Robots Win
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
Ozy
You are absolutely wrong.
Post the Sioux City crash NASA built a system called IFC.
In short it is a neural network that learns to compensate for damage and work around it.
It does not require programming of each mode. It learns and reacts.
Scientific American article: crippled not crashed [Archive] - PPRuNe Forums
That was 20 years ago. We are a long way past that.
My point stands.
If you fly in an Airbus or any modern airliner you are relying on at least some of the computers working or you are dead.
I don't have any problem with this. There are lots of them and multiple redundancies.
Exactly like in a future autonomous airliner.
My point is that relying on computers is not some new paradigm. We have been doing it for years.
There will be issues with unmanned vehicles, but relying on computers is a hurdle we have long passed.
You are absolutely wrong.
Post the Sioux City crash NASA built a system called IFC.
In short it is a neural network that learns to compensate for damage and work around it.
It does not require programming of each mode. It learns and reacts.
Scientific American article: crippled not crashed [Archive] - PPRuNe Forums
That was 20 years ago. We are a long way past that.
My point stands.
If you fly in an Airbus or any modern airliner you are relying on at least some of the computers working or you are dead.
I don't have any problem with this. There are lots of them and multiple redundancies.
Exactly like in a future autonomous airliner.
My point is that relying on computers is not some new paradigm. We have been doing it for years.
There will be issues with unmanned vehicles, but relying on computers is a hurdle we have long passed.
Join Date: Feb 2006
Location: Hanging off the end of a thread
Posts: 33,071
Received 2,939 Likes
on
1,252 Posts
At what, exactly
Please add content.
Please add content.
Last edited by NutLoose; 19th Apr 2015 at 19:50.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
Nutloose
I think you labouring under a misapprehension.
Do you think that these new vehicles are radio controlled?
The word autonomous is there for a reason with the new toys.
Re the DC10
That is the whole point of the article I just posted the link to. They came up with a way 20 years ago. It learns to fly using what controls it has left. Currently it then gives that control back to the pilot in a relatively invisible manner, but I can just as easily do it with an autopilot.
I believe a development of it is now standard on F15 and C17?
I think you labouring under a misapprehension.
Do you think that these new vehicles are radio controlled?
The word autonomous is there for a reason with the new toys.
Re the DC10
That is the whole point of the article I just posted the link to. They came up with a way 20 years ago. It learns to fly using what controls it has left. Currently it then gives that control back to the pilot in a relatively invisible manner, but I can just as easily do it with an autopilot.
I believe a development of it is now standard on F15 and C17?
Guest
Posts: n/a
Tourist,
Already duly chastened as being the old fuddy-duddy that I undoubtedly am, I
must say that I am with you when you say:
"....we have all actually come to trust FADEC because actually they are astonishingly better than humans at their job". No one would dispute that.
And before that:
".....the FADEC computers failing will kill you and there is no back-up for that on any aircraft I know about".
An old story (from when the pilotless airliner was first envisaged):
Robot Jumbo takes off faultlessly and climbs away. Recorded cabin annoucement comes on to reassure the fearful that this carry-on is entirely safe: "You can rest assured that all the automatic systems on this aircraft have been triplicated..... there is no possible way in which they can all fail....all fail....all fail...all fail....Sqwark !....(silence)"
Danny.
Already duly chastened as being the old fuddy-duddy that I undoubtedly am, I
must say that I am with you when you say:
"....we have all actually come to trust FADEC because actually they are astonishingly better than humans at their job". No one would dispute that.
And before that:
".....the FADEC computers failing will kill you and there is no back-up for that on any aircraft I know about".
An old story (from when the pilotless airliner was first envisaged):
Robot Jumbo takes off faultlessly and climbs away. Recorded cabin annoucement comes on to reassure the fearful that this carry-on is entirely safe: "You can rest assured that all the automatic systems on this aircraft have been triplicated..... there is no possible way in which they can all fail....all fail....all fail...all fail....Sqwark !....(silence)"
Danny.
Tourist
I too agree with you.
The difficulty I have is that a couple of months ago, when discussing such safety critical systems in an aircraft you have flown in, you stated you "fully supported" MoD senior staffs who ruled it unnecessary for these systems to work or be integrated properly; which led directly to the deaths of aircrew. Have you changed your mind?
Doesn't matter what clever things you do with the flight controls, the FADEC computers failing will kill you and there is no back-up for that on any aircraft I know about.
the FADEC computers failing will kill you and there is no back-up for that on any aircraft I know about.
