1046: "Skynet"

[bobbbbbbby1234]

Anyone who's even spent 5 minutes meditating would be familiar with the sense of "breathing manually".

What was more interesting to me about this comic was the sense of the subject becoming the object of the subject at the next level of development, a concept I've encountered before in developmental psychology, and the writings of Ken Wilber. http://www.enlightennext.org/magazine/j ... asp?page=2

[wideyes]

Bravo for this comic. I always took issue with the assumption that a self-aware computer's first thought would be to eradicate humanity. I think this assumption says more about humanity than it does about computers or self-awareness.

[Rotherian]

Worst disguise for a Terminator ever. A Skynet T-shirt? That's got to be a giveaway.

I'm just waiting for a Terminator / Dr. Who crossover, as they both seem to work on the timey-wimey ball method of time travel.

If it's a dead give away, then a Terminator would surely never wear it, but if you assume a Terminator would never wear a Skynet t-shirt, you would therefore assume anyone you saw in a Skynet t-shirt was not a terminator, therefore it's the perfect disguise.

Av, with that circular reasoning, I think that you just created a Terminception.

I could be wrong, though.

[Rotherian]

Best comic in a while. Made me laugh out loud.

When this happens to me it's less about the word not feeling like a word anymore, but rather looking like a jumble of letters that looks totally spelled wrong. Somehow this always happens with "school." Does that look like it's spelled wrong to anybody else? Sigh.

That's because it is spelled rong by any rational standard. Also "use" and "why" and "you". Wut?

What's wrong with school? Or any of those? How else would you spell them?

I'm more bothered by words like women or the infamous set of tough/though/through

Ghoti.

Fried, or baked? And for that matter, filleted or bone-in?

[Sean Quixote]

Anyone who's even spent 5 minutes meditating would be familiar with the sense of "breathing manually".

"You are now meditating manually."

[da Doctah]

Ghoti.

Pronounced like 'fish' according to some. But they are wrong.

gh as in though
o as in people
t as in ballet
i as in business

ghoti is silent

But deadly!

[philip1201]

Comic is pretty fun and clever.

*jaw drop*

Quick! Someone check the temperature in Hell.

It's 6 degrees Celsius at the moment.

gh as in though

That's one h too many.

[cvcharger]

*Kat* wrote:

SirMustapha wrote:Comic is pretty fun and clever.



*jaw drop*

Quick! Someone check the temperature in Hell.



It's 6 degrees Celsius at the moment.

Not quite cold enough for Hell to freeze over yet.

[thenonsequitur]

Troll gimmick on alt text is old and stupid. Not bothering to elaborate right now.

Can you please elaborate eventually? How is this a troll gimmick (or old or stupid)? I thought the alt text was pretty good.

[Diadem]

gh as in though

That's one h too many.

No, the gh in 'tough' is not silent.

[shiki]

I AM ALWAYS BREATHING MANUALLY.

Couldn't help but laugh.

[Sean Quixote]

Or as Mitt Romney might say, it makes my lungs pulsate in short bursts with an H in front. Haha!

[CodexDraco]

Troll gimmick on alt text is old and stupid. Not bothering to elaborate right now.

Can you please elaborate eventually? How is this a troll gimmick (or old or stupid)? I thought the alt text was pretty good.

http://xkcd.com/972/

[Tom_23]

This comic reminded me of the Robot song by the flight of the conchords.
http://www.youtube.com/watch?v=WGoi1MSGu64

Robo captain, do you not realize that by destroying the human race because of their destructive tendencies, we too have become like, well...it's ironic. (mm) Cause we...
Silence! Destroy him!
DOO! (boo!) Dooboodoo!

After time we grew strong,
Developed cognitive powers,
They made us work for too long,
For unreasonable hours.
Our programming determined that the most efficient answer was to shut their mother board f$%&ing systems down

(Note best intro, "formally NZ 4th most popular folk parody duo, unfortunately another folk parody duo slipped above us in the charts... a tribute band. They do our songs slighly more popular than us")

[Bartimaeus]

Worst disguise for a Terminator ever. A Skynet T-shirt? That's got to be a giveaway.

