Friday, September 18, 2015

Once you start bangin’, there’s no going back (how AI may be "coming", after all)

Putting the finishing touches on my dissertation, so not much time available to muse on the web about the divine and the humane. Not much reading, either, outside of long dead philosophers (which mostly happen to be more lively and interesting than the vast majority of today’s commenters), and the odd Viennese crackpot (I’m seeing more and more in the media news about  Elizabeth’s Roudinesco new biography of said crackpot, just published in Spanish, and it amuses me to no end that, after the perspective we should have gained on his oeuvre,  there is still people willing to make a fool of themselves defending it… but then again Ms. Roudinesco has been reared in the very noxious fumes of French haute culture, where it is apparently still OK to take such balderdash seriously). But in the rare occasion I’ve been peering outside of my private ivory tower this little article caught my attention:  Call to ban development of "intelligent" sex dolls. But of course! the whole web was, to a considerable extent, launched by pornography (you still can read here and there that up to 40% of the information travelling through the wires –and increasingly airwaves- is pornography that people discreetly download), so why wouldn’t the much heralded Artificial Intelligence be much speeded by it?

As long as it had to be crafted by cranky old doctors in dusty lab coats, struggling in their offices with deep learning, neural networks, semantic loading and whatnot, I stick to my original prediction that one hundred years from now we still will not have anything closely resembling artificial intelligence, neither in a strong nor in a weak sense of the word (that is, anything that would fool anybody above three years old after barely two minutes of conversation). But the moment sex enters in the equation, man that changes everything! I bet any clever Joe is willing to ascribe not just somewhat above average intelligence to his natural-sized doll, but considerable foresight, outstanding sensibility and sensitivity, amazing grace and moral agency in spades as soon as he has banged her to his full satisfaction. And for that she doesn’t even have to open her mouth (to speak, that is). Just givin’ him a lovin’ look may be more than enough. If the thing as much as speaks (even if it is a limited set of predefined phrases)… Jeez, I just hope the Nobel committee members are not between the clients of Abyss Creations or True Companion, as we may be shocked by some of their future choices for the award.

Now let’s get (just a bit) serious. There is a solid reason why I have argued that AI is nowhere near: the key component of intelligence is not the ability to compute (which we know how to replicate in a machine since the invention of the abacus), or the ability to put concepts into ever more abstract categories consistently (something that had to wait a few more millennia, to the advent of neural networks, and which is really all there is to most modern wonders like image and speech recognition). The key component of intelligence is the ability? Capability? Event? Of CARING about something. Of picking up an element of our perceptual field (or of our memory) and assigning some value to it. Of making it differentially important, relevant to us. Not surprisingly, there is one verb synonymous with “caring” which is highly relevant here: when we want to convey that we care about something we can also say that we MIND about it. We can use the gerund verb tense to describe the process and say we are “minding”: our own business, the outcome of some action, the decision someone we care about has to make… that is the true essence of intelligence, inseparable of the reality of having a conscious mind. A mind that “minds” about the things it perceives or the ideas it harbors, or the memories it stores. And that “minding” requires a number of things to happen: it requires a self (I mind about the pain in my knee in a totally different way that I would mind about a pain in yours, regardless of how well you describe it to me) that can identify, and follow, and monitor. It requires phenomenal experiences that I have direct access to (I’d rather use a technical terms that gives the proponents of strong AI the jeevies: it requires qualia, which they don’t have a clue about how to explain or how to start replicating in a machine). It requires a flow of conscience that I can control to a certain extent (so I stop minding my surroundings to do some introversion, to concentrate on some idea or some memory that I want to direct my attention to… but that attention thing is highly suspect to the aforementioned proponents; a homunculus! A ghost in the machine! They cry every time they hear the term) and that can even be interrupted by external causes (so when I’m anesthetized –or sleeping- there isn’t much I mind or feel or realize).

So I’ve always been very skeptic about AI claims because I recognized that we were entirely, utterly, completely, astoundingly clueless about how any of those essential components worked, and I firmly believed that until we had a passingly operational understanding of how they come to be we would be unable to even start working on anything resembling a conscience, and that without a functioning conscience we would never get the first inklings of anything resembling an intelligence. Call it my “working theory of intelligence”: Intelligence is the set of mechanisms our mind has for identifying another conscience. So you want an intelligent machine? Start by creating a conscious machine, and everything else will kinda sort itself out. And for the record, nobody has the darnedest clue on how to build a conscious machine, because nobody has the darnedest clue of what consciousness is (do not be fooled by books titled Consciousness Explained, as Raymond Tallis put it in his excellent Aping Mankind the author should be sued for breaching the Trade Act with such patently misleading description). My own hunch, which I have already presented elsewhere (yep, the shortcomings of dualism:  why monism sucks) is that the fact we are conscious ourselves, we have qualia, we mind what happens not just around us, but in the Universe in general (and specially to people we care about, but we may get somewhat self-referential here), we value things, or states-of-affairs (which is the same as the previous minding), we recognize non-material qualities (beauty, justice and truth, which I posit are more than “categories of categories of material things”) is in the end attributable to the fact that mind is substantially different from matter, and can not be reduced to it. So we can delude ourselves thinking our sex dolls (or the poor slaves we have robbed of their dignity and freedom and thus turned almost in a machine, regardless of their biological origin) have intelligence, or that any other device, for what is worth, has that same slippery quality, but it will be a delusion, and an easily dispelled one at that.

But may be it is not so easy to dispel, after all, as my theory of intelligence has one gigantic loophole: people (but let’s be honest here, I would be willing to bet my hat that it is mostly men) are willing to project the most astounding and bizarre qualities in those things that catch their fancy, specially if “catching their fancy” is a polite way of saying “making them believe they can give them sexual satisfaction”. So the nice “consciousness identifying mechanism” theory collapses when the dick takes charge and sees some opportunity of giving himself some action. I can see droves of guys ordering “robots” (and paying good bucks for ‘em) that can give them a conversation that, in the lips of a “real” human being would point her out as barely literate, but then coming out of the experience feeling they had a thoughtful exchange with a mixture of Dostoevsky and Aristotle (well, as long as it is followed by a not so thoughtful but more physical exchange, that is). If the thing catches on, the demand for some talking capability can only go up, as the window dressing in these matters is as important as the window herself. Call it the “Playboy effect”, as in the reputed magazine of the same name, where you could find some notable pieces of journalism that gave it a veneer of respectability (and the wild dollars they made by selling photos of naked girls to excite the imagination and facilitate the masturbation of kids that probably could have achieved the same effect looking at the photograph of a brick allowed the editors to hire and finely pay some outstanding authors, by the way). If talking sex robots kinda become mainstream I can very well envision the iterative development of more and more sophisticated interaction capabilities, until in the end they become more or less indistinguishable from the real thing (indistinguishable for someone who is sure to get the thing to the sack with no resistance at all, were there some doubts as to the final result I guess people would be more discerning).

So I would expect five years from now to have some of my friends point to me the astounding advances in the conversational abilities of the latest sex toy (I hope that is all they  tactfully point, as I don’t expect to be all that interested in their abilities beyond that, which will surely develop as exponentially) to try to convince me of the imminence of the advent of fully developed AI. Of course, that particular astonishment will have to compensate for self-driving cars never having materialized commercially by then, translation software being almost as clunky as it is today, and no machine yet having won the Loebner prize (for a good report on how that works, see The most human human). Well, at least by then we will be only 25 years from Kurzweil’s predicted date for the arrival of the Singularity, close enough to confirm the idea was bonkers all along       

No comments:

Post a Comment