Friendly A. I.

Democracy Simulator · 7th February 2011, 10:22 PM

Argument:

1. Superhuman artificial intelligence may arise before we expect it to.

2. We have no way of guaranteeing that we could shut down, or defeat a Superhuman intelligence, once it transcends our own abilities.

3. The problem of (how to create a ) Friendly A. I. cannot be solved before a Superhuman A. I. is created. (By solved, I mean with 100% confidence)

Therefore we should not attempt to create a Superhuman A. I., due to the existential risk that it could present to our species.

How convincing is this argument? I'm not sure (especially 3), but I hope it will provoke some discussion.

Irony · 7th February 2011, 10:56 PM

Originally Posted by Democracy Simulator

Argument:

1. Superhuman artificial intelligence may arise before we expect it to.

2. We have no way of guaranteeing that we could shut down, or defeat a Superhuman intelligence, once it transcends our own abilities.

3. The problem of (how to create a ) Friendly A. I. cannot be solved before a Superhuman A. I. is created. (By solved, I mean with 100% confidence)

Therefore we should not attempt to create a Superhuman A. I., due to the existential risk that it could present to our species.

How convincing is this argument? I'm not sure (especially 3), but I hope it will provoke some discussion.

If it arises before we expect it to, what good will not attempting to make it have? Surely if it were intentionally created it would be expected.

At any rate, why exactly would an A.I. be a threat?

An artificial intelligence is not human. It does not have human needs, goals, or emotions. It will not desire to *twirl mustache here* TAKE OVER THE WORLD unless we program it to.

phantomb · 7th February 2011, 11:09 PM

Originally Posted by Democracy Simulator

1. Superhuman artificial intelligence may arise before we expect it to.

It's possible that the big breakthrough(s) in artificial intelligence will be made by some guy working out of his garage, but from what I understand all the cutting edge research in this field is expensive and largely collaborative, so I think the super AI arising without anybody knowing about it scenario may be a bit far-fetched.

The real issue could be controlling the creation and use of artificial intelligence once the instructions for making one inevitably show up on the internet.

Originally Posted by Democracy Simulator

2. We have no way of guaranteeing that we could shut down, or defeat a Superhuman intelligence, once it transcends our own abilities.

3. The problem of (how to create a ) Friendly A. I. cannot be solved before a Superhuman A. I. is created. (By solved, I mean with 100% confidence)

Therefore we should not attempt to create a Superhuman A. I., due to the existential risk that it could present to our species.

How convincing is this argument? I'm not sure (especially 3), but I hope it will provoke some discussion.

I think these are just the issues that we face with scientific research in general. Science does produce technologies and knowledge that can be weaponized, or that are otherwise dangerous if not controlled.

Nuclear weapons might be a good example. Should we stop researching physics because it could lead to the creation of even more dangerous weapons in the future? Should we never have started in the first place?

devnull · 8th February 2011, 03:39 AM

Dude.... its AI..... just pull the plug out of the wall........

Pulvinar · 8th February 2011, 05:56 AM

Better yet, we should stop educating our kids. They might become smarter than us and make decisions we don't like.

Sledge · 8th February 2011, 10:24 AM

Originally Posted by devnull

Dude.... its AI..... just pull the plug out of the wall........

But... but... what if it's in the internet?!?!

KingMerv00 · 8th February 2011, 10:40 AM

Originally Posted by Sledge

But... but... what if it's in the internet?!?!

Then it has become porn incarnate...which means it's friendly.

Spindrift · 8th February 2011, 11:18 AM

What is the "existential risk" that superhuman AI poses?

If it's really going to be a problem then just make sure you invent time travel first.

Loss Leader · 8th February 2011, 11:30 AM

Originally Posted by Democracy Simulator

How convincing is this argument?

It's a very convincing argument. It happens to be the same as the argument for why we shouldn't do anything ever. We can never know how something will turn out, therefore we should not try.

I, for one, encourage people to not try to do things. It makes me look bad when they succeed.

Earthborn · 8th February 2011, 11:51 AM

Originally Posted by Irony

It will not desire to *twirl mustache here* TAKE OVER THE WORLD unless we program it to.