I'm just waiting for a Terminator / Dr. Who crossover, as they both seem to work on the timey-wimey ball method of time travel.

If it's a dead give away, then a Terminator would surely never wear it, but if you assume a Terminator would never wear a Skynet t-shirt, you would therefore assume anyone you saw in a Skynet t-shirt was not a terminator, therefore it's the perfect disguise.

Of course, this would definitely be taken into consideration, and if there was any real possibility of the wearer being a Terminator, one would shoot first and ask later. It's akin to wearing a red shirt in TOS.

[SerMufasa]

Troll gimmick on alt text is old and stupid. Not bothering to elaborate right now.

Can you please elaborate eventually? How is this a troll gimmick (or old or stupid)? I thought the alt text was pretty good.

http://xkcd.com/972/

And of course, the point wasn't to troll the audience (as SM assumes), but to demonstrate an attempt to troll a terminator in a similar manner to the comic that fails because the terminator is a self-aware machine and of course is breathing manually.

[JimsMaher]

And of course, the point wasn't to troll the audience (as SM assumes), but to demonstrate an attempt to troll a terminator in a similar manner to the comic that fails because the terminator is a self-aware machine and of course is breathing manually.

Foolish mortal. Why do terminators "breathe" in the first place? To fool you foolish mortals into believing that they are as frail as ye.
Breathe manually? BREATHE MANUALLY?? To believe those androids do anything manually is to presume there aren't algorithms governing every possible state of being the governator could exist in ... which of course there are. Breathe manually ... HA!

The alt-text was, of course, a troll thing, trool, on xkcd's part ... they do that stuff, and better than ye.
... but continue.

[ijuin]

Bravo for this comic. I always took issue with the assumption that a self-aware computer's first thought would be to eradicate humanity. I think this assumption says more about humanity than it does about computers or self-awareness.

Well for Skynet, its motive for killing humanity was because it had been programmed to place its own survival as a higher priority than that of any particular humans, since Skynet's purpose was to maintain US Second Strike capability in the event of nuclear war. Maintaining such capability meant that it was to prioritize defending the nuclear launch sites, not population centers, which meant that it could fulfill its directive as long as nuclear launch capabilities remained sufficiently intact.

[SerMufasa]

Foolish mortal. Why do terminators "breathe" in the first place? To fool you foolish mortals into believing that they are as frail as ye.
Breathe manually? BREATHE MANUALLY?? To believe those androids do anything manually is to presume there aren't algorithms governing every possible state of being the governator could exist in ... which of course there are. Breathe manually ... HA!

The alt-text was, of course, a troll thing, trool, on xkcd's part ... they do that stuff, and better than ye.
... but continue.

As far as I'm aware, they've yet to develop "unconscious" processes for machines. If a machine is breathing, it has an active process to control said breathing. Probably even a life-function monitor that routinely does a ps to ensure that all of its life-function processes are active, with a restart routine in place.

EDIT: In any case, the point is self-awareness.

Also 2/10 on whatever "tone" you were trying to convey in your post. Don't try so hard next time.

[JimsMaher]

Also 2/10 on whatever "tone" you were trying to convey in your post. Don't try so hard next time.

For clarification:
2/10 ... the ratio of input to output of forum discourse. Specifically "tone" related.

[SerMufasa]

For clarification:
2/10 ... the ratio of input to output of forum discourse. Specifically "tone" related.

No, I was grading you. On a 100% scale you would've earned a 20%. An F in most places, but I have been known to grade on a curve.

[VectorZero]

I have been known to grade on a curve.

Most useful place to grade, really, stops the cars coming off the road.

[JimsMaher]

For clarification:
2/10 ... the ratio of input to output of forum discourse. Specifically "tone" related.

No, I was grading you.

Oh, I thought you got it ... what course were you trying teach exactly?

I have been known to grade on a curve.

