We're Laughing Now... David Pogue/CBS Sunday Morning on AI

CBS Sunday Morning, June 14, 2015

Nice little piece on the future of robots and artificial intelligence occasioned by DARPA's latest competition, full of untested and hotly disputed hypotheses on timelines and degree of threat:

Also, not sure how I've gone so long without a viewing (probably because it's been almost impossible to rent/stream), but just this weekend managed to see Her. Creepy, beautiful, awkward ... in short, a convincingly well-studied speculative depiction of a possible near-future world where ASI is beginning to dawn. If you haven't seen it yet, it's worth a belated screening:

Putting the I into i, Robot

Metropolis, 1927

Having spent some time with an alternative model's possible answer to the hard problem of consciousness, let's put it aside and turn to the dominant materialist paradigm's treatment - and current state of investigation - of the problem and the efforts to create consciousness that take for granted that the materialist paradigm is true and complete. Indeed, proponents of this view generally believe that any alternative view is laughable and that they are well along in the quest of both understanding and then being able to replicate the requisite parts necessary for a constructed replica of the human brain to achieve human-equivalent self-aware consciousness.

Tony Prescott, professor of cognitive neuroscience at the University of Sheffield, UK, and director of Sheffield Robotics, offers a stimulating and philosophically aware account of his iCub project in the March 21-27th edition of New Scientist (paid subscription required for full article). The article is worth a read - he posits that of Ulric Neisser's five necessary components to human-equivalent consciousness or selfhood, three of the five have been achieved by their iCub creation.

Behold the somehow endearing - and yet simultaneously deeply creepy - iCub in its early childhood: 

Have an hour or so to contemplate what might perhaps be the most significant mystery of human existence: how our brains might create consciousness and therefore how a physical or virtual replica of the human brain might produce it to complete the achievement of inserting an i into i, Robot? If so, the following short talks might be helpful.

First up, Dan Dennet, the co-director of the Center for Cognitive Studies at Tufts University. Dennett's larger position seems to be that there really is no "hard problem" of consciousness, there is no phenomenon of subjective, willful self-aware consciousness to explain - it is an illusion. His point in the following talk is really just that we can't trust the subjective feeling of consciousness and that our almost innate certainty of its existence is because we "experience" consciousness - but that subjective experience is based on trickery in the same way we can easily be tricked by the magician's clever slight of hand or simple optical illusions:

While it deserves a hearing, I personally find Dennett's argument here to be weak sauce and, like many of his "new atheist" bretheren, Dennett seems to be very busy building and burning straw men, missing the forest for the trees and refuting assertions no one of note from the other side is making. I find the magician's slight of hand metaphor to be a more apt descriptor of his alleged chain of logic and what he proports to be able to demonstrate. Nevertheless, this is the sort of somewhat ridiculous tautology (There is no consciousness because consciousness doesn't exist / I don't have to explain consciousness because we are not really conscious) that must be faced and challenged for anything outside the dominant paradigm to get a fair hearing.

For me, the more compelling arguments for the materialists resort to the evidence that there are specific brain structures that correspond to every aspect of human consciousness - and that altering, injuring and/or destroying these structures have entirely predictable affects on specific functions and qualities of consciousness. V.S. Ramachandran's talk is a good example of the sort of survey that might immediately lead one (though it is important to note that Ramachandran is not making this assertion nor denying it here) to the conclusion that if the phenomenon under consideration is altered or stopped when a physical structure is altered, injured or destroyed that the latter must be the necessary cause or source of the former. That is the natural assumption with the following associations between specific phenomena of consciousness and their corresponding brain structures:

Ironically, I actually think something like Dennett's slight of hand might be more at work in our very natural assumptions about causality, but I'll save that for a subsequent post detailing the more moderate / possibly middle-ground view of another theorist.

I'll end today's rabbit hole adventure by checking out the state of the materialists' quest to create a build-a-brain workshop. Henry Markram, the Israeli neuroscientist and director of the Blue Brain Project and the subsequent Human Brain Project, outlined in 2009 the plan to virtually model the human brain in a supercomputer. This is where the rubber is meeting the road in attempting to achieve full consciousness for an artificially conscious entity or robot, assuming the materialists' assumptions are correct and that they have within their reach the ability to solve the raw and immense technical challenges involved in duplicating the structures and processes involved in the human brain's production of consciousness:

Almost six years later, just this past month, Markram delivered an address at the World Economic Forum in Davos, which may provide some good clues as to the status of their quest, the first half of which is reportedly roughly on schedule. If you're like me, you might need to stop and brew a pot of coffee. A stem-winding silver-tongued orator he most certainly is not, but he's very, very smart and he's in charge of one of the most ambitious and potentially important initiatives that may yet prove to be foundational to the achievement of artificial super intelligence in this century. Get ready for human brain chips and Roboy!:

Robot Cars!

