This assumption is totally false. The truth is the computer has to assess the situation exactly like a human brain using sensors which have a finite accuracy and some uncertainity on the readings.
Nope. You are assuming that an AI will drive like a human. Here's the situation, you are in LA, its foggy, you observe the average speed on the road to be 80 mph. You can stop from 80 about 3x the visual distance. IF you slow to 40, where you can stop, you'll likely be hit from behind. If you go 80, you'll likely hit something. So, what do you, as a human with hours to think about it, do? The computer will come to a quicker answer. Get off the road. Re route on a slower road. Humans *never* take that option, and it's the safe one.
People are stupid. It doesn't take much for a computer to beat a human. If the conditions are such that the processing power is taxed, then the computer will re-route to a less troublesome route, or call home for a summary algorithm.
Every complaint I've seen about self-driving cars being unable to avoid a crash devolves into a situation that could be restated:
You drive off a cliff. You turn on the self-drive. The AI can't stop the collision with the ground at the end. So AI is no better than a human. Ever.
Decode everything, including images from cameras, reconstruct a 3-D representation of the scene and have some mean to make decisions from the scene and the physics. That's a lot of processing a human brain can do much more faster than a computer today.
That's not how humans work, so why would you assume an AI would work worse than a human? Humans work by identifying "hazards" and tracking those and only those. Yes, the new drivers suffer driving paralysis, where everything is a threat, but a practiced driver will not even "see" a parked car. They are so common, that the pre-processor erases them from view before they are processed. Now, when that car is by a park, and the park has kids, the pre-processor won't erase it and it gets labeled a hazard, as it's an obstruction that could hide a running child. So your blindness is location specific. But it's literally deleted by your vision before you see it, when it's a non-hazard.
As a human, you don't decode "everything". So why would you assume a poor implementation in the computer version? Driving is simple. Identify hazards. Avoid them. When hazards conflict, react.
One when it really mattered, I had the luxury of dodging a pickup truck which I thought was going to swerve into my lane (his lane very suddenly ended due to a construction site which was poorly marked). But the fully loaded dump truck tailgating me didn't have the luxury of braking. Instead of dodging, I braked and totaled my car.
You braked hard for a non-obstruction without regard to your surroundings, and take that as a reason to always swerve? In that case, you should have not braked or swerved. Just take the hit. You caused a crash. Your stupidity doesn't translate to others.
A side-swipe is not something to bother avoiding. That's one reason the fatality rate in the US is not any better than countries with worse conditions. People would rather total their car to blame someone else, than just trade paint with the pickup and deal with it after. A side-swipe is a non-injury collision in almost all cases. Rear-end crashes are often fatal.
The reason to not dodge is simple. Most people can't do it. Deer kill more people who dodge them by driving into a tree than are killed by striking a deer. Oh, and you are dodging a moving object. In most cases, to "dodge" a moving object, you want to steer at it, not away from it. Then it moves before you get there. If the car is moving across, right to left, and you go left to get around it, you end up hitting it anyway, as you moved into its path. Had you gone straight, you'd have not hit it. People in the moment, don't have time to think. Just react. And if you want the best reaction that will save the most lives, the answer is brake, don't swerve. It's not perfect, but it's the optimal reaction.
Self driving cars will transfer the liability from the owner of the car to the manufacturer of the car. This is already happening.
Happened long before AI was considered. If the manufacturer made defective brakes, you'd sue the maker, rather than focus on the owner. The maker had more money, and was responsible for the design.
Same as today. Nothing new. Nothing has changed.
The robot had calculated the chances of saving each and Will won the AI lottery.
The AI didn't cause survivor's guilt. And in the movie, he's exactly the opposite of TFA. He'd rather be dead than alive, having watched the girl die. In his own words (forgive a poor quote, as I'm going from memory of a movie I've seen exactly once) "I'm a cop, when we hit the water, I knew we were all dead." He gave himself and the girl a 0% chance of survival. That he got saved should be a good thing, but the survival guilt got to him. The robot should have tried (and failed) to save the girl, then he wouldn't have survival guilt, and the girl would be dead, but greater effort would have been spent saving her. How is that an efficient outcome?
Yes, if you want a sad example, look at infant mortality. Some places don't (or at least didn't, at some point) count deaths under 1 year as infant mortality, they are considered "stillborn" and so they have less infant mortality than other places, but more dead babies.
I never said it was slavery, so if you want to respond to what people write, feel free, but your words didn't have anything to do with mine, other than I did say "slavery" in my post.