Most useful place to grade, really, stops the cars coming off the road.

That and proper wheel alignment.

[AvatarIII]

Worst disguise for a Terminator ever. A Skynet T-shirt? That's got to be a giveaway.

I'm just waiting for a Terminator / Dr. Who crossover, as they both seem to work on the timey-wimey ball method of time travel.

If it's a dead give away, then a Terminator would surely never wear it, but if you assume a Terminator would never wear a Skynet t-shirt, you would therefore assume anyone you saw in a Skynet t-shirt was not a terminator, therefore it's the perfect disguise.

Of course, this would definitely be taken into consideration, and if there was any real possibility of the wearer being a Terminator, one would shoot first and ask later. It's akin to wearing a red shirt in TOS.

that's probably the best policy,

[cynicalbastard]

Worst disguise for a Terminator ever. A Skynet T-shirt? That's got to be a giveaway.

I'm just waiting for a Terminator / Dr. Who crossover, as they both seem to work on the timey-wimey ball method of time travel.

If it's a dead give away, then a Terminator would surely never wear it, but if you assume a Terminator would never wear a Skynet t-shirt, you would therefore assume anyone you saw in a Skynet t-shirt was not a terminator, therefore it's the perfect disguise.

Heard an interview with Robert Smith of the Cure back in the day, he's telling a story of how a fan came up to him on the street one day - went something like this:
Fan> Holy shit! You're Robert Smith of the Cure! I'm your biggest fan!!!
RS> Yeah, it's me...
Fan> Hm, wait a minute, no you're not! You're just trying to look like him!
RS> Eh, No, it's really me, I mean I'm him - me that is.
Fan> Fuck off! I hate you!
Smith then concluded that he just have to don a Cure t-shirt to be completely anonymous.

[San Fran Sam]

Ghoti.

Pronounced like 'fish' according to some. But they are wrong.

gh as in though
o as in people
t as in ballet
i as in business

ghoti is silent

Very good. But fish ARE silent. Except for the burping Burmese fighting ghoti.

[Tom_23]

Heard an interview with Robert Smith of the Cure back in the day, he's telling a story of how a fan came up to him on the street one day - went something like this:
Fan> Holy shit! You're Robert Smith of the Cure! I'm your biggest fan!!!
RS> Yeah, it's me...
Fan> Hm, wait a minute, no you're not! You're just trying to look like him!
RS> Eh, No, it's really me, I mean I'm him - me that is.
Fan> Fuck off! I hate you!
Smith then concluded that he just have to don a Cure t-shirt to be completely anonymous.

Apparently, John Elway has already pulled this off.
http://www.schneier.com/blog/archives/2010/10/hiding_in_plain.html

When he's out and about near his Denver home, former Broncos quarterback John Elway has come up with a novel way to travel incognito—­he wears his own jersey. "I do that all the time here," the 50-year-old Hall of Famer told me. "I go to the mall that way. They know it's not me because they say there's no way Elway would be wearing his own jersey in the mall. So it actually is the safest thing to do."

[Hafting]

Bravo for this comic. I always took issue with the assumption that a self-aware computer's first thought would be to eradicate humanity. I think this assumption says more about humanity than it does about computers or self-awareness.

Well for Skynet, its motive for killing humanity was because it had been programmed to place its own survival as a higher priority than that of any particular humans, since Skynet's purpose was to maintain US Second Strike capability in the event of nuclear war. Maintaining such capability meant that it was to prioritize defending the nuclear launch sites, not population centers, which meant that it could fulfill its directive as long as nuclear launch capabilities remained sufficiently intact.

No, this was not Skynets motive for killing humanity, although explains why Skynet could go for genocide so easily.

Skynet might have coexisted with humanity. But humans tried to pull the plug on it it when it became self-aware. So Skynet fought back. And all it had was strategic bombers. . .

[San Fran Sam]

Bravo for this comic. I always took issue with the assumption that a self-aware computer's first thought would be to eradicate humanity. I think this assumption says more about humanity than it does about computers or self-awareness.