The difficulty I have is that a couple of months ago, when discussing such safety critical systems in an aircraft you have flown in, you stated you "fully supported" MoD senior staffs who ruled it unnecessary for these systems to work or be integrated properly; which led directly to the deaths of aircrew. Have you changed your mind?
Last edited by tucumseh; 20th Apr 2015 at 06:16.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
Tuc
As usual, you are messing around with words to make people say what you want them to say.
There is a big difference between declaring it unnecessary for a system to work, and declaring it unnecessary to test a system in the manner which you happen to want it tested.
ie If my sunglasses work in a Seaking, it is not unreasonable to infer that they will work in a Merlin.
A senior officer who declares that a new sunglasses trial is not required for a new aircraft is not declaring that it is not required for sunglasses to work in a Merlin, he is declaring that he would like his Merlin crew to have sunglasses before the merlin goes out of service.
As usual, you are messing around with words to make people say what you want them to say.
There is a big difference between declaring it unnecessary for a system to work, and declaring it unnecessary to test a system in the manner which you happen to want it tested.
ie If my sunglasses work in a Seaking, it is not unreasonable to infer that they will work in a Merlin.
A senior officer who declares that a new sunglasses trial is not required for a new aircraft is not declaring that it is not required for sunglasses to work in a Merlin, he is declaring that he would like his Merlin crew to have sunglasses before the merlin goes out of service.
I don't own this space under my name. I should have leased it while I still could
Back to passenger aircraft,
"Mrs PN, would you fly from Gatwick to Barbados without a pilot?"
"No, would you?"
On consideration, after a few thousand flights.
"Mrs PN, would you fly from Gatwick to Barbados without a pilot?"
"No, would you?"
On consideration, after a few thousand flights.
Join Date: Apr 1999
Location: Manchester, UK
Posts: 1,958
Likes: 0
Received 0 Likes
on
0 Posts
In an interesting and probably unintended feat of technical judo, the "anti" computerites have formed a powerful argument in favour of robotic unmanned aircraft. If some beastly enemies break all our toys with an EMP, theres no NOK letters to write; simply roll out another few dozen (comparatively) inexpensive drones and send them off for the return match.
Join Date: Feb 2006
Location: Hanging off the end of a thread
Posts: 33,071
Received 2,939 Likes
on
1,252 Posts
I'm guessing its because despite all the "I don't trust a computer" types on here, we have all actually come to trust FADEC because actually they are astonishingly better than humans at their job.
read all about it
http://www.flightglobal.com/news/art...ne-row-213371/
Tourist, what I am saying one cannot totally eradicate flaws, and it only takes one to make the whole thing go west. A man in the cockpit can at least work through a fix a computer simply cannot do.... a popped CB being a simple example. I also remember reading of one aircraft where they actually rewired part of it in flight. However what may tip the equation in military fighters is the limit of the mans tolerance to G forces that an autonomous aircraft does not suffer from.
.
Last edited by NutLoose; 20th Apr 2015 at 11:56.
Join Date: Apr 1998
Location: Mesopotamos
Posts: 5
Likes: 0
Received 0 Likes
on
0 Posts
I've always said that a pilot with a good sense of self preservation was a bonus to his/her passengers.
Now show me a robot that gives a damn.
FADECs, like most automation, are designed by humans and only capable of handling scenarios they have been designed to handle. As a fuel metering system they are brilliant, but not without there limitations - just ask anyone flying a twin engined helo and how they are taught to deal with identifying the failed engine - it's not quite as obvious as you would think.
Now show me a robot that gives a damn.
FADECs, like most automation, are designed by humans and only capable of handling scenarios they have been designed to handle. As a fuel metering system they are brilliant, but not without there limitations - just ask anyone flying a twin engined helo and how they are taught to deal with identifying the failed engine - it's not quite as obvious as you would think.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
Nutloose
Your Diamond example demonstrates exactly where the advantage of computers lies.
The human pilot did not follow the instructions and broke the system. He then did not manage to fix the problem and crashed. I fail to see how that is an advert for manned flight. The computer would have always followed the checklist.
The fix was not a fix for the FADEC at all. It was an attempt to idiot proof it against bad pilots.
Whatever the disadvantages, and there are currently many, there are many advantages to having no human, not restricted to g tolerance as mentioned earlier. The weighting is shifting all the time.
