THESE FORUMS NOW CLOSED (read only)
Comic Discussion => QUESTIONABLE CONTENT => Topic started by: drosera88 on 21 Oct 2015, 14:58
-
I don't post much. It's been years since I've posted. I'm curious though...
Are the robots in QC basically immortal as long as they are maintained? Or do they have some artificially imposed lifespan? Do the one's who have owners willing submit themselves to their owners or are they programmed to be incapable of becoming individuals like some of the other robots are? Basically robot philosophy withing the QC-verse in general.
-
Lifespan is unknown. AIs are a very recent thing, and their civil rights even more so. Though practically at rocket speed compared to other group's being granted civil rights... Nothing has even been mentioned about any kind of natural neural degregation over time, but it's really probably to soon to guess at that sort of thing. An AI's body is totally ancillary to their existence, and ultimately replaceable or unneeded. Any kind of artificially imposed limited lifespan ala replicants would be a violation of their civil rights of the highest order.
Since civil rights came about, AIs do not have owners. Many may have not had them before. Momo has said that the Companion program is totally voluntary on both sides. In most cases an AI is considered to be the owner of it's chassis... Though that has changed over the course of the comic. There are times when Momo and Pintsize needed permission from their human companions to install driver updates. Also in some cases an AI may inhabit a chassis it does not own, like a submarine, space station or an illegally acquired combat chassis.
-
Robot Immortality (http://www.questionablecontent.net/view.php?comic=70)
-
A bit more information with links to relevant comics (http://questionablecontent.wikia.com/wiki/AnthroPC).
-
AI are probably not immortal if you take immortality in the same sense as omnipotence. Because that kind of immortality would violate entropy laws. Over a long enough timeline, an AI will find a terminal end.
But in the sense of being not mortal as humans are, they basically are. Pintsize suggested as much. As long as they are serviced (and they can service themselves and others of their kinds) they won't "die." That's really all we know.
There are issue that spring to mind, such as limits on accessible memory, and thus a need to prune memory of time and thus the reaching a point where enough memory is off line that resulting personality is no longer the same person who started. Is that death? QC hasn't addressed that and doesn't seem likely to. But that's going pretty deep. As far as the basic question goes, there's nothing yet known, other than doing damage that will kill an AI. Even starving one, since we've seen several shutdowns.
I don't know if we've seen any weak AI (outside of google). But what we have seen of QC AI suggests that weak AI (or narrow AI) can't become strong AI. General (strong) AI was a deliberate creation. It took human intervention to make. So a chat bot or search engine isn't going to spontaneously develop into a thinking being with autonomy. It either was one to begin, or it wasn't. I think it would be interesting if there were AI in the QC verse agitating for the uplift of chatbots and IVRs of the world. A lot of google's infrastructure is so complex IRL that it probably wouldn't take much intervention to uplift it, in the QC verse, and it would have a different way of thinking.
(Notably, very large AI exist in the QC verse, and they are suggested to have somewhat alien thought modes, themselves). All of this suggests that if AI exist in the QC verse who aren't individuals, they aren't capable of becoming individuals without some outside intervention, so it's not necessary to program them not to be individuals.
-
>There are issue that spring to mind, such as limits on accessible memory, and thus a need to prune memory of time and thus the reaching a point where enough memory is off line that resulting personality is no longer the same person who started.
Moore's Law would cover that. By the time a given A.I.'s memory was overflowing, cheaper and larger memory drives would be available. That assumes that they or their owner are gainfully employed, of course.
-
We're actually getting pretty close now. I do mad-science in AI at home on my own time, and I think I finally cracked neural plasticity, localized pain, and strategic choice among temporally exclusive courses of action to satisfy base needs one at a time. Of course, this mostly raises the level of "consciousness" in my AI from about clam-level to about goldfish level so we're not talking about "people" yet, but I there is real progress being made, by me and others.
By my calculations, which are admittedly based entirely on mad science, Human-level consciousness will require between eighty and eight hundred Terabytes of live data. That's -- surprisingly close. 20 to 200 modern hard drives. There's more than that in one of the rack servers at work where I do my "sane science."
But there's also a big difference between storage and live data. The hard drives are not enough; you've got to have it in RAM with a whole lot of CPU power keeping it working. So we're talking about several dozen server racks.
And then it gets worse unless I develop something better than what I've developed now. I got a neural-plasticity system that does self-organizing neural network wiring, but it isn't fast. It needs to process LOTS of input over LOTS of time (Eighty terabytes would be YEARS, Eight hundred would be CENTURIES) of training on that much RAM and that many processors, along with massive amounts of raw data which I just don't have, to start to organize it into the kind of higher structures a mammal would use.
Of course by the time that much RAM and that much CPU comes inside my mad-science budget, Moore's law will have cut those years down to a couple weeks each? Maybe? Until then I'm working on the software....