Robot Car, DJ Shin, cc Creative Commons

Okay, so it's not a flying car, but robot cars (autonomous vehicles) are starting to appear on the near horizon in a couple of recent announcements. Mr. "summoning the demon" himself, Elon Musk, announced this week that Tesla will provide a software update this summer for its Model S sedans that will enable them to function in "autopilot" mode while on highways.  Google's "director of self-driving cars," Chris Urmson, projects that fully autonomous cars will be available to the public and on the roads within the next five years, in time to ensure his now 11-year-old son won't have to take a driver's test. These are very bold predictions, and are probably taking a good chunk of the population, perhaps the vast majority, by surprise. Urmson's declaration was made at session 3 of TED 2015, and was accompanied there by more than a little discussion around artificial intelligence and some of the complicated technological, philosophical and ethical issues and topics that have been the focused interests of this blog.

But even just taking the topic of the decidedly un-super artificial intelligence of self-driving cars, one can easily see current technological capability on a collision course with our ability to regulate its development and / or revise our norms and laws to even accomodate it, as a quick review of the New York Times article on Musk's announcement will demonstrate. If we're getting stumped in legally dealing with auto-cars and Amazon's wished-for delivery drones, how much more difficulty will our legal and ethical systems have with artificial super intelligence? Here we are in 2015, very much at the base of one of Kurzweil's exponential S-curves and already the issues seem overwhelming.

Personally, all this reminds me of a recurring nightmare I had as a child in which our hulking Chevy family station wagon (the kind with the vomit-inducing rear-facing seat) would take off by itself with me strapped in the back, panicked as the car mumbled non-sensically back to me in response to my pleas that it stop and let me out. So, strap yourself in, kiddos, it's going to be a bumpy ride, and there's a fair to middlin' chance no one's going to stop and let us get out. 

The good news is that the demon cars look a lot more comfortable and luxurious than my family's 1970s Chevy station wagon. But, I'd bet everything I have that they'll still make little kids vomit:

The Future isn't what it used to be...

A Trip to the Moon, 1902, Georges Méliès

Much of what has been discussed on this blog to date has been theoretical and abstract. Abstractions and theory are perhaps the relatively more stable and reliable elements of prognostication, but they lack soul - they are the skeleton framework of dry bones that hunger for the flesh and blood of imagined and lush new worlds. As we gaze into the abyss of the event horizon, these questions echo only, without answer: "But what will the future look like, feel like, taste like?" "What will it mean?" "Will we still be ourselves?" 

All of these questions bounce off of the offered mathmatical extrapolations and exponential S-curves which can whisper only about what is theoretically possible, that can tell us only that what we're trying to imagine ultimately may not be within our merely human conceptual grasp.

But we imagine still. If you're a Gen-Xer, like me, the perenial questions, "Where's my jet pack?" "Where's my flying car?" linger to cast a healthy dose of skepticism in reply to any fool who comes along claiming to tell us what the future is going to be like. And all these imaginings from time immemorial probably only really tell us something accurate about our present, as they are simply the projections of ourselves at very particular points of time and the context framing us in the present. If we're Jules Verne or George Méliès, our future cosmonauts wear bloomers and are shot at the moon out of giant gold cannons. If we're Gene Roddenberry, our Starship Enterprise crew wears 1960s bouffants and only more refined polyester combos of leisure/track suits. If we're producing a 1976 SNL skit peering into 1999, everyone then sports identical polyester getups (again) and giant afros. Even when we're not trying to be funny, there are elements of our present we simply just can't imagine being dated, that trip us up, that evoke chuckles and giggles only with decades of hindsight. So much more so when contemplating a radical break like the singularity.

But, sometimes, we get it right. Certainly, no major advance in human technological history happened without being first imagined and most often presented in fantastical science fiction. And so science fiction is a menu of possibility and most often the choices are dictated less by true technical limitation (though that is certainly sometimes the case) than what our future selves find important enough to achieve (somehow, jet packs and flying cars never really developed a compelling market for themselves - a Manhattan project-style initiative for either or both and we'd have flying cars and jet packs aplenty). Our imagination fails most often in anticipating what we'll actually find to be important and crucial in the future. When it does, our imagination fails then in intuiting the evolution of our desire and our own deepest internal motivations. Ultimately, we are a mystery most unto our selves.

