Why does it need to work. It's a search aid, not adding cause or suspicion to anyone for any reason. Thermal imaging hasn't been shown by a peer reviewed study to work. Guns haven't been shown by a peer reviewed study to work. The cops have tasks. They use the tools they need to do them.
Or are you the idiot that sues the judge for having yellow fringe on the flag in the courtroom, and demands to see the calibration test for the pen used to write your report?
Like the time I was in jury duty. The judge asked all seated "do you have a personal problem or history that would keep you from judging a policeman's testimony fairly?" The woman next to me didn't say yes. Those who said yes were dismissed. She was the sole hold-out for not guilty. "The cop said he did it, so obviously he was innocent. My son in law is a cop, so I know all cops are lying assholes."
Hung jury. She should have gone to jail for perjury, but nobody cared.
When were the first 1280x1024? My memory places it about 20 years ago, and I had one. And my most recent work computer has fewer pixels than that monitor did.
The Turing Test was a test of machine intelligence. That was what they thought about AI in the 1950s. So you are saying that scientists of the era didn't consider human-like intelligence to be the "ideal" of AI?
History and facts prove you wrong. I'm not the strawman, you are. Or was Turing not a scientist?
I'm applying the first definitions to the current work and find that the biggest change is in the scientist's definitions, not progress towards the initial definitions. Go read how many comments state Watson or Siri is an AI. That's where the current focus is. On making people think that the device is "smart", not on making it so.
That's impossible. Nobody has ever tried that as a real experiment and it only lives as a contrived (false) thought experiment.
It requires that the person who learns the symbol for "fire" for example, when asking for fire, knows that "fire" results in fire, but doesn't make the association of that character to the fire.
A real person in that situation would end up "translating" by learning the languages in question, not by magically decoding them without understanding.
But hey, even if the door says "open", it's a private house. The "normal" thing to do is knock, go "hey is anyone home?", enter, say hi and state your business.
If the door says "open" then it is a home based business that has invited you in.
Else why the heck are you entering that door?
Because they invited me in. I was curious why.
Street View cars were accessing email, web history and other data
Web history is stored on a computer. If they actually accessed web history, rather than monitoring the broadcast web requests, then they were rifling through locked cabinets after being invited in. If they just captured sites that were accessed while they could see them, then it's no different than driving down the street with a camera pointed to the side, recording the colors of the clothes of anyone they could see in a window (100% legal under all laws that don't involve "on a computer").
I see no way where Google was wrong with this. They didn't "enter" the house. They did the equivalent of looking through a window with a "regular" (not specialized) device from public property. That's explicitly legal until someone says "on a computer".
This is what I came to read. Someone asserting NASA's calculations on air worthiness being wrong. No evidence. Just some assertions that it'd be hard, so we should consider it impossible, even if NASA says it's possible.
The only goalpost moving I see is from the computer scientists. They originally thought AI would be a human analogue. Now, AI is anything that an average human can't do as fast. My 40 year old calculator is AI, according to most of the definitions people are throwing around here to show how far we've come.
70% accuracy quite easily (between romance languages, anyway. It really sucks for Japanese or Russian, for example
When translating into a native toungue, it's better than 70% accuracy, when minor human filtering is used. It's not perfect, and there are lots of idioms it will miss, but, amazingly, the global world has fixed much of that. A green joke (Spanish) is an off-color joke (English), and someone "should" be able to figure out the meaning from context. A "dark horse" in a contest is a "black horse" in Chinese. Someone should be able to figure that out as well, right?
Translation will always work best into one's native language. The translations fail at making it sound natural. Often because things are untranslatable. Like from between Chinese and English. You don't say "no" in Chinese. You say "Yes, but I can't". But saying that in English would not be direct enough, and would imply some lack of conviction that isn't there. But "no" indicates a lack of politeness. I would love to see translations translate and transliterate. So it says:
"Portugal is the dark horse" (translation) "Portugal is (the) black horse" (transliteration)
Conferring both the meaning and the tone requires a re-write, not a simple translation. But giving the literal translation and the meaningful one helps with that. At least with more than tourist words. "Where's the toilet?" loses little in translation in any language.
Plato made more progress in the years he was alive than CS has done in the same time frame.
All we got from the computer scientists working on it is a re-definition of AI to include toasters that monitor "doneness" of the bread, so that they could declare some progress.
Then why does it have so many disclaimers? And you know how you can pick waypoints? Why is it that I can play with waypoints and knock 5 minutes and 10 miles off a trip? It's smarter than the entire human race, and any single human is smarter than it?
Watson, OTOH, is a much more general intelltigence.