I never said it was. You are arguing with other people's arguments, assuming I mean something other than what I said, and am aiming for some specific conclusiong 10 posts from now. But like your opinion on economics, you are 100% wrong. Not even close.
Don't forget to copy the FCC (communications company issue). I was told for 6 months that fixing my DSL was "impossible". Within 48 hours of writing a formal complaint to the FCC (copying my ISP as well), the issue was fixed.
So if your vacuum were proven sentient, it still wouldn't be real. And a person is "real" just because we assume sentience, even if they can't prove it.
It's not a standard pidgin English it's our own English without a workable level of education.
So a dialect can't exist if there are more than one? So Ebonics is different in LA than Chicago. Does that linguistically prove that neither can be a dialect? Your comments about it being hard/impossible for a native speaker of a different dialect to understand it, but the native speakers of that dialect don't have a problem understanding others is more a definition of a dialect than proof it isn't one.
One of the popular AI themes in fiction is that the AI has consciousness, and deliberately hides it. So in the face of a smart enough AI, it may choose to fail the consciousness test.
This is about as retarded as calling your toaster a "person" and assigning it "rights".
That's unrelated to the issues mentioned in TFA.
a) define, and
b) prove a computer has consciousness
THEN we'll talk about laws.
We have laws protecting people and squid, and have no definition of consciousness. So why is that a requirement for computers, when it wasn't there for any previous laws? Your non sequitur is non sequitur.
His test wasn't a test. It was a guess about how one might be able to test. He never wrote up a test plan, or anything like that. It was others around him that ran with the idea, so the Turing Test (as thought of today) was never thought of by Alan Turing.
For what he was guessing about it was more a dismissal of the idea that a calculator that can count faster than a human is "smarter". That's silly, a better test would be to ask it questions, and if you can't tell if it's human or computer, then it's "smart". He knew such a test wouldn't happen in his lifetime, but used the Turing Test to dismiss those asking questions, not as a proposal for a formal test on AI.
You might as well come up with a test to determine whether an alien-human hybrid is more human-like or alien-like. You could make some guesses, but by the time it's an actual concern, someone will have come up with a better test. Turing's test was a dismissal of all the people that thought we were 5 years from AI (and we've been there for 80 or so years).
There is no reason that even a fully sentient robot will even have to have a body.
Yes, there is. If it's a fully sentient robot, it has to have a body, or it's a fully sentient computer. Or are you getting confused about a central AI controlling a dumb body, vs requiring the AI inhabit the body it controls?
I disagree of course. For example, those 62 people employ a vast number of people over many decades as well as creating far more wealth than they own.
I disagree of course. The 62 "create" nothing. Those they employ might, but you are confusing those who "own" the work (as defined by the 62) vs those who did the work.
So a properly configured phone that's powered on before plugged in, or off while charged has a 0% chance of compromising the device (considering only currently known attacks). Only if your phone is improperly configured, or powered on while plugged in is there any possible attack against it.
Ebonics isn't about teaching only a dumbed down English. It was recognizing that there are some language forks that are different enough that they may qualify as a dialect. Recognizing this culture shift isn't the same as making in a national language.
Like most things, the racists made it (education cuts and work to isolate and ostracize unwanted minorities), then, when pointed out what they created, they rebelled against it. PC was created by the conservatives, trying to make fun of people who told them they can't call "those people" Niggers anymore. Just like SJW got a boost in popularity when the conservatives tried to make fun of it.
You'd think they'd learn from their numerous mistakes, but they are conservatives.
You would say that of any regulation. Doesn't make it true, and even if it were, it's not necessarily a bad thing. If you always move your commerce and industry to the poorest nation, that nation will be lifted by the influx of activity, until all nations are at or above some threshold. Hopefully, that line would be a nice and high one.
My definition of "real AI" is an AI that can program an AI smarter than it is. Give the AI 10,000,000 generations (about 10 minutes?) and it'll be perfect, or as close as we'll ever get.
I swear, I just want to start punching people in the head whenever they start talking about the crap we have now as 'artificial intelligence'.
That's why weak AI is being called "machine learning". Because the AI name has been so abused that it's now meaningless. And machine learning isn't AI, it's just made by failed AI researchers.
Slashdot Mirror
This assumption is totally false. The truth is the computer has to assess the situation exactly like a human brain using sensors which have a finite accuracy and some uncertainity on the readings.
