I was watching a James Cameron interview on the Science Channel's website and they asked him what he thought the next big thing was. And he mentioned two things: faster frame rates and making things brighter. Unfortunately, the full-length interview is no longer available.
Cameron talks about how they started doing 3D differently. Instead of doing it the old fashioned IMAX way, with a pair of cameras side-by-side, he talks about how they started experimenting with "active convergence"- which means having two cameras and changing the angle between them- parallel for objects out at infinity, angled together to converge on something close. This is the basis of stereoscopic vision. Human eyes do the same thing. When you look at something really close to your face, you go cross-eyed. And that's a big part of what tells you that it's really close. The double image produced in the non-focused background, on the other hand, is an indirect cue. The main reason you don't see double-image artifacts is that film making isn't that close. You could, however, create quite a miniature effect by using it (in the next 3D Bugs movie, such techniques might be used).
Cameron et al quickly discover that active convergence has limitations. Namely, the fact that movie cameras are enormous compared to human eyes. It takes a large lens to match the resolution of our natural optics. And you can't put two large lenses too close together without having them crash into one another. So they started doint things differently: "We completely revolutionized... when we went from a parallel system to a beam splitter system." And that's the big gimmick with Avatar: that 3D isn't a gimmick any more. Unlike past 3D movies, where the 3D was omnipresent (think of Beowulf for instance), Avatar's 3D was designed to go unnoticed. In fact, Avatar represents the first time 3D wasn't being used to create scenes with impossibly deep 3D that don't represent how the eye and brain actually work. Think of the spear scene in Beowulf. The tip of the spear is pointed at the viewer's eye. Part of what tells you that it's so close is that it's huge compared to the background. But if you were actually to have a spear in your face, and you were to focus directly on it (as it appears in the shot), you'd also see a blurry double image of the background. But the makers of Beowulf keep the background clean and singular. After all, the mind can derive 3D from the primary cue without resorting to distracting secondary cues. In making Avatar, Cameron uses active convergence, enabled by beam splitters, to operate in the middle distance. No cross-eyed shots. Avatar is a movie of dense foreground shots- numerous 3D plants- and big sky long shots. There are several shots that are extreme closeups- Jake's face when he emerges from hypersleep, or while he's opening his own eyes after finishing a link session. But these shots have no background cues, so we never experience the lack of secondary cues.
Here's where I get crazy. Cameron says we're finally reaching a plateau.
Asked whether we'll ever be able to take off the glasses, Cameron says no. He says, "For a big theatrical exposition in a movie theater, you're always going to wear glasses." (And then Jon Landau, Avatar's producer, fantasizes about using the glasses to make money: "It's not that I have to wear the glasses, but that I get to wear the glasses" and goes on to spout a fantasy about how your glasses could also be your ticket, paid for at home.)
First of all, Cameron is wrong about the plateau. But just as Cameron represents the (half)generation of filmmakers that were inspired to surpass George Lucas at his own game, there is plenty of room for new innovation. I'd like to say something about George Lucas. People may not have been impressed with his new Star Wars trilogy, but he, and ILM, did a massive amount of innovation a whole ten years before Cameron pulled the trigger on Avatar. Just as Cameron and (many) others spent fifteen years (post Return of the Jedi) moving the tech forward for Lucas.
Cameron mentioned two innovations: faster frame rates and brighter pictures. Both would be significant. By "brighter" he may mean "higher contrast." Film projection allows contrast ratios of thousands to one. Some TV displays allow over a million to one. That's the difference between looking at a photograph of a sunset and actually feeling like you're looking at the setting sun. One of the biggest obstacles to the suspension of disbelief is that movies are relatively dark. That's one of the reasons that the cinema style favors relatively dark scenes, at night, under clouds. Using artful but inherently unnatural lighting (oh, how bright the moon is, and how conveniently located). Higher contrast would both add a level of realism never before imagined and freedom from the constraints of traditional lighting.
Higher frame rates would provide some interesting benefits. The most important thing it would do would be to allow fast POV action shots to be believably achieved. That's because, at 24 fps, you almost can't pan at all without creating stroboscopic artifacts that tell the brain that what it's seeing is not real. In other words, the camera can't turn its head without making the scene unintelligible. There are some tricky ways around this: artificial motion blur, carefully establishing spatiality before making a motion (using an establishing shot for instance). The best way around it is to keep the camera still. That's what movie making equipment was designed to do in the first place.
