What is so artificial about AI?

[Tango Alpha Delta]

For the above reasons, I've never liked the way I've seen AI's protrayed in fiction.  Why do they always have emotions?  Why do they go nuts?

Because if they behaved as designed and didn't go nuts there wouldn't be much of a story.   

No, you misunderstood the question: why do they always have emotions and go nuts IN REAL LIFE?

[stePH]

For the above reasons, I've never liked the way I've seen AI's protrayed in fiction.  Why do they always have emotions?  Why do they go nuts?

Because if they behaved as designed and didn't go nuts there wouldn't be much of a story.   

Exactly.  You can't have System Shock without SHODAN.

[ClintMemo]

I don't foresee an AI deciding to kill us off.  For en evil AI to survive, it only needs uninterrupted power (so it can stay running on whatever machine it runs on) and physical security (to keep people from switching it off). A back up system to use as an escape route would also be prudent.  So an AI wouldn't care about clean water or clean air. It doesn't need those. Consequently, as long as armies of us didn't try to break into it's home and switch it off, we wouldn't be a threat to it.

You're assuming it doesn't go insane.

Insanity is a brain malfunction.  For an AI to go insane, it would have to become just malfunctioned enough to be dangerous without breaking to the point of not being able to function at all - not a likely scenario.

At first I was going to refute this because scientists are using neural networks in their AI research.  Neural networks, like our own brains and the Internet, function very well after being damaged in isolated areas.

But there is a fundamental difference between the internet and our brain.  Most of the internet is functionally identical - servers which spit out content. You could eliminate all of them but one and that one would still work fine.  A very small number of the machines attached to it  - routers and such -  actually make it work. Take a few of these out and the whole thing collapses.
In our brain, every different part has its own purpose.  Yes, damage one part and another part might try and compensate, but most parts cannot be replaced.
Also, computers are digital. Our brains are not.

[Thaurismunths]

But there is a fundamental difference between the internet and our brain.  Most of the internet is functionally identical - servers which spit out content. You could eliminate all of them but one and that one would still work fine.  A very small number of the machines attached to it  - routers and such -  actually make it work. Take a few of these out and the whole thing collapses.
In our brain, every different part has its own purpose.  Yes, damage one part and another part might try and compensate, but most parts cannot be replaced.
Also, computers are digital. Our brains are not.

Aside from wanting to point out that that was your your 666th post, I wanted to ask if you could clarify what you meant.
I'm not quite following your objection.

[ClintMemo]

But there is a fundamental difference between the internet and our brain.  Most of the internet is functionally identical - servers which spit out content. You could eliminate all of them but one and that one would still work fine.  A very small number of the machines attached to it  - routers and such -  actually make it work. Take a few of these out and the whole thing collapses.
In our brain, every different part has its own purpose.  Yes, damage one part and another part might try and compensate, but most parts cannot be replaced.
Also, computers are digital. Our brains are not.

Aside from wanting to point out that that was your your 666th post, I wanted to ask if you could clarify what you meant.
I'm not quite following your objection.

666 - I hadn't noticed that. 

I'm just saying that the analogy of comparing our brain to the internet doesn't really hold up under enough scrutiny.

 Also, computers are digital.  No matter how sophisticated the software is that run on them, it still all comes down to representing things as 1's and 0's.  Our brains are not digital.  At some point in the future, it might be possible to simulate our brain on a computer, or a network of them, but it would still only be a simulation, not a duplication.  It would be like a digital image of the Mona Lisa.  It might look really good, but it isn't the same.  (Wait, now I'm using an analogy ).  Digital images and digital music are both very good, but we are along way from "digital thought process."   I'm not sure you could ever achieve consciousness on a computer. Perhaps the best you could ever achieve is "an amazing simulation of consciousness."
Now, you could argue that it could be close enough not to matter and maybe that is true, but it's still not the same thing.

Another thought just occurred to me.  People have been arguing for centuries as to whether humans are deterministic or not. A computer AI would be deterministic.  Given a certain situation, it would always respond a certain way. 

[Nobilis]

Given a certain situation, it would always respond a certain way. 

Ah, a Linux user then?

Seriously, though, not all computers are digital.  Look at some of the quantum computing, optical computing, and analog electronic computing experiments that are going on, and you'll find some different answers.

[Tango Alpha Delta]

I notice, too, that you say "given a certain situation" as though you can guarantee that ALL of the variables will be the same each time.  Perhaps the reason you can't see how "deterministic" we are is that our "processors" are affected by more factors than we are perceiving?

[ClintMemo]

Given a certain situation, it would always respond a certain way. 

Ah, a Linux user then?

No, just a meatball programmer (currently I do .net).

Seriously, though, not all computers are digital.  Look at some of the quantum computing, optical computing, and analog electronic computing experiments that are going on, and you'll find some different answers.

Quantum and optical computers are both digital - they store information as 1's and 0's. Quantum computers (if they are ever developed) would store their information bits at the atomic level, so you could have LOTS of bits in a small area.  Optical computers use light to transfer the information around inside the computer itself and would be very fast. Only analog computers (which have been around since the 1960's?) are not digital.  Analog and digital are essential antonyms.

[Tango Alpha Delta]

ClintMemo's description prodded my memory, and I briefly searched for clarification, but couldn't even understand the wiki article, so I'll just pose this as an "I thought quantum computing meant this" and let someone who understands it better shoot me down:

I thought digital computing was tied to binary (0's and 1's) because there are two electrical states; but quantum computing would be based on the six quark states (up, down, top, bottom, strange, and ...um... Doc).   So quantum computers would no longer be digital, per se.