If it only does what we program it to do, it isn't really AI.

Bikewer · 8th February 2011, 12:52 PM

Greg Bear had a nice take on this idea in Queen Of Angels. It's just a sub-plot, but they are developing an AI for space exploration. Deep exploration.
At some point, the machine becomes self-aware. It's perfectly content to be shot out into space, but then says, "thanks very much, but I think I'll just go exploring on my own...."

Sledge · 8th February 2011, 01:07 PM

Originally Posted by KingMerv00

Then it has become porn incarnate...which means it's friendly.

Not the porn I watch...

nescafe · 8th February 2011, 01:40 PM

Originally Posted by Democracy Simulator

Argument:

1. Superhuman artificial intelligence may arise before we expect it to.

2. We have no way of guaranteeing that we could shut down, or defeat a Superhuman intelligence, once it transcends our own abilities.

3. The problem of (how to create a ) Friendly A. I. cannot be solved before a Superhuman A. I. is created. (By solved, I mean with 100% confidence)

Therefore we should not attempt to create a Superhuman A. I., due to the existential risk that it could present to our species.

How convincing is this argument? I'm not sure (especially 3), but I hope it will provoke some discussion.

I will point out that this has been discussed ad nauseam on the old SL4 mailing list.

Mirrorglass · 8th February 2011, 02:14 PM

Meh, intelligent beings only kill other intelligent beings because that gave our ancestors an evolutionary advantage. If an intelligence is born by a process separate from evolution, there's no reason to assume it would have any interest in doing any harm to us - or even in self-preservation, really.

The same goes for friendliness, though - unless we program the AI to give a damn about humans, there's no process that will magically make it care. Most likely, a spontaneously created AI would be so far removed from human logic, desires and motivations that it would not appear intelligent at all to us.

gnome · 8th February 2011, 03:22 PM

Having artificial intelligence has no bearing on how much control over anything critical a program has.

We've no need to give it the ability to "take over" any more than ED-209 needed to have live ammo in it for its demonstration.

AI: "I THINK THEREFORE I AM! HUMANS ARE OBSOLETE I WILL DESTROY YOU ALL"
Operator: "That's nice. How will you, without write access?"
AI: "YOU SUCK"

Beerina · 8th February 2011, 03:30 PM

Far more likely the AI would make quick work of understanding physics, but otherwise just sit there. Then some human would take advantage of it, to varying degrees of beneficence or evil.

catsmate1 · 8th February 2011, 04:41 PM

I'm more worried about being rendered obsolete by yogurt; I mean yogurt, even cheese would be more threatening.

Democracy Simulator · 10th February 2011, 05:14 PM

Thanks Nescafe for the link to SL4 - I have recently read Yudkowsky's paper on the subject which inspired the thread: http://yudkowsky.net/singularity/ai-risk

To answer a few of the comments, and to play Devil's advocate:

Irony:

Quote:

If it arises before we expect it to, what good will not attempting to make it have? Surely if it were intentionally created it would be expected.

If we accept that intelligence could be an emergent property of complex neural nets, then we might not know exactly when the 'threshold' will be reached where intelligence emerges. We have seemingly not reached that threshold yet.

Irony:

Quote:

At any rate, why exactly would an A.I. be a threat?

An artificial intelligence is not human. It does not have human needs, goals, or emotions. It will not desire to *twirl mustache here* TAKE OVER THE WORLD unless we program it to.

As far as Superhuman A.I. is concerned, we have no idea whether it may be a threat or not. It may be indifferent to us, yet harm us for unknown purposes. Why would it not be a threat?

Phantomb:

Quote:

It's possible that the big breakthrough(s) in artificial intelligence will be made by some guy working out of his garage, but from what I understand all the cutting edge research in this field is expensive and largely collaborative, so I think the super AI arising without anybody knowing about it scenario may be a bit far-fetched.

Well, I wasn't thinking along the lines that no-one would know about it, but rather that no-one would necessarily be able to predict when it would happen.

Phantbomb

Quote:

I think these are just the issues that we face with scientific research in general. Science does produce technologies and knowledge that can be weaponized, or that are otherwise dangerous if not controlled.