Cattle truck
Giving a Damn is meaningless. Really really wanting something is always trumped by reality.
I really really want to fly better than my autopilot on the approach. Rarely happens.
Please read earlier links about learning neural networks. They are decades old and are in currently flying aircraft.
Thanks for the mention of twin helos, something of which I am very experienced on.
They are a beautiful example of where a computer would be massively better.
When you have let's say Tq instability, a pilot has to try to work out from a variety of waving needles what is going on. It can be confusing because humans are easily overloaded by inputs.
Computers are not.
There are far more metrics produced by a modern engine than we would ever consider showing to a pilot. He would be instantly overloaded by a tenth of what a modern engine can produce in terms of data. An Airbus is sending home vastly more info to home base than the pilot ever sees.
That data makes it elementary for a computer to diagnose the problem. It is in fact very unlikely to ever come to that since continuous monitoring will pick up the problem before any pilot could ever dream of noticing the first needle wibble.
Your Diamond example demonstrates exactly where the advantage of computers lies.
The human pilot did not follow the instructions and broke the system. He then did not manage to fix the problem and crashed. I fail to see how that is an advert for manned flight. The computer would have always followed the checklist.
The fix was not a fix for the FADEC at all. It was an attempt to idiot proof it against bad pilots.
Whatever the disadvantages, and there are currently many, there are many advantages to having no human, not restricted to g tolerance as mentioned earlier. The weighting is shifting all the time.
Cattle truck
Giving a Damn is meaningless. Really really wanting something is always trumped by reality.
I really really want to fly better than my autopilot on the approach. Rarely happens.
Please read earlier links about learning neural networks. They are decades old and are in currently flying aircraft.
Thanks for the mention of twin helos, something of which I am very experienced on.
They are a beautiful example of where a computer would be massively better.
When you have let's say Tq instability, a pilot has to try to work out from a variety of waving needles what is going on. It can be confusing because humans are easily overloaded by inputs.
Computers are not.
There are far more metrics produced by a modern engine than we would ever consider showing to a pilot. He would be instantly overloaded by a tenth of what a modern engine can produce in terms of data. An Airbus is sending home vastly more info to home base than the pilot ever sees.
That data makes it elementary for a computer to diagnose the problem. It is in fact very unlikely to ever come to that since continuous monitoring will pick up the problem before any pilot could ever dream of noticing the first needle wibble.
Join Date: Nov 2010
Location: Tamworth, UK / Nairobi, Kenya
Posts: 616
Likes: 0
Received 0 Likes
on
0 Posts
Today, it's probably safe to say that a "computer" isn't better than a person at flying a plane in all circumstances.
But with the rate at which machine learning is advancing (not necessarily neural networks, but Markov and other probability "networks"), it is also safe to say that in the very near future a "computer" will be far better at handling abnormal situations than a human pilot. (at the extreme a decade, more likely within 3 to 5 years)
We have already seen cases where autonomous vehicles are better at handling many (not yet all) abnormal situations than a human. And the rate at which the technology is advancing would lead anyone who's studied these vehicles to conclude that the "computers" that are driving them will be more capable than humans within just a few years.
These "computers" are capable of absorbing the data from hundreds and thousands of sensors within milliseconds, and then predicting the hundreds and thousands of possible futures based on any number of combinations of actions, and thus determining the best action, all within fractions of a second. Then, after running around for awhile, they begin to "tune" their cost/values for inputs and actions and reach a point where they're no longer using the original logic "proposed" by the developers.
(I'm putting "computers" in quotes, because they're not like what we normally call a computer, they still have a CPU, and memory, etc. but they don't usually have much in the way of user interface equipment).
But with the rate at which machine learning is advancing (not necessarily neural networks, but Markov and other probability "networks"), it is also safe to say that in the very near future a "computer" will be far better at handling abnormal situations than a human pilot. (at the extreme a decade, more likely within 3 to 5 years)
We have already seen cases where autonomous vehicles are better at handling many (not yet all) abnormal situations than a human. And the rate at which the technology is advancing would lead anyone who's studied these vehicles to conclude that the "computers" that are driving them will be more capable than humans within just a few years.
These "computers" are capable of absorbing the data from hundreds and thousands of sensors within milliseconds, and then predicting the hundreds and thousands of possible futures based on any number of combinations of actions, and thus determining the best action, all within fractions of a second. Then, after running around for awhile, they begin to "tune" their cost/values for inputs and actions and reach a point where they're no longer using the original logic "proposed" by the developers.