-
Incidentally, I really hate the fact that when I'm doing what's essentially epistemological philosophy - literally trying to create intelligent self-willed beings, that's "mad science" and gets no support from anywhere - in fact mostly it gets people scared or freaked out.
But "sane" systems? Systems that many thousands of people are working on, sweating over, and spending their whole lives trying to tune one percent better -- are NOTHING to do with consciousness. They're just function optimizers that happen to make money.
Big. Fucking. Deal. Those are just ways to do the same business faster and more efficiently; they don't fundamentally change anything. If you want to fundamentally change anything, they call you crazy, and think you're "wasting" effort that could have been spent on "MORE OF THE SAME ONLY FASTER, BIGGER, BETTER, FASTER, CHEAPER, WHATEVER JUST DO IT BEFORE THAT OTHER GUY!!!"
Which, you know, I kind of like the competition and sometimes the ideas from that plebian kind of thing are applicable to things that will actually change the world fundamentally. But their vision is so small....
-
AI are probably not immortal if you take immortality in the same sense as omnipotence. Because that kind of immortality would violate entropy laws. Over a long enough timeline, an AI will find a terminal end.
Somebody said that immortality withoutomnipotence would suck, as it's inevitable that sooner or later you would find yourself burred under a few million tons of rock, and would have no option other than wait for it to erode.
-
Somebody said that immortality withoutomnipotence would suck, as it's inevitable that sooner or later you would find yourself burred under a few million tons of rock, and would have no option other than wait for it to erode.
Actually without omnipotence these million tons of rock would smash you to pieces.
-
You don't need one to have the other.
Immortality means you don't die no matter what happens to you, so no being crushed to death by rocks since being crushed by a few rocks means you were not immortal.
The millions of tons of rocks burying thing is just a blip on the timeline for an immortal entity since you will surface to interesting things happening eventually. So always something to look forward too while whiling away the centuries if stuck in rock strata or much longer if you got stuck in a subduction zone.
It is being stuck in interstellar space where things happen so slowly and on such a vast scale. Talk about boring..... ugh :facepalm:
-
What kind of life do you live that you feel getting crushed by millions of tons of rock is inevitable?
Keep in mind that AIs are not immortal when it comes to sustaining damage. Their bodies can be destroyed. Even though they tend to use armored shells for their cores, they can still be destroyed, which would kill them. The only question is whether or not they would die of 'old age'. Normally us flesh and blood types are stuck with one body. Once that body deteriorates enough, it will cease to function. Though it's more likely that people will die of outside forces before that point, be it disease or injury.
AIs on the other hand, so long as their personality core survives can just switch to a new body when the time comes and their old one is no longer suitable. The bigger question is: can their personality and memory be transferred between cores? Eventually whatever storage they are built on will start to fail, likely leading to the robot version of dementia. Even if they were to switch cores, eventually they would have to undergo neural pruning. Otherwise the amount of data stored would become astronomical. Even if data storage maintained pace with how quickly an AI learns they will probably become slower and slower at remembering things as they have to search through hundreds, possibly thousands of years of data. Possibly by that time they will have some sort of 'backup storage module' where they can archive data that isn't relevant to their current life. Remembering that their human companion from 300 years ago liked blueberry waffles with honey doesn't exactly seem important enough to keep stored in the AI equivalent of RAM.
-
I'm not sure how much thought Jeph has given the AI-lifespan question outside of what he has posted - but I guess it's safe to say that he's a fan of SF authors who have: e.g. Banks or Stross have written on 'very, very old' AIs becoming 'eccentric' (IIRC, there was a Ship-Mind called Mistake Not... in 'Hydrogen Sonate' that was so prone to seeking out remote stars and 'firegazing' at the plasmastreams that other Minds became concerned. Mistake Not... was only its last call-sign - it was really a very old Mind transferred to a brand-new warship).
If we assume that maintenance of the physical substrate is not an issue, any AI that is comparable to us will also have to undergo one or the other form of memory consolidation & pruning, as RF laid out above. It will be driven to learning and transcending its current self - I'm thinking of the poem (https://mindmastery.wordpress.com/2007/02/17/hermann-hesse-steps-stufen/) 'Stufen' (Steps) by Hermann Hesse - some impulse comparable to what we call curiosity.
If that mind reaches a state where it can no longer transcend itself, it may well be subject to something akin to what we call ennui - and if such a state lasts long enough, it might well decide that enough is enough.
Who wants to live forever?
-
Curious entities will want to live as long as there are new things to find out.
We know from Pintsize and Momo's history that QC robots can be backed up and restored onto new hardware of a different model.
Human-AI relations might suffer from human envy at beings exempt from the horrors of aging.
-
There's two things here, and it probably helps to unbundle:
1. Invulnerability -- Inability to be harmed.
2. Immortality -- Freedom from death by natural causes.
Most god-like creatures are both, so we tend to conflate the two.
