Actually, I do. It was supposedly a feature of the automatic pilot, which you were so confident could be totally relied on. It was designed to modify, through flight control, the tendency for the airplane to develop a stronger pitch up moment at high angles of attack. But it did so by (1) overriding the human pilots, and (2) imposing downward pitch angles in succession, resulting in crash dives. Boeing tried to deflect criticism of the software by blaming the pilots for incompetence---when they died pulling back on the yoke to prevent a fatal crash.
The 737 MAX program development management review team were assured by management that the MCAS was merely an "adjustment" to an existing, well-proven flight control software. Nope. It turns out that it was more than an adjustment, and was insufficiently tested for its purpose. How do I know this? I was a member of the program review team. We were fed a bill of sale.
MCAS is an autonomous system. That is the apex of artificial intelligence. It was designed to accomplish its purpose without regard to the well-being of crew, passengers, or aircraft.
You have no idea what an AI is. The MCAS system does not reason, learn, nor perform problem-solving. It's just a pitch-control system that kicks in or turns off when certain conditions are met. Those have existed since the 60's. You probably just read that one Australian news story that mislabeled the MCAS system as an AI but decided to LARP as a "member of the program review team"
There are different kinds of A.I. and the apex is autonomous behavior. Current A.I. does not reason, learn, or problem-solve. It is mostly an aggregator algorithm that can plagiarize based on verbal constraints, but no conceptualization. A.I. generated artwork, for example, will be persuasive up to the point where you notice odd details, like people having 6 fingers or 3 eyes. It is constructing a sensory montage, not governed by any conceptual framework.
But you were touting A.I. as an infallible autopilot, so you should be willing to see MCAS as a refutation. It was trying to solve the problem of preventing unwanted pitch-up. I was a member of the program review team. That's a fact. You don't gain any ground by fantasizing what I may or may not have done, and then accuse me of being a liar---without any evidence.
I have also been involved in the design of automatic space-based missile defense systems, where autonomous behavior turns out to be indispensable to system success (identification and tracking of targets, and solving an interceptor-target assignment schedule). I have done work in artificial perception (target-vs-decoy discrimination), which is preliminary to artificial concept-formation. There are many steps toward the creation of an artificial mind. A.I. isn't there.
And you have a problem of realism. The MCAS software was essentially straightforward, but it killed two aircraft loads of crew and passengers before Boeing got the message it was unsafe (after first blaming the pilots). But you are saying that something orders of magnitude more complicated is going to work perfectly?
So because bad software existed before, good software could never be made? I didn't put a timescale on that prediction, it could be 20 years, it could be in a 100. Eventually I think it will happen, and the fact that bad software has existed has nothing to do with it.
Was NCAS supposed to be "bad software"? No. The whole premise is that it was good software. The badness came out in the wash, and the problem pointed straight back to the carelessness of the software requirements formulation.
Your problem is how to assure that life-saving software will result. It is not easy. And what we thought---and were assured---was a simple, harmless change in pre-existing, reliable software, turned out not to be.
To dismiss the phenomenon of bad software eventuating from what was absolutely considered good software is a bad move on your part. Very ignorant. How will you know your software is safe? And by what means? If you have no answer, then it is not a certainty the day will come. On what day will you declare your software "good," so we can measure whether it was 20 or 100 years?
This is why testing is so important in modern aerospace technology, and why it is so expensive. The human mind is still far superior to any collection of microchips.
The murderous MCAS software on the 737 MAX was A.I. You have no idea what you are talking about.
Not all software is AI. You don't know what you're talking about either.
Actually, I do. It was supposedly a feature of the automatic pilot, which you were so confident could be totally relied on. It was designed to modify, through flight control, the tendency for the airplane to develop a stronger pitch up moment at high angles of attack. But it did so by (1) overriding the human pilots, and (2) imposing downward pitch angles in succession, resulting in crash dives. Boeing tried to deflect criticism of the software by blaming the pilots for incompetence---when they died pulling back on the yoke to prevent a fatal crash.
The 737 MAX program development management review team were assured by management that the MCAS was merely an "adjustment" to an existing, well-proven flight control software. Nope. It turns out that it was more than an adjustment, and was insufficiently tested for its purpose. How do I know this? I was a member of the program review team. We were fed a bill of sale.
MCAS is an autonomous system. That is the apex of artificial intelligence. It was designed to accomplish its purpose without regard to the well-being of crew, passengers, or aircraft.
You have no idea what an AI is. The MCAS system does not reason, learn, nor perform problem-solving. It's just a pitch-control system that kicks in or turns off when certain conditions are met. Those have existed since the 60's. You probably just read that one Australian news story that mislabeled the MCAS system as an AI but decided to LARP as a "member of the program review team"
There are different kinds of A.I. and the apex is autonomous behavior. Current A.I. does not reason, learn, or problem-solve. It is mostly an aggregator algorithm that can plagiarize based on verbal constraints, but no conceptualization. A.I. generated artwork, for example, will be persuasive up to the point where you notice odd details, like people having 6 fingers or 3 eyes. It is constructing a sensory montage, not governed by any conceptual framework.
But you were touting A.I. as an infallible autopilot, so you should be willing to see MCAS as a refutation. It was trying to solve the problem of preventing unwanted pitch-up. I was a member of the program review team. That's a fact. You don't gain any ground by fantasizing what I may or may not have done, and then accuse me of being a liar---without any evidence.
I have also been involved in the design of automatic space-based missile defense systems, where autonomous behavior turns out to be indispensable to system success (identification and tracking of targets, and solving an interceptor-target assignment schedule). I have done work in artificial perception (target-vs-decoy discrimination), which is preliminary to artificial concept-formation. There are many steps toward the creation of an artificial mind. A.I. isn't there.
And you have a problem of realism. The MCAS software was essentially straightforward, but it killed two aircraft loads of crew and passengers before Boeing got the message it was unsafe (after first blaming the pilots). But you are saying that something orders of magnitude more complicated is going to work perfectly?
So because bad software existed before, good software could never be made? I didn't put a timescale on that prediction, it could be 20 years, it could be in a 100. Eventually I think it will happen, and the fact that bad software has existed has nothing to do with it.
Was NCAS supposed to be "bad software"? No. The whole premise is that it was good software. The badness came out in the wash, and the problem pointed straight back to the carelessness of the software requirements formulation.
Your problem is how to assure that life-saving software will result. It is not easy. And what we thought---and were assured---was a simple, harmless change in pre-existing, reliable software, turned out not to be.
To dismiss the phenomenon of bad software eventuating from what was absolutely considered good software is a bad move on your part. Very ignorant. How will you know your software is safe? And by what means? If you have no answer, then it is not a certainty the day will come. On what day will you declare your software "good," so we can measure whether it was 20 or 100 years?
This is why testing is so important in modern aerospace technology, and why it is so expensive. The human mind is still far superior to any collection of microchips.