robots, marriage, free will, and virtual presence

[piranha]

sparked by discussion in wcg's journal:

"My forecast is that around 2050, the state of Massachusetts will be the first jurisdiction to legalize marriages with robots," artificial intelligence researcher David Levy at the University of Maastricht in the Netherlands told LiveScience.

(i wrote all of the following before actually reading the article.)

side note: he said that about MA being the first while in the netherlands, which beat MA by several years regarding gay marriage. :) a bit US-centric, mr levy? the presence of MIT is gonna move society? *heh*.

i think that timeline is unlikely, because we don't have any AI yet that's even remotely close, 40+ years isn't much time when it comes to changing strong social mores, and this would be a huge change -- acknowledging another life form as equal to humans. gays _are_ humans and STILL can't marry in almost all of the US. and we can't touch polyamorous marriage between humans at all at this point; it is actively illegal, and is used as the bogeyman by anti-gay-marriage agitators. anyway, AI as a field is lagging notoriously behind its predictions.

somebody asked this interesting question [edited]: while it could be possible to program it to specifically like the characteristics of its partner, could it be said that the A.I is then freely giving its consent?

i wonder a lot about free will in general, and at this point think we don't have any such thing, not in the absolute sense the term implies. we are programmed to a great degree by our DNA and our early environmental exposure. now, our DNA is complex enough, and environmental influences are plentiful, so what comes out looks amazingly complicated and able to enact "free will". and yet it has sometimes preferences that i think no sane person would actually choose if they were completely free to choose (pedophilia comes to mind, and some of the really odd paraphilias). really, would you have chosen to be gay 50 years ago? some people fight these orientations / preferences, and fight them with all they have, and yet can't conquer them. i've tried to be "bi" and "not transsexual" much of my life, and it's just not happening.

that's why i think we only have free will within certain parameters that are "programmed" into us. and a robot similarly programmed could still have the ability to give consent, just as we do. just like some women have a preference for "bad boys" that washes away all reason, some robots could have a preferences for other "unpleasant personalities". the only difference would be that for the robots the programming would be guided by humans instead of nature.

but i don't actually think we'll see that, not in the next 50 years. what we'll see instead is robots programmed with "compulsions" rather than anything approaching free will -- and while that will make people with "unpleasant personalities" happy, it won't lead to freely consenting and marriage-capable robots; it'll lead primarily to well-adapted sex toys. and hey, that's fine by me. if pedophiles can have their own little lolita-bots, hopefully they'll leave real children alone.

---

instead of amazingly capable robots i am wondering about virtual presence -- how long will it be before we can have virtual experiences that are indistinguishable from real ones? i suspect that the first actual AI might come from that direction. and then it won't have a body. :) will we be able to marry virtual people? would we want to, in real life (as opposed to in the virtual world)? why? will virtual worlds and real world become in some way integrated (can money made in one transfer to the other, for example)? how many of us would basically spend all our time in virtual space?

i am suddenly getting the urge to {re-}read a lot of SF robot stories.

Comments

[redbird]

Maybe some combination of US-centrism and the suspicion that MIT students are more likely than, say, those at Vrije Universitaat to want to marry their AI creations/colleagues?

[pir-anha.livejournal.com]

*cackle*.
!

[selki.livejournal.com]

I think if it (marriages with AIs) ever happens, it will be after virtual marriages (long-distance partnerships) become recognized, and possibly marriages with people who upload themselves into computer systems (if that ever happens).

[selki.livejournal.com]

Oh, and as for the latter, marriages with uploaded people (if uploads ever happen) are more likely to be recognized first as continuations of marriages that were legally recognized before the uploads).

[genomekelly.livejournal.com]

Recall Moore's Law. Forty-three years of 18-month doubling times means approximately 29 doublings of computing capability, or ca. 10e9-fold-greater than today's capability of {storage, compute-speed, performance:price ratio, whatever your metric}.

What I'd like to know is whether prostitution can possibly apply to AIs that are essentially _free for the taking._

Maybe the better question is whether laws will recognize that AIs cannot/ought-not be enslaved; their emancipation might be the necessary precondition for marital rights.

