths not present in the original product and therefore untested. Has the FAA changed its position on this?
*Call me when they build a flight controls computer that can log onto PPRuNe after hours and shoot the bull with real pilots to pick up handy advice from experienced practitioners.
Join Date: Oct 2009
Location: The Isle Dordt
Posts: 290
Likes: 0
Received 0 Likes
on
0 Posts
*Call me when they build a flight controls computer that can log onto PPRuNe after hours and shoot the bull with real pilots to pick up handy advice from experienced practitioners.
With drones already flying around (with incentive to improve performance) there will soon be a choice of "robotic flight controllers", all too poor for passenger transport. However, some controllers will be suitable for (small, Cessna sized) unmanned freight transport. How much domestic use will the FAA allow? That's the major question that will ultimately decide upon acceptance.