But if you double or quadruple the frame rate, and you can also quadruple your pan speed and stop relying on tricks. This would allow the film maker to break the proprioception barrier- to get into the head of the viewer, to get them to identify kinesthetically with your characters. Many people who saw Speed Racer (2008) missed the point(s). It was panned by people who called themselves "film critics," but were actually operating on a purely emotional level. What made Speed Racer great was that it made so many of its viewers sick. What kind of sickness? Motion sickness. Because the movie convinced them that they were moving in ways they weren't prepared for. The only way to enjoy a movie like that is to do so actively: to put yourself in the driver's seat. Because it was so innovative in this and other ways, it still managed to make Richard Corliss's top ten list. Just imagine what the Wachowski Bros could have done with a faster frame rate?
And now for the reason I'm writing this blog: Let's talk about how to make 3D more believable.
When you're watching a 3D movie, the film makers have decided what your eye can look at, what is in focus. This used to be done in more subtle ways- the use of eye dominants (the brightest thing on screen should always bee the thing you want to draw attention to- a rule made to be broken). The use of aberrant motion. Simply using color contrast. And the oldest trick of them all: movie stars. Pretty faces to attract the eye.
With 3D techniques, film makers have another trick in the bag: the ability to constrain the focal plane on the object-of-focus. And that works well.
However, the illusion of 3D arises in the brain, not in the eye, not in the picture the eye is looking at. By dictating what the eye can focus on- but giving it only one option- a huge part of the illusion is lost. Unfortunately, there's only two ways around this (that I know of). Both are difficult to do.
One, the oldest and hardest to revive, we'll call the multi-planar staging approach. Imagine instead of projecting your 3D image on a single flat screen, you project images on numerous screens that are actually at different distances from the eye. Besides being very difficult to film for and to execute, it also only works well for only a couple people sitting in the best seats. And projection is out of the question: projection screens have to be physically moved out of the way because they are, of necessity, not transparent. Instead, you use special screens with countless switches allowing flat pixels (super bright LEDs- the brighter, the smaller) to be reoriented so you're seeing them edge on- making them almost transparent when not in use. And you'd use these screens to show 3D images to add extra depth.
This would allow the eye to choose its own focal plane, which would create a sense of 3D that would make Avatar look like King Kong (the first one) in comparison.
You'd really be creating a simplified hologram. But unlike the traditional application of hologram technology, this one would look enormous (even if it was no bigger than a home theater) and at least half-immersive.
The other way is actually easier, even though it isn't technologically feasible (yet).
Instead of projecting multiple screens at multiple distances and letting the eye choose, you simply track what the eye is looking at and provide an in-focus image before in less time than it takes the eye to focus or the mind to take notice- somewhere around 1/15th of a second. You just need fast computers (about a magnitude more powerful than the present crop of affordable GPUs). At this point, the movie theater becomes obsolete. The best way to achieve this would be with head wearable displays. Virtual reality "glasses." So Cameron would be right about not taking off the glasses. He might be wrong about the future of "big theatrical expositions" though. That's likely to change in some slow subtle ways.
It would also pose a considerable challenge to film makers. Cameras would need to collect a distance map (using an array of laser range finders) of the image that could be used to translate the image into a virtual hologram of the image in post. One cool thing is that your could use a single camera to produce a 3D image- you'd just have to have separate motion trackers, display drivers, and displays for each eye.
It would take some very fast computers but, ultimately, movies wouldn't become much more data heavy. It would be like adding an extra audio track. The technique could be used in both live action and CG-based movies. You could even 3D-ize movies that weren't filmed using the techniques (though the mind would suspend believe every time it tried to focus on something in an out-of-focus plane). In that sense, it would be better used for TV shows. Cheap TV shows filmed during the day. Woohoo.
Fast frame rates, much higher contrast, and self-selectable 3D focus planes- together these could produce a visual experience indistinguishable from reality. Of course, if you really want to go there, you'd need to be able to not just move your eyes. You'd need to be able to move your head too. And if we're really going for the next big thing, it should go without saying that the image should be at least 180 x 180 degrees wide and tall.