However it'd be possible to be invulnerable, but have a finite lifespan -- unkillable but eventually dying of old age.
It'd also be possible to be immortal but vulnerable. Like a person benefitting from anti-aging biotechnology. You wouldn't die of natural causes, but you would still be as vulnerable as any other human to poison, fire, gunshots, etc.
-
Let's not forget "eternal youth" in the mix. Immortality doesn't necessarily confer eternal youth, as the Cumaean Sybil discovered to her sorrow.
-
Human-AI relations might suffer from human envy at beings exempt from the horrors of aging.
Not only relations might suffer, as this piece of QC fan-art (https://forums.questionablecontent.net/index.php/topic,4954.msg1055465.html#msg1055465) demonstrates. Consider how we feel when our own shorter-lived companions age and die.
When the body that lived at your single will,
With its whimper of welcome, is stilled (how still!),
When the spirit that answered your every mood
Is gone - wherever it goes - for good,
You will discover how much you care,
And will give your heart to a dog to tear!
-
Let's not forget "eternal youth" in the mix. Immortality doesn't necessarily confer eternal youth, as the Cumaean Sybil discovered to her sorrow.
Good point.
-
We're actually getting pretty close now. I do mad-science in AI at home on my own time, and I think I finally cracked neural plasticity, localized pain, and strategic choice among temporally exclusive courses of action to satisfy base needs one at a time. Of course, this mostly raises the level of "consciousness" in my AI from about clam-level to about goldfish level so we're not talking about "people" yet, but I there is real progress being made, by me and others.
By my calculations, which are admittedly based entirely on mad science, Human-level consciousness will require between eighty and eight hundred Terabytes of live data. That's -- surprisingly close. 20 to 200 modern hard drives. There's more than that in one of the rack servers at work where I do my "sane science."
Yes, they are getting pretty close now. See this article from the University of Washington:
http://www.washington.edu/news/2015/09/21/ai-system-solves-sat-geometry-questions-as-well-as-average-human-test-taker/
Granted, there is still a long way to go, but things are looking better in the AI front.
-
There's two things here, and it probably helps to unbundle:
1. Invulnerability -- Inability to be harmed.
2. Immortality -- Freedom from death by natural causes.
Most god-like creatures are both, so we tend to conflate the two.
However it'd be possible to be invulnerable, but have a finite lifespan -- unkillable but eventually dying of old age.
It'd also be possible to be immortal but vulnerable. Like a person benefitting from anti-aging biotechnology. You wouldn't die of natural causes, but you would still be as vulnerable as any other human to poison, fire, gunshots, etc.
I argued for this before: 2 is better described as amortality. Immortality means you literally cannot die, amortality means that your death is not inevitable, barring the heat death of the universe.
-
It occurs to me there's also:
3. Immunity from diseases.
Without this, you could be amortal, and still die from the flu.
-
We've seen the AIs get viruses.
-
We've seen the AIs get viruses.
Yes! If it is properly maintained.
-
We've seen the AIs get viruses.
PT410X has advice on how to avoid that little problem.
-
PT410X might be a little overconfident... (http://www.theregister.co.uk/2015/11/06/omnirat_malware_android_poc_mac_linux/)
:evil:
-
They haven't been around long enough to answer the original question. There could be some hidden design flaw that will only show up after, say, 30 years.
-
Who wants to live forever?
Me.
-
We've seen the AIs get viruses.
PT410X has advice on how to avoid that little problem.
Actually, I read about a cryptolocker for Linux the other day xD
Who wants to live forever?
Me.
I mean, if I was 22 forever and a couple close friends were 22 forever that'd be great.
-
We've seen the AIs get viruses.
PT410X has advice on how to avoid that little problem.
Actually, I read about a cryptolocker for Linux the other day xD
As have I. But a *lot* of the OSS zealots will pretend that it's not an issue. I know that I *do* use a lot of software that isn't in the main repos. Yeah, part of me is a bit more wary of non-official sources like I use for Netflix-desktop, Bloom, and proprietary drivers (obviously). There are also *very* good reasons to make Flash "click-to-play" rather than just running it, and avoiding Adobe reader like one would Typhoid Mary.
-
I've never had any issues with Adobe Reader myself, but I only have it on the Compie for those rare occasions that some Story i'm reading is in that Format.
-
What kind of life do you live that you feel getting crushed by millions of tons of rock is inevitable?
The mostly immortal Captain Jack Harkness on Torchwood, the BBC spin-off from Doctor Who, found himself entombed in a mass of concrete when elements of the British government came to fear his existence. Fortunately, he had allies who eventually came to his rescue.
-
It occurs to me there's also:
3. Immunity from diseases.
Without this, you could be amortal, and still die from the flu.
Most people would consider diseases to be "natural causes", since they are caused by Nature.