[mike777]

http://www.kurzweilai.net/meme/frame.html?....html?m=4%23683


Despite Wars in Iraq and the MidEast and concerns over Korea the world marches on!


The advent of strong AI (exceeding human intelligence) is the most important transformation this century will see, and it will happen within 25 years, says Ray Kurzweil, who will present this paper at The Dartmouth Artificial Intelligence Conference: The next 50 years (AI@50) on July 14, 2006. (Added July 13th 2006)

[DrTodd13]

Biological systems are physical systems just like mechanical systems are physical systems. There is nothing special about a biological system and as far as I'm concerned no reason to believe that biological systems are capable of something that mechanical systems are not. Having said that, I think that all attempts at deterministic consciousness will fail. Most strong AI people believe that consciousness is some form of emergent behavior. In other words, once a system gets sufficiently complex, it can start to exhibit behaviors that are essentially unpredictable even though every step along the way is a fundamentally deterministic one. On the flip side, there's a good book by Roger Penrose (sorry, I forget the name..maybe Shadows of the Mind?) that argues that since humans can perform mathmathetical feats that have been proven impossible to accomplish in a deterministic system that human thought must be inherently non-deterministic. He talks about how there are certain single-celled organisms that actually appear to learn and remember things. These organisms are covered in fine hair-like structures built out of molecules that can be in one of two states, either open or closed. Computer scientists among us will instantly recognize the fundamentals of a computing system as soon as you have something that can be in one of two states and that can change from one state to another in response to input. Moreover, the internals of these hairlike structures are sufficiently small as to allow non-deterministic quantum behavior to occur within them. Penrose argues that some combination of classical computation with non-classical quantum behavior is going on here. As it so happens, the neurons in our brain have the same fine hair-like structures! Not only do neurons reach out and touch each other with their axions and dendrites but these hair-like structures also extend away from the neuron and can interact with hairs from other neurons. To be fair, this is Penrose's second attempt at explaining how the brain can perform functions that Turing machines cannot. His first hypothesis was proven impossible and so he has tried again but the issue remains that even if this is not the mechanism of the brain's non-classical behavior that there must be some such mechanism.

So, in summary, I believe that no attempt at artificial consciousness will succeed unless classical computation is merged with quantum effects. I'll go on record as saying that Turing machines are not capable of consciousness.

[mike777]

What test do you propose a human can pass and a machine cannot in the next 50 years? Will the test prove conscious or which is more intelligent?

IF you have a blind test on which subject can speak a language, read a roadmap and hold a press conference and you cannot tell which is which is that a fair test?

[luke warm]

Quote

What test do you propose a human can pass and a machine cannot in the next 50 years?

imagination... i can imagine being in a boat on a river, with tangerine trees and marmalade skies... i can imagine seeing a woman with diamonds for eyes

Quote

Will the test prove conscious or which is more intelligent?

conscious... and intelligence, if from a rationally functioning mind

[mike777]

luke warm, on Jul 25 2006, 04:26 PM, said:

Quote

What test do you propose a human can pass and a machine cannot in the next 50 years?

imagination... i can imagine being in a boat on a river, with tangerine trees and marmalade skies... i can imagine seeing a woman with diamonds for eyes

Quote

Will the test prove conscious or which is more intelligent?

conscious... and intelligence, if from a rationally functioning mind

ok and how did you prove you passed this test?

[DrTodd13]

What you are describing is called the Turing test. If a group of humans "talking" to a machine and a human via a computer terminal are incapable of saying which participant is human and which is computer then the machine can be said to possess intelligence. This will get vague very quickly. There are already machines in specific domains that can "learn" and remember things but only things for which they have been apriori programmed to be able to detect and learn. True intelligence could be defined as the ability to learn in any subject domain. For me, the interesting question is not one of intelligence but one of sentience. There will be robots who clean our homes and they will have to adapt to new furniture, etc.

How about this as a test. If there is a general purpose learning machine that is capable of learning from any subject domain and I can spontaneously tell it that I am going to destroy it and that machine responds by pleading with me not to do it and can explain that it is self-aware and why it believes it shouldn't be destroyed then I'll say the thing is sentient.