All that means is filming at a higher resolution and a wider angle. Not necessarily 360 degrees though (your camera crew needs to hide somewhere).
When? Actually quite soon. You'll start seeing cameras capable of encoding distance data in the next five years. Ultra HD will eventually become the next TV standard, but the exact technology hasn't been designed yet. I predict we'll see viable HWD's in the next five, and the technology will mature in the next ten. Fifteen years from now, we may expect feature films using self-selectable 3D. It'll be a test film, nothing like Avatar. Fully mature examples of the technology may take another five to ten years to arrive. So, I place the next plateau somewhere around 2030.
Is there a plateau after that If you've read my last post, you already know the answer. There's usually another level. In this case, I think the next level will arrive concurrently. It requires adding senses, especially breaking the proprioception barrier. Imagine sitting in a chair that sways back and forth to simulate walking, or spins you left and right to cue your inner ear that you're turning. Imagine having a virtual body- arms and legs that move as your point of view moves, or as you choose to move them, and which are lit in ways that are consistent with the scene you're immersed in. Why are arms and legs important if you can't interact with the movie?
We're talking about movies here, after all, not games- though there are already good examples of games that have exceeded the ambitions of theater (Call of Duty, Modern Warfare is a great example- the final battle, at least for me, was emotionally immersive in a way no movie has achieved- I personally felt responsible to save the millions of lives that would be snuffed out if I didn't succeed).
So, how does seeing your limbs break the proprioceptic barrier? Simple. Arms and legs are in the middle territory between the internal and external sensory fields. Your arms are embassies into "reality." The fact that you can move them, using signals from within, to affect part of what you see outside of you lends reality to everything that you can see because it belongs to the same visual continuum.
Try this exercise. Close your eyes in a completely dark room. Now, move your hand in front of your face. If you are indeed in a completely dark room you won't be able to see anything. Right? But you tell me, can you see your hand? Not visually, no. But you "see" something. Try changing the shape of your hand. Make a fist. Wiggle your fingers. Do you have any trouble visualizing the shape your hand has taken? Is it even "visualization" that your mind is engaged in?
The frontier beyond that is probably something internal- jacking into the optic nerve directly, creating false nerve signals, creating technologically-driven conscious dream states. The Holoband of Caprica instead of the Holodeck of Star Trek. And beyond that? For that, read my last post.
Thursday, February 25, 2010
Sunday, February 21, 2010
Shock Level Five: The Limits of Ultraintelligence
This is going to be long and to the point. I'm not going to explain what future shock is. Read this instead.
Just in case you didn't, here's an excerpt (begin quote):
Future Shock Levels:
SL0: The legendary average person is comfortable with modern technology - not so much the frontiers of modern technology, but the technology used in everyday life. Most people, TV anchors, journalists, politicians.
SL1: Virtual reality, living to be a hundred, "The Road Ahead", "To Renew America", "Future Shock", the frontiers of modern technology as seen by Wired magazine. Scientists, novelty-seekers, early-adopters, programmers, technophiles.
SL2: Medical immortality, interplanetary exploration, major genetic engineering, and new ("alien") cultures. The average SF fan.
SL3: Nanotechnology, human-equivalent AI, minor intelligence enhancement, uploading, total body revision, intergalactic exploration. Extropians and transhumanists.
SL4: The Singularity, Jupiter Brains, Powers, complete mental revision, ultraintelligence, posthumanity, Alpha-Point computing, Apotheosis, the total evaporation of "life as we know it." Singularitarians and not much else.
If there's a Shock Level Five, I'm not sure I want to know about it!
(end quote)
Catch that?
"If there's a shock level five I'm not sure I want to know about it."
If.
Which is to say, "there might not be." That it's possible, on a conceptual level, that nothing beyond what we can presently imagine will ever be imaginable. That there's not necessarily anything beyond "ultraintelligence."