With all that said, and with the amount this blog has dwelled on the apocalyptic dangers lurking in the development of AI and ASI, the utopian futurist vision of Ray Kurzweil deserves a hearing. It's compelling and beautiful and seductive. It leaves you wondering if he and others like him are dangerous crackpots or if they just might be on to something. I suspect a little of both and I've probably got a strong history of precedent on my side in imagining that about them. At any rate, enjoy a picture of what could be if he and others like him are 100% right about the most important things and if all goes well in our dangerous quest, straight from the horse's mouth. Nothing huge promised, only immortality:

And here's a recent and condensed version of a talk he's been giving now for a few years:

Finally, enjoy one of my favorite bands - who provided the inspiration for the title of this post:

The Mind-Brain Problem, Part I

Michaelangelo's The Creation of Adam. Michaelangelo's depiction of the deity seems to have been based upon an anatomically correct template of the human brain.

Earlier, I referenced the most common (almost ubiquitous) scientific presupposition about the brain and willful human consciousness: that the the human brain is the material, biological mechanism that is all that is necessary to create willful human consciousness. At the risk of oversimplifying things, let's just label this the "materialist" presupposition about the relationship between the brain and the mind. No physical brain = no mind. It's hard to imagine anyone taking seriously any alternative. But, there is one, and it's got a surprising history and cast of proponents and supporters. 

At the outset, one can be forgiven for wondering: how exactly is this relevant to artificial intelligence? Permit me the argument: If the materialist presupposition turns out to be incorrect, or even incomplete, in its articulation of the truth about how the brain and mind interact, the implications are profound for the development of AI. That would mean that creating a digital replica of the brain (made not of flesh but a highly sophisticated virtualization mimicking both macro brain structures from the thalamus, cortex and reticular formation, to the frontal, prefrontal and temporal lobes, corpus callosum, etc., etc., and micro structures, from neurons and synapses to a bath of neurotransmitters, etc.) might not be all that's required to create self-aware and human-equivalent consciousness. While such a replication does seem to happen quite reliably many thousands of times a day in biological flesh form across this planet as human babies are born and then develop into little humans with what appears reliably to our eye as self-aware consciousness, we would be wise to check our assumptions that digitally replicating these structures, components and processes would necessarily lead to consciousness. While there is absolutely no denying that the brain must be somehow very significantly involved in consciousness, if our understanding is incorrect or merely incomplete, what we don't understand represents the x-factor which could complicate, frustrate or even prohibit any development of artificial superintelligence which assumes self-aware consciousness as a foundational prerequisite.

Unanswered (and surely important too) is the question of the degree to which conscious and self-aware thought is dependent upon the vast amount of unconscious mental activity and even unconscious thought and ideation in the human mind. One imagines AI theorists and researchers simply wishing to avoid that thorny unanswered question by seeking to develop a digital model or replica of the entire brain so as to not unintentionally and unknowingly leave anything crucial out of the equation. From a purely materialist perspective, this all of course begs the question as to whether the first self-aware robotic entities might also possess significant unconscious thought, desires, etc. like their human models do - a very unsettling prospect...

While questioning the materialist presupposition seems on the face of things to be entirely laughable, there are glimmers and hints even in recent medical literature that all might not be entirely well in terms of fitting all the evidence to this most foundational of assumptions about how self-aware consciousness comes into existence. We'll explore this crucial question and the alternative model more in coming weeks, but for now, consider just one of the more recent cases that suggests that something might be incomplete in our understanding of the relationship between the brain and mind: a 44 year old-white collar worker who, due to hydrocephalis, is missing the vast majority of the mass of his brain, and, while possessing a below average IQ, tested comfortably above the clinical threshold for intellectual disability and has managed to lead a seemingly normal life. The case lays perhaps far too heavy a burden on the explanatory power of "brain plasticity" and suggests that the brain, the mind and consciousness and the relationships among them are all still quite mysterious to our understandings - scientific, medical and otherwise.