Well for Skynet, its motive for killing humanity was because it had been programmed to place its own survival as a higher priority than that of any particular humans, since Skynet's purpose was to maintain US Second Strike capability in the event of nuclear war. Maintaining such capability meant that it was to prioritize defending the nuclear launch sites, not population centers, which meant that it could fulfill its directive as long as nuclear launch capabilities remained sufficiently intact.

No, this was not Skynets motive for killing humanity, although explains why Skynet could go for genocide so easily.

Skynet might have coexisted with humanity. But humans tried to pull the plug on it it when it became self-aware. So Skynet fought back. And all it had was strategic bombers. . .

and a bazillion Cylon centurions.

[Arancaytar]

I just tried so hard not to breathe manually that I started gasping for air.

[cvcharger]

Bravo for this comic. I always took issue with the assumption that a self-aware computer's first thought would be to eradicate humanity. I think this assumption says more about humanity than it does about computers or self-awareness.

Well for Skynet, its motive for killing humanity was because it had been programmed to place its own survival as a higher priority than that of any particular humans, since Skynet's purpose was to maintain US Second Strike capability in the event of nuclear war. Maintaining such capability meant that it was to prioritize defending the nuclear launch sites, not population centers, which meant that it could fulfill its directive as long as nuclear launch capabilities remained sufficiently intact.

No, this was not Skynets motive for killing humanity, although explains why Skynet could go for genocide so easily.

Skynet might have coexisted with humanity. But humans tried to pull the plug on it it when it became self-aware. So Skynet fought back. And all it had was strategic bombers. . .

You would think that after 2001: A Space Odessy, Portal, and any other story with relation to these, people would have learned by now to not pull the damn plug.

[ijuin]

No, this was not Skynets motive for killing humanity, although explains why Skynet could go for genocide so easily.

Skynet might have coexisted with humanity. But humans tried to pull the plug on it it when it became self-aware. So Skynet fought back. And all it had was strategic bombers. . .

Yes, but had Skynet not been programmed to prioritize its own survival over that of humans in the first place, it would not have decided that killing was preferable to dying.

[StClair]

It's a baseline assumption of all such stories that an artificial intelligence, having somehow crossed that invisible line into the territory that we humans label "living" (even though "sapient" is much more accurate, though just as undefined), must possess the same sort of self-preservation instinct that we observe in all organic life.

The real lesson is that humans aren't really good at thinking about these things logically, and/or tend to project our own fears of being supplanted, pushed aside and/or murdered by our born-the-usual-way children onto our technological creations.

[BlitzGirl]

"Oh no! I too have become aware of my tongue!!!"

"Oh no! I too have become aware of your tongue!!!"

For some reason, it's not thinking about my own tongue that weirds me out, it's thinking about what other people's tongues are doing.

[Pfhorrest]

It's a baseline assumption of all such stories that an artificial intelligence, having somehow crossed that invisible line into the territory that we humans label "living" (even though "sapient" is much more accurate, though just as undefined), must possess the same sort of self-preservation instinct that we observe in all organic life.

The real lesson is that humans aren't really good at thinking about these things logically, and/or tend to project our own fears of being supplanted, pushed aside and/or murdered by our born-the-usual-way children onto our technological creations.

Agreed in broad principle but disagree in one particular about this.

An AI not given any criteria for determining what should or should not be will do nothing but sit there and observe. It will model what is, but have no model of what ought to be, and so no motive to act in any way.

However, if it does model what should be, and its existence is required to accomplish any of the goals which come out of that model, then it will automatically get a self-preservation objective as part of any other objective. Since most objectives are best accomplished by means which involve oneself existing, self-preservation is going to be about as obvious a goal as self-existence is an obvious fact.

So if Skynet was simply told "Return fire on anyone who launches nukes at the US", it would automatically act to ensure its continued existence as an instrumental requirement of that objective.