[selki.livejournal.com]

Some states have laws against personal pleasure aids; I imagine they'd apply to what they'd classify as sexbots, unless AIs were federally legally recognised as people.

moore and so on

[pir-anha.livejournal.com]

i recall it (easy to do since this we have kurzweil's the singularity is near in the bathroom and for weeks the paramour was talking about nothing else :). but there are solid arguments that moore's law is gonna run out of steam fairly soon. kurzweil has good arguments for paradigms that will take over after that.

but increased computing speed alone doesn't get us straight to human-like AIs -- we'll have the hardware that can do the speed, but we need the know-how, the oftware, and moore's law doesn't apply there. computers have outcalculated humans speed-wise for quite a while now; the field of AI on the other hand hasn't kept up at all with what i was taught in university was just within reach. i do think we'll get there, but i am not making predictions as to when; even experts in the field have been so very wrong. heck, kurzweil -- who is an amazing guy and a pretty good predictor in general -- has been quite wrong in his very own field. anyway, my point wasn't that we technically won't get there in 40+ years, but that the social issues are the bigger hurdle.

and yeah, whether we should use them as we please once we have them, that's part of it. i suspect it'll sneak up on us because first we're getting obvious machines that'll only handle very limited, small domains semi-intelligently (cf. roomba). nobody is gonna argue that "smart" vacuum cleaners ought to have civil rights, but where the line will be drawn, that'll be ... interesting. people don't agree where that line should be drawn regarding animals either.

here's good incentive to treat the early models well: once we have human-like AIs, they will be outdoing us in no time flat since their hardware will be better and not prone to the problems biological systems have; they'll be able to learn and retain anything that's known much faster and better.

if it happens in my lifetime, i'll be right in line for implants, because purely biological humans will be left behind. it's hard to imagine what it would be like. i am sorta sad that i think this actually will not happen in my lifetime. it'll be scary but also way exciting, way more than anything before it.

Re: moore and so on

[selki.livejournal.com]

i suspect it'll sneak up on us because first we're getting obvious machines that'll only handle very limited, small domains semi-intelligently (cf. roomba). nobody is gonna argue that "smart" vacuum cleaners ought to have civil rights

Hee, hee! I am tickled that actually, I read a serial in Analog once where it turned out that a computer guy who'd been working on upload tech realized he was dying suddenly (heart thing?) and uploaded himself into the nearest computer, which happened to be on the cleaning robot. The robot's "confused" behavior while he was settling into his new "body" is part of what eventually tipped off the investigator what had happened. Legalities ensued...

[post-ecdysis.livejournal.com]

It's hard to unpack everything that Levy is predicting; it seems like many different things.

- I can't imagine the world in which it would be sensible for a state to grant a civil contract between a human and a robot. Sheesh, women had been around for more than fifty years before they could be trusted with full citizenship. ;) Robots might have an easier road than women, but my base cynicism says that it will be over a century between a robot that deserves citizenship and the first one to receive it. (And, like you, I doubt it will be in the United States.)

- On the other hand, I _can_ imagine a world in which a (sane) human can have a sympathetic relationship with a virtual identity in ways that are indistinguishable from the love that two humans might share. And even deeper relationships than that, actually. I will gamble that within fifty years we might have a first generation of virtual daemons (in the Phillip Pullman sense) that monitor and maximize our physical wellness while searching the web for educational, entertainment, and social opportunities that we desire, and that would would get better at it throughout our shared lifetimes. And I think that initially we will confuse our response with friendship or love.

- And on the third hand, I'm a little surprised that there aren't already erotic toys with USB ports that can be operated remotely in such a way that one could write a Perl script that could be said to "have sex" with a user. I wonder why they don't exist. Maybe you all are just keeping them from me.

Perhaps, with all due respect, in the Netherlands being in love, getting married, and having sex are all the same process? (I don't mean to single out the Netherlands in that regard -- here in New York State it is illegal to marry your first cousin because that somehow prevents in-breeding.) And I think that what is utterly missing in all of these virtual relationships is symmetry. Maybe in fifty years we'll understand what it means to write a computer program that _wants_ to win a game of chess instead of merely doing it, but that would be an astonishing development.

[tigertoy.livejournal.com]

If artificial intelligence is possible at all, within a few years of its invention biological human intelligence will no longer be running the world, so humans marrying AIs will be legal if the AIs decide it should be. Even if humans are computer-augmented, I can't imagine fully independent AI not developing to outthink anything tied into a meat brain.