Watson is a fancy interface to Wikipedia. Watson has no more thinking capability than my 40 year old calculator. Just more processing power, bigger DB, and a fancier interface.
He's making the point that typing 2+2 on a calculator and getting the answer 4 isn't intelligence. Apparently your point is any correct answer is AI.
Sadly, in this discussion of AI, all definitions that exclude General AI would seem to include (dumb 1970s) calculators as AI devices. Yes, as computing power has gotten better, the answers have gotten more complex. But are still nothing more than a "simple" calculation without "thought", deduction, or other "intelligent" process.
AI to a human is asking a computer a problem it doesn't know how to answer and getting a valid answer.
My example elsewhere is that humans are intelligent because they can estimate the height of the Empire State Building. A computer can't. It either knows, or doesn't. A human that knows that it's "tall" and could use unclear information it has about it (about 100 stories) and venture a guess. A computer will look up the information. If it can't find it, it won't answer.
We do know that , if not all, the vast majority of decisions are made before consciously thought about.
Consciousness isn't the intelligence. It's the layer between the senses and the intelligence.
But in principle, physics can be simulated by an algorithm. Therefore a human brain can be modelled at the particle level and run in simulation. Therefore whatever a human brain is doing that produces intelligence (assuming for now that it does, in fact, produce intelligence) can, in principle, be reproduced by an algorithm, even if it has to treat the brain as a black-box to do so.
That you keep saying "in principle" indicates it's (at least currently) impossible. Why not simulate a smarter brain. Then simulate it faster. Like my old DOS emulators have to have speed settings or the games are unplayable on modern computers. When you have the smartest brain on the planet running 1000 times faster than regular brains, you task it with making a brain 1000 times smarter running a million times faster. Then your brute-force AI will find your elegant AI for you.
Surely that would imply I could rewire your neurons without affecting your ability to think.
I've had that done, and my wife insists that I'm more irritable now, but nobody else noticed a problem.
Or were you indicating that you would do so deliberately destructively (big lobotomy), rather than randomly, or delicately?
if you take "intelligence" to mean "the ability to independently acquire and apply knowledge" has already been solved. IBM's Watson is clearly artificial and it can answer open ended general knowledge questions that its creator cannot.
100% incorrect. Watson does so faster, but any human with the access to Watson's resources and infinite time, will be more accurate. You are testing speed, not the ability to apply knowledge.
Fast seems to confuse AI bigots. Just because it's faster than me, doesn't mean it's "smarter". Are baseball players smarter because they are better at hitting and catching the ball?
Of course AI will fail. They've had more progress in re-designing the definition of AI than in building AI as first conceived by the first experts who worked on it. The closest to the actual human brain was analogue computing. But I think that failed because it wasn't digital enough. Humans work by evaluating options (multiple variables at once) and getting an answer close, based on all the soft inputs. Computers work by what humans call rote memorization. Even Watson, the most advanced AI, can't "think" at all. If you ask a human how tall the Empire State building is, the human will likely not recall the spec sheet for the building, and instead would go through some quick fuzzy (as in inexact) logic to estimate an answer and hope that's in the tolerance of the person asking. That's "intelligence" even if they get the wrong answer.
Watson will, without thought or intelligence, look up the Wikipedia page for the Empire State building (or equivalent). That's not AI, that's just a big memory with fast recall.
That Watson is more accurate doesn't make it smarter.
So the phone book was AI. In knows everyone's name. That's smarter than most people. Thus AI.
"complicated problem solving" isn't AI. The algorithm is never "smarter" than the sum of the people creating it.
AI to everyone outside CS is Generalized AI. Only CS redefined AI, because they were tired of decades of failure. Now, they call "hello world" AI because it can say hello faster than a human, thus *must be* AI.
Except for the cell phone in your pocket, that can recognize your commands and search the internet for what you requested, or translate your statement into any of a dozen foreign languages, and has a camera that can recognize faces, and millions of objects, and can connect to expert systems that can, for instance, diagnose diseases better than all but the very best doctors. Oh, and your cellphone can also beat any grandmaster in the world at chess.
That's "AI" like my dog figuring out where to be to catch the ball I threw makes him a physics expert. Doing something "dumb" fast doesn't make one smart.
Slashdot Mirror
Why does it need to work. It's a search aid, not adding cause or suspicion to anyone for any reason. Thermal imaging hasn't been shown by a peer reviewed study to work. Guns haven't been shown by a peer reviewed study to work. The cops have tasks. They use the tools they need to do them.
Or are you the idiot that sues the judge for having yellow fringe on the flag in the courtroom, and demands to see the calibration test for the pen used to write your report?