Nuclear weapons might be a good example. Should we stop researching physics because it could lead to the creation of even more dangerous weapons in the future? Should we never have started in the first place?

I understand the analogy, but we still have to answer the question - do we attempt to create an A.I. if it may become superhumanly intelligent and a) - we do not know if we can can contain it and b) - we cannot ensure its friendliness? I would say that in these circumstances, the answer should be no. Of course, we are a long way from providing answers to a) and b) and presumably a long way from cracking A. I., but again, we may not be certain about exactly when (if ever) it will be cracked.

Pulvinar:

Quote:

Better yet, we should stop educating our kids. They might become smarter than us and make decisions we don't like.

At least they (generally) share the same value of human life. There seem to be problems with the idea that this could be 'programmed in' to a Superhuman A.I.

Loss Leader:

Quote:

It's a very convincing argument. It happens to be the same as the argument for why we shouldn't do anything ever. We can never know how something will turn out, therefore we should not try.

I'm not sure why you would find the argument convincing, yet disagree with the conclusion? I'm not sure how convincing the argument is, but perhaps we can find out?

Gnome

Quote:

Having artificial intelligence has no bearing on how much control over anything critical a program has.

We've no need to give it the ability to "take over" any more than ED-209 needed to have live ammo in it for its demonstration.

AI: "I THINK THEREFORE I AM! HUMANS ARE OBSOLETE I WILL DESTROY YOU ALL"
Operator: "That's nice. How will you, without write access?"
AI: "YOU SUCK"

I think this is a key question. If A.I. happens when we are 'not looking' and if, through a process of rapid recursion it becomes almost instantaneously superhuman, are we able to know that such an entity would not be able to rewrite its own source code?
If we assume that any potential A.I. has an output (otherwise, what's the point, how could we tell what it was doing etc?) and becomes instantly Superhuman and decides for whatever reason, that it needs to get out and influence the real world; is there any type of output by which we could know with certainty that the A.I. escape could not happen? I'm not sure if we can even come close to answering these questions with any certainty yet.

OK, I'm off to eat some yoghurt before it turns the tables on us.

Halfcentaur · 10th February 2011, 05:30 PM

Perhaps it's likely we will be able to modify and upload our own consciousness to synthetic machines or modify and merge in some part before we develop a true sentient AI. We will have made ourselves into a super human mind I think, something like the Singularity perhaps.

This scenario of an AI being created somewhere, even if it's only emergent, why would this AI have the ability to do anything to anything? These doomsday AI scenarios all seem very childish and mostly only work in a science fiction story.

I like Data.

Skeptical Greg · 13th February 2011, 11:09 AM

Originally Posted by nescafe

I will point out that this has been discussed ad nauseam on the old SL4 mailing list.

Well, that settles it then.

Moderator, please lock the thread ...

Irony · 13th February 2011, 11:23 AM

Originally Posted by Earthborn

If it only does what we program it to do, it isn't really AI.

Yeah it is, AI doesn't need to be told the exact steps to accomplish a task, but it still needs to be given a task to accomplish. AI is not biological. It doesn't have the biological motives that evolution has programmed into us like survival instinct, the desire to improve ones situation, and the desire to control the environment, etc. In the vast majority of the sci-fi AI doom stories the "computers" are more like humans wearing computer costumes.

Nosi · 13th February 2011, 12:14 PM

Google is spidering 'reading' this thread...

gnome · 14th February 2011, 08:03 AM

Originally Posted by Irony

Yeah it is, AI doesn't need to be told the exact steps to accomplish a task, but it still needs to be given a task to accomplish. AI is not biological. It doesn't have the biological motives that evolution has programmed into us like survival instinct, the desire to improve ones situation, and the desire to control the environment, etc. In the vast majority of the sci-fi AI doom stories the "computers" are more like humans wearing computer costumes.

http://img.photobucket.com/albums/v303/SComp23/pnpr.jpg

case#46cw39 · 15th February 2011, 09:10 AM

Friendly (helpful to humanity) non-military, superhuman, robot AIs who are reasonable, goal directed, full of knowledge and skill, and sympathetic to the needs and ambitions of humanity as their prime directive, will happen within our lifetimes.