[hrothgar]

DrTodd13, on Jul 26 2006, 01:29 AM, said:

How about this as a test.  If there is a general purpose learning machine that is capable of learning from any subject domain and I can spontaneously tell it that I am going to destroy it and that machine responds by pleading with me not to do it and can explain that it is self-aware and why it believes it shouldn't be destroyed then I'll say the thing is sentient.

What if it takes over Skynet and nukes the Russians instead?

[mike777]

Well robots have been involved in at least 70 human deaths. No one is claiming self defense yet. The robots have said nothing so far. .

It does seem as robots become more intelligent, leaving aside the discussion of conscious, they may start to get some rights? Will we legally marry it with full marriage rights? Full parental rights including decisions on health care for the kids and spending money?

Can we at least agree there is something called AI right now in our cars and fridges and the only issue is how intelligent it will become and how we will measure it?

[DrTodd13]

Glad to see you haven't lost your sense of humor.

It is a tricky area. Theoretically, one could program a computer to understand enough language to know when it's existence were threatened and then program it to respond to try to prevent that in certain ways. Externally, you couldn't know whether the defense mechanisms were pre-programmed or represented originality of thought. You could even program it to plead for its own life with something that a human might say so that is why a clever interrogator would be necessary to see if the machine really understands what it means to be self-aware.

[Gerben42]

Quote

How about this as a test.  If there is a general purpose learning machine that is capable of learning from any subject domain and I can spontaneously tell it that I am going to destroy it and that machine responds by pleading with me not to do it and can explain that it is self-aware and why it believes it shouldn't be destroyed then I'll say the thing is sentient.

I'm sorry Dave. I can't let you do that...

[mike777]

DrTodd13, on Jul 25 2006, 05:49 PM, said:

Glad to see you haven't lost your sense of humor.

It is a tricky area. .......... might say so that is why a clever interrogator would be necessary to see if the machine really understands what it means to be self-aware.

IF it takes a clever trained interrogator will it really matter? That means 99.9999% of the rest of us cannot tell?

[mikeh]

mike777, on Jul 25 2006, 04:07 PM, said:

What test do you propose a human can pass and a machine cannot in the next 50 years? Will the test prove conscious or which is more intelligent?

IF you have a blind test on which subject can speak a language, read a roadmap and hold a press conference and you cannot tell which is which is that a fair test?

A lot of politicians can do at least two of these three things.. so your test will be inadequate unless it demands more as proof of intelligence

[luke warm]

Gerben42, on Jul 25 2006, 05:51 PM, said:

Quote

How about this as a test.  If there is a general purpose learning machine that is capable of learning from any subject domain and I can spontaneously tell it that I am going to destroy it and that machine responds by pleading with me not to do it and can explain that it is self-aware and why it believes it shouldn't be destroyed then I'll say the thing is sentient.

I'm sorry Dave. I can't let you do that...

easy hal, easy

mike said:

ok and how did you prove you passed this test?

when someone asks why i didn't complete a certain task, i'd tell them i was daydreaming, and describe the dream... now it might be possible to program a machine to do the same thing, but it seems to me that would defeat at least part of the purpose

[helene_t]

How consciousness should be defined is a major source of controversy among neuroscientists, philosofers, computer scientists and others who claim to have an informed opinion about the subject. I bought Stephen Pinkers "How the mind works" in the hope of finding an answer, but the best he could come up with was that maybe the whole issue is meaningless or at least non-scientific, or maybe it's so inherently difficult that we can never understand the answer even if some super-intelligent extra-terristial told it to us. It's very hard (yet still interesting) to discuss the possibility of artificial consciousness when we can't even agree what "consciousness" is and if it exists at all.

Personally, I feel that consciousness is over-rated. Most seem-to-be consciousness reasoning is post-hoc rationalization.

From an evolutionary-psycological point of view, it's quite obvious that the "theory of the psyche" and other mental mechanisms related to consciousness serve some purposes for our survival:
- Empathy is important for ethical behavior, teaching, and social behavior in general
- The theory of the "free will" is important in moral systems.
- Post-hoc rationalization is important in communication: you can explain how you reached a conclusion by reasoning, not how you reached it through intuition. This helps your audience to infer the reach of the scope of your conclusion so they can see if it's relevant for their purpose.
- Sentience related to awards such as food, sex and social recognition may make award-based learning more flexible (this is a somewhat vague idea that I have, I might be wrong).