(I'm putting "computers" in quotes, because they're not like what we normally call a computer, they still have a CPU, and memory, etc. but they don't usually have much in the way of user interface equipment).
Join Date: Jul 2006
Location: bristol
Age: 56
Posts: 1,051
Likes: 0
Received 0 Likes
on
0 Posts
I don't think there is much use in debating a large part of an autonomous mission as its very clear the unmanned aircraft could get itself to the battle area or target every time, and with perfect precision. What the unmanned aircraft can't do are the jobs the mk1 eyeball can do.
A pilot in the cockpit using his or her mk1 eyeball can (normally) differentiate between a Middle eastern wedding and rebels, friendly or opposition troops, civilian cars with innocent drivers, civilian cars with opposition rebel drivers or pickups mounted with 12.7mm machine guns etc. the mk1 eyeball can also tell if opposing troops want to surrender or attack.
While computers are better at navigation and have much faster processing and reaction times there is a very useful measure of artificial intelligence that no computer has achieved yet. When a robot/computer can be given a random basket of washed clothing and successfully iron each item then it will match the mk1 eyeball and human decision making. So far, no computer can even get close.
A pilot in the cockpit using his or her mk1 eyeball can (normally) differentiate between a Middle eastern wedding and rebels, friendly or opposition troops, civilian cars with innocent drivers, civilian cars with opposition rebel drivers or pickups mounted with 12.7mm machine guns etc. the mk1 eyeball can also tell if opposing troops want to surrender or attack.
While computers are better at navigation and have much faster processing and reaction times there is a very useful measure of artificial intelligence that no computer has achieved yet. When a robot/computer can be given a random basket of washed clothing and successfully iron each item then it will match the mk1 eyeball and human decision making. So far, no computer can even get close.
Ecce Homo! Loquitur...
Thread Starter
When a robot/computer can be given a random basket of washed clothing and successfully iron each item then it will match the mk1 eyeball and human decision making. So far, no computer can even get close.
I don't own this space under my name. I should have leased it while I still could
Really?
Unfortunately the Mk 1 Eyeball, connected to a Mk 1 Brain has failed every test you cite:
I accept you say Normally but the failure rate is sufficiently high to suggest the Mk 1 Eyeball alone is not infallible.
The A6 cable strike in Italy was a human error. A drone with proper sensors, in peacetime, would fly over or around, in wartime it might fly under; it should not hit it.
Unfortunately the Mk 1 Eyeball, connected to a Mk 1 Brain has failed every test you cite:
IWhat the unmanned aircraft can't do are the jobs the mk1 eyeball can do.
A pilot in the cockpit using his or her mk1 eyeball can (normally) differentiate between a Middle eastern wedding and rebels, friendly or opposition troops, civilian cars with innocent drivers, civilian cars with opposition rebel drivers or pickups mounted with 12.7mm machine guns etc. the mk1 eyeball can also tell if opposing troops want to surrender or attack.
A pilot in the cockpit using his or her mk1 eyeball can (normally) differentiate between a Middle eastern wedding and rebels, friendly or opposition troops, civilian cars with innocent drivers, civilian cars with opposition rebel drivers or pickups mounted with 12.7mm machine guns etc. the mk1 eyeball can also tell if opposing troops want to surrender or attack.
The A6 cable strike in Italy was a human error. A drone with proper sensors, in peacetime, would fly over or around, in wartime it might fly under; it should not hit it.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
"While computers are better at navigation and have much faster processing and reaction times there is a very useful measure of artificial intelligence that no computer has achieved yet. When a robot/computer can be given a random basket of washed clothing and successfully iron each item then it will match the mk1 eyeball and human decision making. So far, no computer can even get close."
Capability to do a given, defined task is one thing. But how do you design a robot to CARE or to TAKE RESPONSIBILITY? I lost count of human errors blamed on computer error, as if the computer did anything other than what its design made it do. Humans make mistakes, but we try to make them responsible for them and we try to change behaviors that led to or permitted the mistakes to go uncorrected.
Capability to do a given, defined task is one thing. But how do you design a robot to CARE or to TAKE RESPONSIBILITY? I lost count of human errors blamed on computer error, as if the computer did anything other than what its design made it do. Humans make mistakes, but we try to make them responsible for them and we try to change behaviors that led to or permitted the mistakes to go uncorrected.