Like the time I was in jury duty. The judge asked all seated "do you have a personal problem or history that would keep you from judging a policeman's testimony fairly?" The woman next to me didn't say yes. Those who said yes were dismissed. She was the sole hold-out for not guilty. "The cop said he did it, so obviously he was innocent. My son in law is a cop, so I know all cops are lying assholes."
Hung jury. She should have gone to jail for perjury, but nobody cared.
When were the first 1280x1024? My memory places it about 20 years ago, and I had one. And my most recent work computer has fewer pixels than that monitor did.
The Turing Test was a test of machine intelligence. That was what they thought about AI in the 1950s. So you are saying that scientists of the era didn't consider human-like intelligence to be the "ideal" of AI?
History and facts prove you wrong. I'm not the strawman, you are. Or was Turing not a scientist?
I'm applying the first definitions to the current work and find that the biggest change is in the scientist's definitions, not progress towards the initial definitions. Go read how many comments state Watson or Siri is an AI. That's where the current focus is. On making people think that the device is "smart", not on making it so.
Because it performs keyword hits against it.
That's impossible. Nobody has ever tried that as a real experiment and it only lives as a contrived (false) thought experiment.
It requires that the person who learns the symbol for "fire" for example, when asking for fire, knows that "fire" results in fire, but doesn't make the association of that character to the fire.
A real person in that situation would end up "translating" by learning the languages in question, not by magically decoding them without understanding.
It also fails to mention when they are installing GPS around Mars.
The summary mentions NASA, And I'm not an American. I didn't bother to read well because the details of something that doesn't exist don't matter.
But hey, even if the door says "open", it's a private house. The "normal" thing to do is knock, go "hey is anyone home?", enter, say hi and state your business.
If the door says "open" then it is a home based business that has invited you in.
Else why the heck are you entering that door?
Because they invited me in. I was curious why.
Street View cars were accessing email, web history and other data
Web history is stored on a computer. If they actually accessed web history, rather than monitoring the broadcast web requests, then they were rifling through locked cabinets after being invited in. If they just captured sites that were accessed while they could see them, then it's no different than driving down the street with a camera pointed to the side, recording the colors of the clothes of anyone they could see in a window (100% legal under all laws that don't involve "on a computer").
I see no way where Google was wrong with this. They didn't "enter" the house. They did the equivalent of looking through a window with a "regular" (not specialized) device from public property. That's explicitly legal until someone says "on a computer".
This is what I came to read. Someone asserting NASA's calculations on air worthiness being wrong. No evidence. Just some assertions that it'd be hard, so we should consider it impossible, even if NASA says it's possible.
It's called moving the goalposts.
The only goalpost moving I see is from the computer scientists. They originally thought AI would be a human analogue. Now, AI is anything that an average human can't do as fast. My 40 year old calculator is AI, according to most of the definitions people are throwing around here to show how far we've come.
Watson, on the other hand, comes remarkably close.
Watson is more like a fancy interface for Wikipedia than AI. Being fast doesn't make one "smart"
70% accuracy quite easily (between romance languages, anyway. It really sucks for Japanese or Russian, for example
When translating into a native toungue, it's better than 70% accuracy, when minor human filtering is used. It's not perfect, and there are lots of idioms it will miss, but, amazingly, the global world has fixed much of that. A green joke (Spanish) is an off-color joke (English), and someone "should" be able to figure out the meaning from context. A "dark horse" in a contest is a "black horse" in Chinese. Someone should be able to figure that out as well, right?
Translation will always work best into one's native language. The translations fail at making it sound natural. Often because things are untranslatable. Like from between Chinese and English. You don't say "no" in Chinese. You say "Yes, but I can't". But saying that in English would not be direct enough, and would imply some lack of conviction that isn't there. But "no" indicates a lack of politeness. I would love to see translations translate and transliterate. So it says:
"Portugal is the dark horse" (translation)
"Portugal is (the) black horse" (transliteration)
Conferring both the meaning and the tone requires a re-write, not a simple translation. But giving the literal translation and the meaningful one helps with that. At least with more than tourist words. "Where's the toilet?" loses little in translation in any language.
It is completely impossible to say whether any entity, living or not, possesses that kind of introspective intelligence.
No, it's not. It's just impossible to demonstrate it. The chatbot lies. I don't. That's the difference.
Plato made more progress in the years he was alive than CS has done in the same time frame.
All we got from the computer scientists working on it is a re-definition of AI to include toasters that monitor "doneness" of the bread, so that they could declare some progress.
None of your links work.