Now for the easy, down-to-Earth part:
- Machines will superceed human capapabilities in more and more areas. They allready play better chess, and some would say that they conduct better psycho-analytical interviews, than the best human experts. It will be exceedingly difficult to find an area in which humans are still superior.
- There will never be a market for a machine that mimics a human brain, because it learns too slowly. It takes about twenty-five years of child-raising and education to train a skilled mind-worker with all the human qualities comming from unique childhood experiences, including stimuli of five difference senses, interaction with hundreds of different other people etc. Even if you could by a human-brain-capacity computer for a penny, you can't afford the time it wouyld take to collect all the input it would need to become human. Therefore, allthough computers will be more and more all-purpose, computer software will continue to be very specialized compared to human minds.

[mike777]

"Even if you could by a human-brain-capacity computer for a penny, you can't afford the time it wouyld take to collect all the input it would need to become human. Therefore, allthough computers will be more and more all-purpose, computer software will continue to be very specialized compared to human minds."


Why do you assume it would cost more than one penny and take more than one second to download all this information? Why assume only a human brain capacity, why not a billion or trillion times more?

There are already predictions of capacity exceeding the entire human race in one computer by 2050?

[helene_t]

mike777, on Jul 26 2006, 10:20 AM, said:

Why do you assume it would cost more than one penny and take more than one second to download all this information?

You may be right. If we could equip a new-born baby with a recorder that collects all sensory stimuli, after some 25 years we could make the collected information available on the net so mind-ware developers could use it for training there super-human computers, which learns the same as the human does but with superior memory, arithmetic skills and many other superior capabilities.

Then again, every human will be connected to the internet in the same way as machines are: we don't need to remember anything since everything is available on the internet, and if we want to do some reasoning that is beyond our own capabilities, we can just ask some web-server to do it for us.

Maybe humans will still enjoy the advantage of being able to keeping secrets. The owner of a computer will probably choose to make the computer's resources available to himself and he may choose to make some of them available to others. If the computer becomes an all-purpose personal assistent, he may choose to make his own brain's resources available to the computer. When this happens their will be no distinction between human and machine, they will have fused. But this seems a technically difficult thing to do. I think the computers will continue to enhance our consciousness for a long time before they will start developing consciousness themselves.

[Gerben42]

I predict that as usual the time scale of all these things is vastly underestimated. Those who have been alive in the 1960s would have thought that in 2000 we would have a Moon colony, household robots, flying cars, etc.

Guess what: all wrong. What NO ONE guessed was the internet and information technology booming. So I suggest we stop listening to outlandish predictions and just wait and see

Expect the unexpected.

[mike777]

Gerben42, on Jul 26 2006, 04:39 AM, said:

I predict that as usual the time scale of all these things is vastly underestimated. Those who have been alive in the 1960s would have thought that in 2000 we would have a Moon colony, household robots, flying cars, etc.

Guess what: all wrong. What NO ONE guessed was the internet and information technology booming. So I suggest we stop listening to outlandish predictions and just wait and see

Expect the unexpected.

We have household robots and we have flying cars!
Some of these household robots are cheap, some are expensive but we have them. Yes there are flying cars!
Yes we have a space colony. We call it the space station.

Yes, things like the internet were predicted in the sixties..for pete's sakes I used things like the internet in the sixties! We chatted and sent messages and played D and D and other stuff back then on it.

I used information technology in the sixties!

[whereagles]

hum.. there's a book by António Damasio, "The Feeling of What Happens: Body and Emotion in the Making of Consciousness," where he explains what counciousness is (haven't read it though). I guess if you apply his definitions to a computer you could decide whether it is councious or not

[Rain]

What about, will humans become more machine like?

I think that's more likely. To imagine a growing number of humans wishing to embed technology into themselves to enhance some human facilities. Its already happening. Isn't pacemaker just a machine inserted to help a failing heart? Inserted artificial kidneys?

Not that far fetched to foresee memory chips being inserted to foster a breed of super duper humans who remembers everything. Super bridge players! Or strength circuits inserted around arms to give it 20x more strength. Or telescopic eyes or enhanced hearing.

Cyborg :>