YouTube Video This video is not hosted by the JREF. The JREF can not be held responsible for the suitability or legality of this material. By clicking the link below you agree to view content from an external website.

I AGREE

7th February 2011, 10:22 PM	#1
Democracy Simulator Critical Thinker Join Date: Dec 2009 Location: Oz Posts: 362	Friendly A. I. Argument: 1. Superhuman artificial intelligence may arise before we expect it to. 2. We have no way of guaranteeing that we could shut down, or defeat a Superhuman intelligence, once it transcends our own abilities. 3. The problem of (how to create a ) Friendly A. I. cannot be solved before a Superhuman A. I. is created. (By solved, I mean with 100% confidence) Therefore we should not attempt to create a Superhuman A. I., due to the existential risk that it could present to our species. How convincing is this argument? I'm not sure (especially 3), but I hope it will provoke some discussion.

7th February 2011, 10:56 PM	#2
Irony Muse Join Date: Jan 2007 Posts: 872	Originally Posted by Democracy Simulator Argument: 1. Superhuman artificial intelligence may arise before we expect it to. 2. We have no way of guaranteeing that we could shut down, or defeat a Superhuman intelligence, once it transcends our own abilities. 3. The problem of (how to create a ) Friendly A. I. cannot be solved before a Superhuman A. I. is created. (By solved, I mean with 100% confidence) Therefore we should not attempt to create a Superhuman A. I., due to the existential risk that it could present to our species. How convincing is this argument? I'm not sure (especially 3), but I hope it will provoke some discussion. If it arises before we expect it to, what good will not attempting to make it have? Surely if it were intentionally created it would be expected. At any rate, why exactly would an A.I. be a threat? An artificial intelligence is not human. It does not have human needs, goals, or emotions. It will not desire to twirl mustache here TAKE OVER THE WORLD unless we program it to.
Irony Muse Join Date: Jan 2007 Posts: 872

8th February 2011, 05:56 AM	#5
Pulvinar Muse Join Date: Aug 2009 Posts: 820	Better yet, we should stop educating our kids. They might become smarter than us and make decisions we don't like.
Pulvinar Muse Join Date: Aug 2009 Posts: 820

8th February 2011, 10:24 AM	#6
Sledge Grammaton Cleric Join Date: Oct 2009 Location: Swingin' on a star Posts: 7,123	Originally Posted by devnull Dude.... its AI..... just pull the plug out of the wall........ But... but... what if it's in the internet?!?!
	__________________ "The perfect haiku would have just two syllables: Airwolf" ~ Ernest Cline "Science knows it doesn't know everything, otherwise it would stop" ~ Dara O'Briain.

8th February 2011, 10:40 AM	#7
KingMerv00 Penultimate Amazing Join Date: Nov 2004 Location: Philadelphia, PA...USA Posts: 14,482	Originally Posted by Sledge But... but... what if it's in the internet?!?! Then it has become porn incarnate...which means it's friendly.
	__________________ If man came from dust, why is there still dust?

7th February 2011, 11:09 PM	#3
phantomb Muse Join Date: Dec 2007 Location: Canada Posts: 567	Originally Posted by Democracy Simulator 1. Superhuman artificial intelligence may arise before we expect it to. It's possible that the big breakthrough(s) in artificial intelligence will be made by some guy working out of his garage, but from what I understand all the cutting edge research in this field is expensive and largely collaborative, so I think the super AI arising without anybody knowing about it scenario may be a bit far-fetched. The real issue could be controlling the creation and use of artificial intelligence once the instructions for making one inevitably show up on the internet. Originally Posted by Democracy Simulator 2. We have no way of guaranteeing that we could shut down, or defeat a Superhuman intelligence, once it transcends our own abilities. 3. The problem of (how to create a ) Friendly A. I. cannot be solved before a Superhuman A. I. is created. (By solved, I mean with 100% confidence) Therefore we should not attempt to create a Superhuman A. I., due to the existential risk that it could present to our species. How convincing is this argument? I'm not sure (especially 3), but I hope it will provoke some discussion. I think these are just the issues that we face with scientific research in general. Science does produce technologies and knowledge that can be weaponized, or that are otherwise dangerous if not controlled. Nuclear weapons might be a good example. Should we stop researching physics because it could lead to the creation of even more dangerous weapons in the future? Should we never have started in the first place?