Nope. You are assuming that an AI will drive like a human. Here's the situation, you are in LA, its foggy, you observe the average speed on the road to be 80 mph. You can stop from 80 about 3x the visual distance. IF you slow to 40, where you can stop, you'll likely be hit from behind. If you go 80, you'll likely hit something. So, what do you, as a human with hours to think about it, do? The computer will come to a quicker answer. Get off the road. Re route on a slower road. Humans *never* take that option, and it's the safe one.
People are stupid. It doesn't take much for a computer to beat a human. If the conditions are such that the processing power is taxed, then the computer will re-route to a less troublesome route, or call home for a summary algorithm.
Every complaint I've seen about self-driving cars being unable to avoid a crash devolves into a situation that could be restated:
You drive off a cliff. You turn on the self-drive. The AI can't stop the collision with the ground at the end. So AI is no better than a human. Ever.
Decode everything, including images from cameras, reconstruct a 3-D representation of the scene and have some mean to make decisions from the scene and the physics. That's a lot of processing a human brain can do much more faster than a computer today.
That's not how humans work, so why would you assume an AI would work worse than a human? Humans work by identifying "hazards" and tracking those and only those. Yes, the new drivers suffer driving paralysis, where everything is a threat, but a practiced driver will not even "see" a parked car. They are so common, that the pre-processor erases them from view before they are processed. Now, when that car is by a park, and the park has kids, the pre-processor won't erase it and it gets labeled a hazard, as it's an obstruction that could hide a running child. So your blindness is location specific. But it's literally deleted by your vision before you see it, when it's a non-hazard.
As a human, you don't decode "everything". So why would you assume a poor implementation in the computer version? Driving is simple. Identify hazards. Avoid them. When hazards conflict, react.
One when it really mattered, I had the luxury of dodging a pickup truck which I thought was going to swerve into my lane (his lane very suddenly ended due to a construction site which was poorly marked). But the fully loaded dump truck tailgating me didn't have the luxury of braking. Instead of dodging, I braked and totaled my car.
You braked hard for a non-obstruction without regard to your surroundings, and take that as a reason to always swerve? In that case, you should have not braked or swerved. Just take the hit. You caused a crash. Your stupidity doesn't translate to others.
A side-swipe is not something to bother avoiding. That's one reason the fatality rate in the US is not any better than countries with worse conditions. People would rather total their car to blame someone else, than just trade paint with the pickup and deal with it after. A side-swipe is a non-injury collision in almost all cases. Rear-end crashes are often fatal.
The reason to not dodge is simple. Most people can't do it. Deer kill more people who dodge them by driving into a tree than are killed by striking a deer. Oh, and you are dodging a moving object. In most cases, to "dodge" a moving object, you want to steer at it, not away from it. Then it moves before you get there. If the car is moving across, right to left, and you go left to get around it, you end up hitting it anyway, as you moved into its path. Had you gone straight, you'd have not hit it. People in the moment, don't have time to think. Just react. And if you want the best reaction that will save the most lives, the answer is brake, don't swerve. It's not perfect, but it's the optimal reaction.
Great example. Of course, the answer tends to UBI, when you don't try to punish the poor for being "lazy" and take the emotion out of it.
Self driving cars will transfer the liability from the owner of the car to the manufacturer of the car. This is already happening.
Happened long before AI was considered. If the manufacturer made defective brakes, you'd sue the maker, rather than focus on the owner. The maker had more money, and was responsible for the design.
Same as today. Nothing new. Nothing has changed.
The robot had calculated the chances of saving each and Will won the AI lottery.
The AI didn't cause survivor's guilt. And in the movie, he's exactly the opposite of TFA. He'd rather be dead than alive, having watched the girl die. In his own words (forgive a poor quote, as I'm going from memory of a movie I've seen exactly once) "I'm a cop, when we hit the water, I knew we were all dead." He gave himself and the girl a 0% chance of survival. That he got saved should be a good thing, but the survival guilt got to him. The robot should have tried (and failed) to save the girl, then he wouldn't have survival guilt, and the girl would be dead, but greater effort would have been spent saving her. How is that an efficient outcome?
Yes, if you want a sad example, look at infant mortality. Some places don't (or at least didn't, at some point) count deaths under 1 year as infant mortality, they are considered "stillborn" and so they have less infant mortality than other places, but more dead babies.
I never said it was slavery, so if you want to respond to what people write, feel free, but your words didn't have anything to do with mine, other than I did say "slavery" in my post.