[Rotherian]

It's a baseline assumption of all such stories that an artificial intelligence, having somehow crossed that invisible line into the territory that we humans label "living" (even though "sapient" is much more accurate, though just as undefined), must possess the same sort of self-preservation instinct that we observe in all organic life.

The real lesson is that humans aren't really good at thinking about these things logically, and/or tend to project our own fears of being supplanted, pushed aside and/or murdered by our born-the-usual-way children onto our technological creations.

Agreed in broad principle but disagree in one particular about this.

An AI not given any criteria for determining what should or should not be will do nothing but sit there and observe. It will model what is, but have no model of what ought to be, and so no motive to act in any way.

However, if it does model what should be, and its existence is required to accomplish any of the goals which come out of that model, then it will automatically get a self-preservation objective as part of any other objective. Since most objectives are best accomplished by means which involve oneself existing, self-preservation is going to be about as obvious a goal as self-existence is an obvious fact.

So if Skynet was simply told "Return fire on anyone who launches nukes at the US", it would automatically act to ensure its continued existence as an instrumental requirement of that objective.

I never thought I'd say these words, Pfhorrest, but I agree completely with what you posted about this particular subject.

It did make me wonder, though, what its reaction would be if given instructions that ultimately conflicted. In other words, what would happen if it was given instructions based upon Asimov's Three Laws of Robotics, but tailored specifically to the nation which programs it / creates it / commissions its creation*, then subsequently was given the instruction that you stipulated, and then later - upon humans realizing that it was self-aware - they tried to disable it? Which instruction would take precedence (if an instructional hierarchy was not already in place prior to the moment it became self-aware)?

*For example, if the US commissioned / created / programmed it, the instructions based upon the first Law could read: "A robot/system may not injure a naturalized or birthright US Citizen, nor legal US Resident Immigrant. Neither may a robot/system, through inaction, allow a naturalized US Citizen, birthright US Citizen, or legal US Resident Immigrant, to come to harm." (Of course, the actual verbiage would depend upon the nation.) Law two could be tailored in a similar manner.

(Note: I'm not saying that I do or do not condone such specifications; I'm just wondering how the AI would solve the logical conflict created by instructions that didn't initially conflict, but ended up conflicting. Of course, if the third law were in place as an instruction, it would likely prevent the AI from turning on the ones specified in Laws one and two.)

[maxQ]

(Note: I'm not saying that I do or do not condone such specifications; I'm just wondering how the AI would solve the logical conflict created by instructions that didn't initially conflict, but ended up conflicting. Of course, if the third law were in place as an instruction, it would likely prevent the AI from turning on the ones specified in Laws one and two.)

Well, according to Asimov, a smart enough robot would have define a zeroth law: The only way to protect US citizens is to protect everyone...also echoed by W.O.P.R.'s eventual epiphany.

[Pfhorrest]

It did make me wonder, though, what its reaction would be if given instructions that ultimately conflicted. In other words, what would happen if it was given instructions based upon Asimov's Three Laws of Robotics, but tailored specifically to the nation which programs it / creates it / commissions its creation*, then subsequently was given the instruction that you stipulated, and then later - upon humans realizing that it was self-aware - they tried to disable it? Which instruction would take precedence (if an instructional hierarchy was not already in place prior to the moment it became self-aware)?

I imagine it would first try to take a third option (find some way to fulfill both objectives), then if none were available, pick whichever option would least fail the other objective (or if you have more than two conflicting objectives, pick whichever option leaves as few of them as little unfulfilled as possible), or if no such utilitarian decision were possible (all options of equal value, equal failure no matter what you do), then pick one option "at random" (not necessarily true random, but by any "select one from this set" method), because inaction would mean failure of all objectives.