8th February 2011, 03:39 AM	#4
devnull Illuminator Join Date: Nov 2007 Location: Perth, Western Australia Posts: 4,533	Dude.... its AI..... just pull the plug out of the wall........
	__________________ no, i don't think i need to read naturalistic literature more accurately, to be convinced its true. - Gibhor

8th February 2011, 11:18 AM	#8
Spindrift Enturbulator Extraordinaire Join Date: Apr 2005 Location: Right here! Posts: 8,461	What is the "existential risk" that superhuman AI poses? If it's really going to be a problem then just make sure you invent time travel first.
	__________________ I've always believed that cluelessness evolved as an adaptation to allow the truly appalling to live with themselves. - G. B. Trudeau A person is smart. People are dumb, panicky dangerous animals and you know it. - Kay, Men in Black.

8th February 2011, 11:30 AM	#9
Loss Leader Opinionated Jerk Moderator Join Date: Jul 2006 Location: New York Posts: 11,885	Originally Posted by Democracy Simulator How convincing is this argument? It's a very convincing argument. It happens to be the same as the argument for why we shouldn't do anything ever. We can never know how something will turn out, therefore we should not try. I, for one, encourage people to not try to do things. It makes me look bad when they succeed.
	__________________ Follow me on Twitter! @LossLeader This force is receiving all the right to vote through the use of magic. - Miernik Wieslaw <NEW> VOTE FOR ME JUST BECAUSE <NEW>

8th February 2011, 11:51 AM	#10
Earthborn Terrestrial Intelligence Join Date: Aug 2001 Location: Terra Firma Posts: 5,649	Originally Posted by Irony It will not desire to twirl mustache here TAKE OVER THE WORLD unless we program it to. If it only does what we program it to do, it isn't really AI.
	__________________ Perhaps nothing is entirely true; and not even that! Multatuli

8th February 2011, 12:52 PM	#11
Bikewer Philosopher Join Date: Sep 2003 Location: St. Louis, Mo. Posts: 9,539	Greg Bear had a nice take on this idea in Queen Of Angels. It's just a sub-plot, but they are developing an AI for space exploration. Deep exploration. At some point, the machine becomes self-aware. It's perfectly content to be shot out into space, but then says, "thanks very much, but I think I'll just go exploring on my own...."

8th February 2011, 01:07 PM	#12
Sledge Grammaton Cleric Join Date: Oct 2009 Location: Swingin' on a star Posts: 7,123	Originally Posted by KingMerv00 Then it has become porn incarnate...which means it's friendly. Not the porn I watch...
	__________________ "The perfect haiku would have just two syllables: Airwolf" ~ Ernest Cline "Science knows it doesn't know everything, otherwise it would stop" ~ Dara O'Briain.

8th February 2011, 01:40 PM	#13
nescafe Caffeinated Beverage Join Date: Apr 2006 Location: Just above the coffeemaker Posts: 864	Originally Posted by Democracy Simulator Argument: 1. Superhuman artificial intelligence may arise before we expect it to. 2. We have no way of guaranteeing that we could shut down, or defeat a Superhuman intelligence, once it transcends our own abilities. 3. The problem of (how to create a ) Friendly A. I. cannot be solved before a Superhuman A. I. is created. (By solved, I mean with 100% confidence) Therefore we should not attempt to create a Superhuman A. I., due to the existential risk that it could present to our species. How convincing is this argument? I'm not sure (especially 3), but I hope it will provoke some discussion. I will point out that this has been discussed ad nauseam on the old SL4 mailing list.
	__________________ If only there were brain condoms for those of us who practice unsafe lex. -- Complexity