On an unrelated side note, in college, my roommate and I both took an electronic music studio course in order to play with samplers, sequencers, drum machines, etc.  He happened to be a much better piano player than I was, despite the fact he is missing his index and middle fingers on his left hand.  Most of his music is digital - programmed on synthesizers - but when he plays something himself, he jokes about it being analog due to the missing digits.

And, of course, he always ends up having to explain what "analog" means.  (The price uber-geeks pay for our humor.)

[Heradel]

I thought digital computing was tied to binary (0's and 1's) because there are two electrical states; but quantum computing would be based on the six quark states (up, down, top, bottom, strange, and ...um... Doc).   So quantum computers would no longer be digital, per se.

Not exactly. Quantum means 1, 0, or a superposition of both. and they're called Qubits. And quantum computing acts at the atomic, not subatomic level. And it's magnetic states for bits on the hard drive, electrical on the flash/RAM. 

[ClintMemo]

I used the word "digital", but "binary" more be accurate (if it even makes a difference).

[Tango Alpha Delta]

I thought digital computing was tied to binary (0's and 1's) because there are two electrical states; but quantum computing would be based on the six quark states (up, down, top, bottom, strange, and ...um... Doc).   So quantum computers would no longer be digital, per se.

Not exactly. Quantum means 1, 0, or a superposition of both. and they're called Qubits. And quantum computing acts at the atomic, not subatomic level. And it's magnetic states for bits on the hard drive, electrical on the flash/RAM. 

By process of elimination, I guess that leaves me with stamp collecting. 

[wakela]

I'm not sure you could ever achieve consciousness on a computer. Perhaps the best you could ever achieve is "an amazing simulation of consciousness."
Now, you could argue that it could be close enough not to matter and maybe that is true, but it's still not the same thing.

From The Age of Spiritual Machines:
"Will computers be self aware?  I don't know.  But they will say they are, and we will treat them as if they are."

An AI doesn't have to be a simulation of a human mind.  Birds need feathers to fly.  Planes don't have feathers, but they fly anyway.  And they fly in a way more useful to us. 

[ClintMemo]

I'm not sure you could ever achieve consciousness on a computer. Perhaps the best you could ever achieve is "an amazing simulation of consciousness."
Now, you could argue that it could be close enough not to matter and maybe that is true, but it's still not the same thing.

From The Age of Spiritual Machines:
"Will computers be self aware?  I don't know.  But they will say they are, and we will treat them as if they are."

An AI doesn't have to be a simulation of a human mind.  Birds need feathers to fly.  Planes don't have feathers, but they fly anyway.  And they fly in a way more useful to us. 

...but again, that's not an "artificial consciousness" - it's just a really, really, really good deterministic user interface.

[Nobilis]

I notice, too, that you say "given a certain situation" as though you can guarantee that ALL of the variables will be the same each time.  Perhaps the reason you can't see how "deterministic" we are is that our "processors" are affected by more factors than we are perceiving?

The human brain is a "chaotic" system in that it is highly sensitive to initial conditions.

As such, it would probably be more realistically simulated with analog neural networks, which can display similar behavior.

[eytanz]

I'm not sure you could ever achieve consciousness on a computer. Perhaps the best you could ever achieve is "an amazing simulation of consciousness."
Now, you could argue that it could be close enough not to matter and maybe that is true, but it's still not the same thing.

From The Age of Spiritual Machines:
"Will computers be self aware?  I don't know.  But they will say they are, and we will treat them as if they are."

An AI doesn't have to be a simulation of a human mind.  Birds need feathers to fly.  Planes don't have feathers, but they fly anyway.  And they fly in a way more useful to us. 

...but again, that's not an "artificial consciousness" - it's just a really, really, really good deterministic user interface.

Wait, I'm not sure I understand this. Assuming that nebulous notions such as souls are left out of the equation (and if a soul is needed for consciousness then the no technological approach would ever work, unless someone figures out how to make artifical souls), then it is not clear to me why you would assume that only human minds are capable of having the necessary properties to seperate a "deterministic user interface" from "conciousness" (scare quotes because I'm not even sure the human mind is anything but a very good deterministic human interface once you look at it from the right level of granularity).

[ClintMemo]

I'm not sure you could ever achieve consciousness on a computer. Perhaps the best you could ever achieve is "an amazing simulation of consciousness."
Now, you could argue that it could be close enough not to matter and maybe that is true, but it's still not the same thing.

From The Age of Spiritual Machines:
"Will computers be self aware?  I don't know.  But they will say they are, and we will treat them as if they are."

An AI doesn't have to be a simulation of a human mind.  Birds need feathers to fly.  Planes don't have feathers, but they fly anyway.  And they fly in a way more useful to us. 

...but again, that's not an "artificial consciousness" - it's just a really, really, really good deterministic user interface.

Wait, I'm not sure I understand this. Assuming that nebulous notions such as souls are left out of the equation (and if a soul is needed for consciousness then the no technological approach would ever work, unless someone figures out how to make artifical souls), then it is not clear to me why you would assume that only human minds are capable of having the necessary properties to seperate a "deterministic user interface" from "conciousness" (scare quotes because I'm not even sure the human mind is anything but a very good deterministic human interface once you look at it from the right level of granularity).

I'm not assuming that only a human mind can achieve this, I just doubt it can be achieved with a binary based machine.   You also make a very important point.  If don't know exactly what consciousness is, how will we duplicate it  - or - know if we do?