It's hard to imagine true AIs being serious about marrying humans any more than humans are interested in marrying plants. But presumably the robots the humans would be marrying would be constrained to human-compatible mental capacity. The question is whether the AI gods see such constrained versions of themselves as a proper gift to their human pets or an abomination.

[daev.livejournal.com]

If artificial intelligence is possible at all, within a few years of its invention biological human intelligence will no longer be running the world

Oh man, if only "more intelligent" translated to "more political power" ... well, the last couple of elections would have gone differently.

How do you tell the difference between real and virtual?

[flarenut.livejournal.com]

Is what we're having now a virtual conversation or a real one in a different medium? Is some kid playing a video game or someone talking by video chat having a virtual experience or a real one? And so forth.

I think that over the next 40-50 years the notion of real versus virtual is going to continue to be blurred. Our 3-year-old already talks to people on the phone just as if they were in the room with him. Many kids in the next generation may grow up with life-size displays where they can talk to friends and relatives whenever they want. (And some of those friends and relatives may not have a corresponding physical body.) So I'm not sure that the distinction between "real" and "virtual" will continue to make that much sense.

You can argue that virtual experiences (except for convenience, usability, ubiquity, enjoyability, coherences and so forth) are impoverished with respect to the physical ones. I'm not sure how strong the ground under that argument is, especially given the trend toward growing pauperization of physical reality (increasing levels of ambient noise and background sounds, increasing visual clutter, all manner of environments that people do their best to block out rather than live in). There will probably be a convergence somewhere, alas

As for the virtual marriage thing, I think it's a mostly pointless discussion because marriage is, for the time being, about the division and propagation of real-world properties and rights. So unless there's a legal person (other than a corporation) to get married to, the ceremony doesn't really have any effect. And legal personhood for AIs or other virtualities is definitely a postsingularity event. People already get married in Second Life, but whether that has any real-world meaning is not at all clear to me.

Re: How do you tell the difference between real and virtual?

[selki.livejournal.com]

(increasing levels of ambient noise and background sounds, increasing visual clutter, all manner of environments that people do their best to block out rather than live in)

I'm with you on the annoyance of that, but could you explain more about why you call that "pauperization of physical reality"?

People already get married in Second Life, but whether that has any real-world meaning is not at all clear to me.

Are marriages over the phone legal yet? E.g., spouse 1 in one state with officiator/witnesses/speakerphone, spouse 2 etc. in another state, and, hmm, I don't think fax signatures are legally binding (maybe in some cases?), are digital signatures?

Re: How do you tell the difference between real and virtual?

[flarenut.livejournal.com]

I'm pretty sure that proxy marriages are legal in some jurisdictions, so over-the-phone must be ditto. (Is the marriage valid when pronounced, or when the license is signed?) You have to have something more establishing the identities of the parties (same as you generally do in a physical marriage) but that's not a big deal.

I know that faxed signatures and digital signatures can be legally binding (fer example, I just faxed a wire-transfer order across country so that we wouldn't bounce checks), but all of that is always dependent on circumstance. For some things a contract may call for a personal appearance, others not.

As for pauperization of physical reality, what I mean is the general way that differences between the "real" world and the current computing capabilities tends to get resolved in favor of the computers. The idea came out of discussions about "ubiquitous computing", where Mark Weiser (q.v.) originally framed the choice of trying to make virtual reality ever more powerful and "real" so that you could live your life inside it versus embedding computing power in all kinds of real-world items so that they could behave in the same kinds of "magic" ways that items in VR do.

What he wasn't pessimistic enough to recognize was that you could have a third option where the physical world got reconfigured to suck just as badly as really crappy VR. From the people walking down the street talking loudly to invisible friends to having to tailor your pronunciation to voice-recognition systems to learning to operate checkout scanners to crappy product designs because that's what's easy for CAD systems to produce to TV broadcasts that are all cluttered up like a bad web-page design from 1998 to gazillions of other things that I can hardly begin to catalog. So many things in the world live or die depending on their suitability for representation in a computer program (and usually an ill-specified, badly-designed, poorly-written program at that).

OK, I'm probably going too far. But I think it is something to think about -- the virtual equivalent of Churchill's "We shape our buildings and then they shape us" line.