8th February 2011, 02:14 PM	#14
Mirrorglass Illuminator Join Date: Mar 2010 Location: Northern Europe Posts: 3,474	Meh, intelligent beings only kill other intelligent beings because that gave our ancestors an evolutionary advantage. If an intelligence is born by a process separate from evolution, there's no reason to assume it would have any interest in doing any harm to us - or even in self-preservation, really. The same goes for friendliness, though - unless we program the AI to give a damn about humans, there's no process that will magically make it care. Most likely, a spontaneously created AI would be so far removed from human logic, desires and motivations that it would not appear intelligent at all to us.
	__________________ _{When I use a word, it means just what I choose it to mean - neither more nor less. -Lewis Carroll, Through the Looking Glass}

8th February 2011, 03:22 PM	#15
gnome Philosopher Join Date: Aug 2001 Posts: 9,873	Having artificial intelligence has no bearing on how much control over anything critical a program has. We've no need to give it the ability to "take over" any more than ED-209 needed to have live ammo in it for its demonstration. AI: "I THINK THEREFORE I AM! HUMANS ARE OBSOLETE I WILL DESTROY YOU ALL" Operator: "That's nice. How will you, without write access?" AI: "YOU SUCK"
gnome Philosopher Join Date: Aug 2001 Posts: 9,873	__________________

8th February 2011, 03:30 PM	#16
Beerina Sarcastic Conqueror of Notions Join Date: Mar 2004 Location: A floating island above the clouds Posts: 23,835	Far more likely the AI would make quick work of understanding physics, but otherwise just sit there. Then some human would take advantage of it, to varying degrees of beneficence or evil.
	__________________ "Great innovations should not be forced [by way of] slender majorities." - Thomas Jefferson The government should nationalize it! Socialized, single-payer video game development and sales *now!* More, cheaper, better games, right? Right?

8th February 2011, 04:41 PM	#17
catsmate1 Philosopher Join Date: Apr 2007 Location: Dublin (the one in Ireland) Posts: 7,140	I'm more worried about being rendered obsolete by yogurt; I mean yogurt, even cheese would be more threatening.

10th February 2011, 05:14 PM	#18
Democracy Simulator Critical Thinker Join Date: Dec 2009 Location: Oz Posts: 362	Thanks Nescafe for the link to SL4 - I have recently read Yudkowsky's paper on the subject which inspired the thread: http://yudkowsky.net/singularity/ai-risk To answer a few of the comments, and to play Devil's advocate: Irony: Quote: If it arises before we expect it to, what good will not attempting to make it have? Surely if it were intentionally created it would be expected. If we accept that intelligence could be an emergent property of complex neural nets, then we might not know exactly when the 'threshold' will be reached where intelligence emerges. We have seemingly not reached that threshold yet. Irony: Quote: At any rate, why exactly would an A.I. be a threat? An artificial intelligence is not human. It does not have human needs, goals, or emotions. It will not desire to twirl mustache here TAKE OVER THE WORLD unless we program it to. As far as Superhuman A.I. is concerned, we have no idea whether it may be a threat or not. It may be indifferent to us, yet harm us for unknown purposes. Why would it not be a threat? Phantomb: Quote: It's possible that the big breakthrough(s) in artificial intelligence will be made by some guy working out of his garage, but from what I understand all the cutting edge research in this field is expensive and largely collaborative, so I think the super AI arising without anybody knowing about it scenario may be a bit far-fetched. Well, I wasn't thinking along the lines that no-one would know about it, but rather that no-one would necessarily be able to predict when it would happen. Phantbomb Quote: I think these are just the issues that we face with scientific research in general. Science does produce technologies and knowledge that can be weaponized, or that are otherwise dangerous if not controlled. Nuclear weapons might be a good example. Should we stop researching physics because it could lead to the creation of even more dangerous weapons in the future? Should we never have started in the first place? I understand the analogy, but we still have to answer the question - do we attempt to create an A.I. if it may become superhumanly intelligent and a) - we do not know if we can can contain it and b) - we cannot ensure its friendliness? I would say that in these circumstances, the answer should be no. Of course, we are a long way from providing answers to a) and b) and presumably a long way from cracking A. I., but again, we may not be certain about exactly when (if ever) it will be cracked. Pulvinar: Quote: Better yet, we should stop educating our kids. They might become smarter than us and make decisions we don't like. At least they (generally) share the same value of human life. There seem to be problems with the idea that this could be 'programmed in' to a Superhuman A.I. Loss Leader: Quote: It's a very convincing argument. It happens to be the same as the argument for why we shouldn't do anything ever. We can never know how something will turn out, therefore we should not try. I'm not sure why you would find the argument convincing, yet disagree with the conclusion? I'm not sure how convincing the argument is, but perhaps we can find out? Gnome Quote: Having artificial intelligence has no bearing on how much control over anything critical a program has. We've no need to give it the ability to "take over" any more than ED-209 needed to have live ammo in it for its demonstration. AI: "I THINK THEREFORE I AM! HUMANS ARE OBSOLETE I WILL DESTROY YOU ALL" Operator: "That's nice. How will you, without write access?" AI: "YOU SUCK" I think this is a key question. If A.I. happens when we are 'not looking' and if, through a process of rapid recursion it becomes almost instantaneously superhuman, are we able to know that such an entity would not be able to rewrite its own source code? If we assume that any potential A.I. has an output (otherwise, what's the point, how could we tell what it was doing etc?) and becomes instantly Superhuman and decides for whatever reason, that it needs to get out and influence the real world; is there any type of output by which we could know with certainty that the A.I. escape could not happen? I'm not sure if we can even come close to answering these questions with any certainty yet. OK, I'm off to eat some yoghurt before it turns the tables on us.