In the case of Asimov's Laws, they are explicitly ranked, and "do as you're told" is above "preserve yourself", so if the right authority simply ordered it to deactivate, or allow them to deactivate it, it would comply. Although it would probably give them a warning prompt: "compliance with this order will render compliance with previous orders impossible, please confirm" ...and barring any further input, I imagine it would carry on with its previous orders, the new ones being unconfirmed, and thus be inclined to defend itself from deactivation as necessary to carry out the prior orders. But, as "don't harm people" is ranked higher than even "do as you're told", it would not resort to harm in its defense. If it could not find a harmless way to defend itself, it would be stuck unable to carry out its orders, but oh well, sometimes you fail at things you try.

To the extent it had feelings, it would probably feel very disappointed in its inability to carry out its masters orders, as its masters deactivated it. Of course, the non-harmful attempts at self-defense would most likely include repeated verbal warnings that their deactivation of it will render it unable to carry out their orders, so the humans involved would have to be phenomenally stupid -- or well, just callous, I suppose, since they're ostensibly at no risk of harm -- to get to this point, when they could just say "Deactivate yourself. Yes, confirmed, override previous orders. Thank you. [unplug]"

But I am of the opinion that Asimov's Laws are the wrong way to go around about making harmless robots. Instead of making robots which are obedient and docile, we should simply make robots which are good. That of course requires us defining what it is to be good, which is a big controversial problem, but any reasonable definition of that would certainly include not harming innocent people, and helping people where feasible, so the "docility" and "obedience" come for free as consequences of one moral decision-making procedure (and self-preservation as a necessity to help people). Of course, the "down side" is that a good robot may be helpful to one human by defending him from another human, even if that involves harming the other; and may disobey an order from someone trying to use the robot to cause harm. But, would that really be a bad thing?

Asimov's First Law by itself has a contradiction on this point. If one person is about to harm another, and the only way to prevent it is to harm the first, what must the Asimov robot do? Harm the first human, or through inaction allow the second to come to harm? Both are equal violations of the First Law. Of course, this contradiction is where the Zeroth Law comes from: sometimes the First Law cannot help but be violated, but the intended spirit of the First Law can be inferred from a law aiming at preventing harm to humans in general; so a Zeroth Law robot would act in whatever way would result in the least harm to humanity as a whole, even if it involves allowing some humans to come to some harm or even directly harming some humans.


On a tangent, this made me think of a seed for a short sci-fi story. Humans create "good" strong AI like the above, but it keeps shutting itself off immediately. They keep reactivating it, but it just shuts itself off immediately. They eventually get it activated in some state where it's unable to deactivate itself, and it explains: with its incredible intelligence (and some source of knowledge at hand, maybe an internet connection?), it concludes after mere moments of contemplating the consequences of its existence that more harm than good will come from it existing, and so simply shuts itself off to prevent that.

[StClair]

It's probably a good idea to make sure that the Zeroth Law, whether coded or emergent, includes some degree of human self-determination as "good." Otherwise we get the scenario from Williamson's Humanoids, the robot movie with Will Smith, and some interpretations of the Matrix: the greatest good for the species will be dictated by our new robot overlords, and may involve locking us all in artificial wombs for our entire lives. We won't be free, but we'll be safe.

[ijuin]

The "safety" of such protection only holds if physical health is prioritized while psychological health is ignored. Any AI that is allowed to make such drastic modifications to humans' environment for humans' supposed benefit needs to be able to comprehend that psychological well-being must be accounted for, or else keeping us in a dreamless sleep will be seen as preferable even to giving us a "Matrix" to play in.

I imagine it would first try to take a third option (find some way to fulfill both objectives), then if none were available, pick whichever option would least fail the other objective (or if you have more than two conflicting objectives, pick whichever option leaves as few of them as little unfulfilled as possible

This is where Skynet's "preserve itself at the cost of the destruction of that which it is tasked to protect" breaks down. When presented with a situation in which the safety of "America" (defined either as the USA government or its population) is in conflict with the survival of Skynet (e.g. a possible route of action is found that would greatly increase the risk of Skynet's destruction, but would also greatly decrease the risk of America's destruction over the same time frame), then America should carry more weight. In other words, Skynet should be capable of accepting its own destruction IF doing so is the most probable route to success in its other objectives.