Then why does it have so many disclaimers? And you know how you can pick waypoints? Why is it that I can play with waypoints and knock 5 minutes and 10 miles off a trip? It's smarter than the entire human race, and any single human is smarter than it?
Watson, OTOH, is a much more general intelltigence.
Watson is a fancy interface to Wikipedia. Watson has no more thinking capability than my 40 year old calculator. Just more processing power, bigger DB, and a fancier interface.
Sadly, in this discussion of AI, all definitions that exclude General AI would seem to include (dumb 1970s) calculators as AI devices. Yes, as computing power has gotten better, the answers have gotten more complex. But are still nothing more than a "simple" calculation without "thought", deduction, or other "intelligent" process.
AI to a human is asking a computer a problem it doesn't know how to answer and getting a valid answer.
My example elsewhere is that humans are intelligent because they can estimate the height of the Empire State Building. A computer can't. It either knows, or doesn't. A human that knows that it's "tall" and could use unclear information it has about it (about 100 stories) and venture a guess. A computer will look up the information. If it can't find it, it won't answer.
We do know that , if not all, the vast majority of decisions are made before consciously thought about.
Consciousness isn't the intelligence. It's the layer between the senses and the intelligence.
yeah, my 40 year old calculator has a M+ button. It's AI.
No, AI has just become a useless word, re-defined by CS until everything is AI because all the CS people got tired of decades of failure.
But in principle, physics can be simulated by an algorithm. Therefore a human brain can be modelled at the particle level and run in simulation. Therefore whatever a human brain is doing that produces intelligence (assuming for now that it does, in fact, produce intelligence) can, in principle, be reproduced by an algorithm, even if it has to treat the brain as a black-box to do so.
That you keep saying "in principle" indicates it's (at least currently) impossible. Why not simulate a smarter brain. Then simulate it faster. Like my old DOS emulators have to have speed settings or the games are unplayable on modern computers. When you have the smartest brain on the planet running 1000 times faster than regular brains, you task it with making a brain 1000 times smarter running a million times faster. Then your brute-force AI will find your elegant AI for you.
Surely that would imply I could rewire your neurons without affecting your ability to think.
I've had that done, and my wife insists that I'm more irritable now, but nobody else noticed a problem.
Or were you indicating that you would do so deliberately destructively (big lobotomy), rather than randomly, or delicately?
if you take "intelligence" to mean "the ability to independently acquire and apply knowledge" has already been solved. IBM's Watson is clearly artificial and it can answer open ended general knowledge questions that its creator cannot.
100% incorrect. Watson does so faster, but any human with the access to Watson's resources and infinite time, will be more accurate. You are testing speed, not the ability to apply knowledge.
Fast seems to confuse AI bigots. Just because it's faster than me, doesn't mean it's "smarter". Are baseball players smarter because they are better at hitting and catching the ball?
Of course AI will fail. They've had more progress in re-designing the definition of AI than in building AI as first conceived by the first experts who worked on it. The closest to the actual human brain was analogue computing. But I think that failed because it wasn't digital enough. Humans work by evaluating options (multiple variables at once) and getting an answer close, based on all the soft inputs. Computers work by what humans call rote memorization. Even Watson, the most advanced AI, can't "think" at all. If you ask a human how tall the Empire State building is, the human will likely not recall the spec sheet for the building, and instead would go through some quick fuzzy (as in inexact) logic to estimate an answer and hope that's in the tolerance of the person asking. That's "intelligence" even if they get the wrong answer.
Watson will, without thought or intelligence, look up the Wikipedia page for the Empire State building (or equivalent). That's not AI, that's just a big memory with fast recall.
That Watson is more accurate doesn't make it smarter.
Actually it is AI.
So the phone book was AI. In knows everyone's name. That's smarter than most people. Thus AI.
"complicated problem solving" isn't AI. The algorithm is never "smarter" than the sum of the people creating it.
AI to everyone outside CS is Generalized AI. Only CS redefined AI, because they were tired of decades of failure. Now, they call "hello world" AI because it can say hello faster than a human, thus *must be* AI.
Except for the cell phone in your pocket, that can recognize your commands and search the internet for what you requested, or translate your statement into any of a dozen foreign languages, and has a camera that can recognize faces, and millions of objects, and can connect to expert systems that can, for instance, diagnose diseases better than all but the very best doctors. Oh, and your cellphone can also beat any grandmaster in the world at chess.
That's "AI" like my dog figuring out where to be to catch the ball I threw makes him a physics expert. Doing something "dumb" fast doesn't make one smart.
It's not always 20 years. http://moller.com/dev/index.ph... has been 5 years away from a commercial Skycar for about 50 years now.