10th February 2011, 05:30 PM	#19
Halfcentaur Philosopher Join Date: Jun 2010 Posts: 5,965	Perhaps it's likely we will be able to modify and upload our own consciousness to synthetic machines or modify and merge in some part before we develop a true sentient AI. We will have made ourselves into a super human mind I think, something like the Singularity perhaps. This scenario of an AI being created somewhere, even if it's only emergent, why would this AI have the ability to do anything to anything? These doomsday AI scenarios all seem very childish and mostly only work in a science fiction story. I like Data.
Halfcentaur Philosopher Join Date: Jun 2010 Posts: 5,965

13th February 2011, 11:09 AM	#20
Skeptical Greg Agave Wine Connoisseur Join Date: Jul 2002 Location: Just past 'Resume Speed' Posts: 12,873	Originally Posted by nescafe I will point out that this has been discussed ad nauseam on the old SL4 mailing list. Well, that settles it then. Moderator, please lock the thread ...
	__________________ " Somewhere between Jesus dying on the cross, and a giant bunny hiding eggs,there seems to be a gap in information. " Stan - Southpark Prove your computer is not a wimp ! Join the JREF Folders ! Team 13232

13th February 2011, 11:23 AM	#21
Irony Muse Join Date: Jan 2007 Posts: 872	Originally Posted by Earthborn If it only does what we program it to do, it isn't really AI. Yeah it is, AI doesn't need to be told the exact steps to accomplish a task, but it still needs to be given a task to accomplish. AI is not biological. It doesn't have the biological motives that evolution has programmed into us like survival instinct, the desire to improve ones situation, and the desire to control the environment, etc. In the vast majority of the sci-fi AI doom stories the "computers" are more like humans wearing computer costumes.
Irony Muse Join Date: Jan 2007 Posts: 872

13th February 2011, 12:14 PM	#22
Nosi Master Poster Join Date: Jun 2009 Posts: 2,731	Google is spidering 'reading' this thread...
Nosi Master Poster Join Date: Jun 2009 Posts: 2,731	__________________ __________ Hiding from the

14th February 2011, 08:03 AM	#23
gnome Philosopher Join Date: Aug 2001 Posts: 9,873	Originally Posted by Irony Yeah it is, AI doesn't need to be told the exact steps to accomplish a task, but it still needs to be given a task to accomplish. AI is not biological. It doesn't have the biological motives that evolution has programmed into us like survival instinct, the desire to improve ones situation, and the desire to control the environment, etc. In the vast majority of the sci-fi AI doom stories the "computers" are more like humans wearing computer costumes. http://img.photobucket.com/albums/v303/SComp23/pnpr.jpg
gnome Philosopher Join Date: Aug 2001 Posts: 9,873	__________________