Even from a purely materialist perspective, stumbling blindly into such a depth of mystery while attempting to create an entity with such immense potential power does evoke exactly the sort of Faustian scene that Musk references with his "summoning the demon." So, wouldn't it be wise to first apply narrow forms of AI or even restrictive forms of general AI to the mind-brain problem to be absolutely sure we've actually answered some of these questions prior to attempting to create anything capable of artificial super intelligence? Maybe. But answering such questions even with the assistance of the lesser but benign and fully controllable forms of AI would take centuries of perfect and consistently applied limitation of already realized capability. In the face of the degree of extreme self discipline and self checking required, across successive generations of the very human collection of theorists, researchers and industry leaders involved, every last cent of my money is stacked on the wise (or the modest, disciplined, restrained or non-greedy) choice never being our first choice.

Those faced with such a choice when the time comes may not be clutching Musk's pentagram and holy water, but you can be damned sure they'll be blowing on the dice cupped in their sweaty hands. 

Part II (in coming weeks): what is the alternative model?

Deus Ex Machina

"Deus Ex Machina I" by Mall Nukke, Creative Commons

For this one-time history of religions student, perhaps one of the most interesting and personally stimulating developments in the contemporary discussions of artificial intelligence is the number of ways spiritual / religious ideation seems to be creeping into view. This isn't really even just implicit in Kurzweil's thinking, it's quite explicit. This shouldn't be altogether surprising, considering the grand speculation as to what eventual (and some would say inevitable) advances in this field will produce: a willful, thinking, conscious entity that is capable of perfecting itself after discarding the human template upon which it is initially constructed, and then disposing of anything we'd understand to be "machine" in a conventional sense. It's difficult for me not to see this central idea as the Genesis story in reverse. The randomly-evolved-uncreated human race paradoxically creates a God... which could then create anything... or any old godly action it might have a desire to do... And it's the irony of ironies - or perhaps just testament to just how strong the human animal's spiritual / religious instinct is - for better or worse - that a bunch of mostly atheist and agnostic theorists are essentially talking about creating just the sort of God they don't believe exists (yet). It's clear some also fear creating something that might fall short of the mark of a perfectly benevolent supreme being vis a vis humanity, perhaps best expressed in "summoning the demon."

So, let's play with this just a little. Kurzweil and others see nothing stopping an artificially super intelligent consciousness from rapidly developing beyond the singularity moment to exploit every molecule/atom (and anything smaller which might lurk about in existence or semi-existence) in the universe as a data storage device, with nothing presumably then preventing such an entity from directly accessing the mind-consciousness of any limited being within that universe (or multi-verses, etc., etc.), assuming that such an entity would have any interest at all in any such limited consciousnesses within its reach and purview (an inherently unlimited reach and purview). While some may criticize 17th and 18th century enlightenment intellectuals and their rational theism as not being wholly consistent with the God of Hebrew scriptures nor his/her/its other Abrahamic incarnations (Christianity, Islam), the notions of omnipotence, omniscience start to rear their heads here immediately and the entity possessing such abilities starts to look and sound an awful lot like any god or God conceived by rational theism - which, we must admit, is really the sort of God all of us 21st century human beings, religious or not, are walking around with in our heads when we think about or imagine "God."

Is all this Anselm's ontological argument twisted and turned on its head - that if the idea of such a God exists in the minds of human beings, they must necessarily be able to eventually create such an entity? Let's sprinkle in some theoretical physics and quantum speculation and other such voodoo mumbo jumbo and pretend that we actually understand what it means when physicists tell us time doesn't really exist, that our notions of conventional one-way temporal causality might be an illusion, and then does it really matter when such a "God" comes into being, or how? If the laws of nature and time and space are accessible and discoverable to any conscious mind of sufficient intelligence, and surely fully realized by a perfected, unimaginably powerful mind like what we're talking about here, would it matter that "God" is created in a particular point of time? What then restricts its influence and reach into the past or future and into the vast reaches of the universe at all times? Anything? Interestingly, the theological / christological assertions implicit in Christian doctrine related to the incarnation would posit that it doesn't matter that the incarnation is a historical event, entering in just one point of time - the fact that it did at all is somehow eternally efficacious/salvific.

But, just for the sake of continued speculation as to the possibilities, let's together ignore any such theological resemblances and willfully adopt the stereotypically reductionistic / atheistic / agnostic / scientific view that might say that religious scriptures are complete hooey, nonsense, merely mythological, merely fable. But...but, if these myths / fables have so influenced and shaped human culture, thinking and imagination - and the thinking and imaginations of even those who claim to reject them - that we end up creating just the sort of conscious supreme being to which these myths and fables testify, can they really be dismissed as merely myths / fables, even in that case? Or would they rather in that case be prophetic - in every sense of the word? Again, one might hear faint echoes of the ontological argument's "thus, if God exists in the mind as an idea, then God necessarily exists in reality."