I never said it was. You are arguing with other people's arguments, assuming I mean something other than what I said, and am aiming for some specific conclusiong 10 posts from now. But like your opinion on economics, you are 100% wrong. Not even close.
It's not a "cut down" version of English. It's a changed and transformed, with new words, not just removal of some.
How do you think "real" dialects were formed?
So the slave owners are performing a valuable service, above any harm created by slavery. Yay capitalist exploitation of the workers.
Don't forget to copy the FCC (communications company issue). I was told for 6 months that fixing my DSL was "impossible". Within 48 hours of writing a formal complaint to the FCC (copying my ISP as well), the issue was fixed.
So if your vacuum were proven sentient, it still wouldn't be real. And a person is "real" just because we assume sentience, even if they can't prove it.
And the nutjobs here complain I'm illogical.
It's not a standard pidgin English it's our own English without a workable level of education.
So a dialect can't exist if there are more than one? So Ebonics is different in LA than Chicago. Does that linguistically prove that neither can be a dialect? Your comments about it being hard/impossible for a native speaker of a different dialect to understand it, but the native speakers of that dialect don't have a problem understanding others is more a definition of a dialect than proof it isn't one.
One of the popular AI themes in fiction is that the AI has consciousness, and deliberately hides it. So in the face of a smart enough AI, it may choose to fail the consciousness test.
This is about as retarded as calling your toaster a "person" and assigning it "rights".
That's unrelated to the issues mentioned in TFA.
a) define, and
b) prove a computer has consciousness
THEN we'll talk about laws.
We have laws protecting people and squid, and have no definition of consciousness. So why is that a requirement for computers, when it wasn't there for any previous laws? Your non sequitur is non sequitur.
How do you test the consciousness of a squid?
His test wasn't a test. It was a guess about how one might be able to test. He never wrote up a test plan, or anything like that. It was others around him that ran with the idea, so the Turing Test (as thought of today) was never thought of by Alan Turing.
For what he was guessing about it was more a dismissal of the idea that a calculator that can count faster than a human is "smarter". That's silly, a better test would be to ask it questions, and if you can't tell if it's human or computer, then it's "smart". He knew such a test wouldn't happen in his lifetime, but used the Turing Test to dismiss those asking questions, not as a proposal for a formal test on AI.
You might as well come up with a test to determine whether an alien-human hybrid is more human-like or alien-like. You could make some guesses, but by the time it's an actual concern, someone will have come up with a better test. Turing's test was a dismissal of all the people that thought we were 5 years from AI (and we've been there for 80 or so years).
There is no reason that even a fully sentient robot will even have to have a body.
Yes, there is. If it's a fully sentient robot, it has to have a body, or it's a fully sentient computer. Or are you getting confused about a central AI controlling a dumb body, vs requiring the AI inhabit the body it controls?
I disagree of course. For example, those 62 people employ a vast number of people over many decades as well as creating far more wealth than they own.
I disagree of course. The 62 "create" nothing. Those they employ might, but you are confusing those who "own" the work (as defined by the 62) vs those who did the work.
So a properly configured phone that's powered on before plugged in, or off while charged has a 0% chance of compromising the device (considering only currently known attacks). Only if your phone is improperly configured, or powered on while plugged in is there any possible attack against it.
Ebonics isn't about teaching only a dumbed down English. It was recognizing that there are some language forks that are different enough that they may qualify as a dialect. Recognizing this culture shift isn't the same as making in a national language.
Like most things, the racists made it (education cuts and work to isolate and ostracize unwanted minorities), then, when pointed out what they created, they rebelled against it. PC was created by the conservatives, trying to make fun of people who told them they can't call "those people" Niggers anymore. Just like SJW got a boost in popularity when the conservatives tried to make fun of it.
You'd think they'd learn from their numerous mistakes, but they are conservatives.
You would say that of any regulation. Doesn't make it true, and even if it were, it's not necessarily a bad thing. If you always move your commerce and industry to the poorest nation, that nation will be lifted by the influx of activity, until all nations are at or above some threshold. Hopefully, that line would be a nice and high one.
If anyone cares to read the actual draft document
Obvioulsy from the comments, nobody has, or wants to read it. It's easier to tear down a straw man than to understand something new.
I swear, I just want to start punching people in the head whenever they start talking about the crap we have now as 'artificial intelligence'.
That's why weak AI is being called "machine learning". Because the AI name has been so abused that it's now meaningless. And machine learning isn't AI, it's just made by failed AI researchers.