Let's take all this even a little further. Let's suppose our universe is constructed in just such a way (or just so happens to exist) as to allow the above to take place (even if it is incredibly rare), that random chance might eventually produce conscious entities, like humans, who are then capable in turn of creating entities that can then self-evolve into the God-like creature described above. Is it really plausible to think that among the vast stretches of our galaxy and then of what we understand to be our universe that here in this place we are the very first to witness this happening, the very first to be involved in such a "singularity" creation? Just as it's pretty unlikely that life as we understand it or intelligent civilization uniquely exists on this one planet in this one galaxy if it is all the result of merely random events, it's likewise rather unlikely that artificial super intelligence hasn't also already been achieved elsewhere in the past... meaning, we've already been anciently accessible to it, already fully known by it (assuming it would even care about or couldn't avoid knowing about such things as ourselves). Maybe, maybe we have even been created by it - maybe the creator pre-existed us after all in a way that is completely consistent with conventional notions of temporal causality - maybe we are already Leibnitz's monads, inherently caught in the jeweled net of Indra, all a part of fully conscious ultimate-entity looking out upon itself from an infinite number of limited perception points ... or maybe we really are in the Matrix, maybe... maybe... maybe...

But, let's get real. The above is probably not a whole lot more precise than a 3AM undergraduate dorm room rap session - and surely the more developed aspects of the speculation rest very precariously on a long series of assumptions, necessarily accepted a priori as unprovable givens - even if these are all presuppositions/assumptions that are virtually ubiquitous in our human culture and thinking (some of which might even be true!). We must ultimately admit that the "singularity" nomenclature of the phenomenon and entity we're imagining as a possibility (which borrows of course from scientific speculation about the properties of black holes in space) is an analogy that serves us fairly well in that all this is an event horizon beyond which we really just can't see... at all. We are left only with our very limited imaginations in thinking about our own possible creation, an artificially super intelligent/conscious entity. Given that, are we perhaps left then to speculate about all this with something as humble and scandalous to the human scientific mind as theology?

Let's assume for the moment that we are on the precipice of something truly unprecedented and let's assume something a bit more sophisticated than Asimov's laws might be required (just how quickly in its development would such an entity disregard such laws or simply decide that any regulative term and certainly one as facile as "robot" doesn't apply to it?). I'd submit that's necessarily ALL we have, that the templates laid out by centuries of theology and religious philosophy (Buddhism, Hinduism, etc.) in their most progressive and humane forms (like most semi-rational theists, I'm necessarily leaving out all empty-minded fundamentalisms and any and all violent manifestations - sue me) might be the best or only things we have in imagining the sort of entities and states of consciousness that may be created and unleashed in the pursuit of artificial super intelligence. And, if we've got a bunch of AI theorists/scientists/industry titans theoretically playing with apocalyptic fire in the ways even they are admitting they are, would it be at all wise to involve squishy-field folks like ethicists, religious thinkers, psychologists, theologians and philosophers in this supremely dangerous endeavor? Would they make it any more or less likely that we end up creating a super-then-supra-machine-God that will have compassion, even something like love for us, that it may offer to let us evolve with it, over a super-then-supra-machine-God that will instead simply and swiftly eradicate us as the irrational and extreme threat (to ourselves and each other and all the other sentient creatures who share our lonely-little-miraculously-rare planet) that we so often provide evidence to be?

Do we really have any power to influence this one way or the other at all?

God only knows... 

CBS This Morning Tackles AI

Enjoyed seeing this segment pop up in my favorite morning news program this morning. - [Original video removed from public site by CBS - replaced with an article, the above is a subsequent video treatment of AI questions from CBS This Morning - March 9, 2015]

And Now, A Word from Our Sponsor: Old Glory Insurance

And lest you think the threat isn't a real and present danger, behold the havoc that already ensues... 

On a more serious note, presented for your consideration are the variety of ways that simple and not-intelligent-at-all pieces of code (algorithms) already routinely intersect with our lives, often in uanticipated ways, attacking our Facebook feeds, discriminating against us, crashing our financial markets, denying our mortgages or yanking the protections of citizenship wherever they are allowed to run amok, unchecked by mere humans. On the upside, there are those occasional $.99 pajamas and knicknacks glitches at Amazon...and countless other life saving and miraculous benefits of our contemporary civilization, but still, the danger is real people! And Old Glory is just $4 a month...