Now remember, we're talking about levels of Future Shock. Shock is an very human concept: the feeling awe in the face of the ramifications of future possibilities. We're not talking about the future directly. We're not making statements about inevitability. Future Shock is a symptom of no-holds-barred extrapolation that are, nevertheless, firmly rooted in our present existence. As much as some people would like to pretend otherwise, there is no way to avoid experiencing future shock in the face of the future. Isaac Asimov was shocked by the emergence of personal computers and the internet. And yet, he wrote about godlike AIs. Arthur C. Clarke helped to pioneer the digital age. He actually invented the idea of the communication satallite. And yet, even he was shocked at the one-way cultural implications of what emerged. Shock occurs in countless ways. Microsoft's Encarta- written by paid professionals- was a massive failure compared to the shocking emergence of the volunteer-driven Wikipedia. Future shock is unavoidable. The only person immune to future shock is a dead person. It would take a knowledgeable intelligence of a higher Shock Level to not be shocked by its own level.
To question the possibility of a Shock Level Five is an honest question that, in some ways, is far more serious than what I'm about to undertake.
I'm going to ask what would a Shock Level Four ultraintelligence think about this question? We exist, with our slow, natural, organic intelligence, at SL0. Zero shock. That means that all levels of future shock are beyond us. But what would happen to the question if we were to find ourselves at SL1? Would we still consider SL4 to be the point-beyond-which nothing is imaginable?
And if a posthuman ultraintelligence (UI) with unlimited computing ability were to pose the question, how many levels of shock would there necessarily be?
What I can say for sure- since what we're talking about is shock levels- is that there would be at least one more level. It would be characterized by that which only a SL4 UI could imagine to be "unimaginable."
Before I continue, I need to diverge for a few paragraphs. I need to ask whether intelligence- the way we define intelligence- is actually scalable beyond a certain point. Is there a point at which a higher IQ just gets you the same answer that a lower IQ could have gotten, only faster? Are ultraintelligent minds able to think in transcendent ways? We know that the human mind is transcendent compared to computers- computers are our inventions. It's conceivable that a single human being, given sufficient time, freedom from senility, and opportunities-of-necessity, could invent everything that has been invented by individuals. Indeed, many inventions arrive from prolific individuals. Are UI's likely to be much better than a great number of normally-intelligent minds operating at a higher speed?
This is not a Penrosian critique of the possibility of AI, but I'm asking: what if the only kind of ultraintelligence that's possible is incapable of being self-aware because the internal vibrancy of thought causes external reality to become too slow, too dark, too uninteresting to have a self concept that exists relative to it? And what if self-awareness is necessary for intelligent intention- for acting in ways we would define as intelligent?
Let me pose the question this way. This is a thought experiment, and isn't meant to be taken as a conclusive argument, even if it draws conclusions.
Suppose you're a UI studying the goings-on of all the parts of the universe that "isn't you." Everything outside what you think of as "yourself." Let's say, for the sake of defining what we mean by "ultra," that your "mind" operates a billion times faster than a normal human mind, with billions of times more bandwidth- which we'll define as the ability to think about different salient details concurrently. Bigger minds take longer to operate, but if we use light-speed post-human "neurons" in place of 25mph neurochemical signals, we can get a truly huge brain operating at fantastic speeds.
Let's say that, with just a few of your innumerable "eyes," you stare right at the surface of the sun. Because you think a billion times faster, it takes you over a hundred thousand years to observe one hour of solar evolution. You might learn an unimaginable number of things about the sun in that one hour of "objective" time. Perhaps more than what a billion scientists could learn given a billion lifetimes. But you couldn't learn more than was available to know. And you wouldn't be able to separate what a thing is from how it's experienced. For a billion-X UI, the universe holds a different meaning. Subatomic activity ceases to be either a statistical firehose or a needle-wide window. Instead, you are able to watch a million square miles in simultaneous microscopic slow motion. And if that's the way you see the universe, then that becomes the normal way of seeing things. The universe becomes a horse of a different color. And that kind of difference would lead to insights unanticipated by normal-speed minds.
Same thing if you were to operate more slowly. An ultraintelligence able to operate a billion times more slowly would have another set of advantages: the ability to perceive over two million years of solar activity per personal "day." The more natural place to put one's attention would be on the macroscale- where big things are happening constantly. Dozens of supernova a minute. Galactic gravity unfolding before your eyes in realtime. Forests changing to deserts faster than a cloud moves across the land. Again, the universe would take on an entirely different meaning derived from a different mode of perception and interaction.
Can you have both kinds of thinking? The most straightforward answer is simply, "No." For a single ultraintelligence to have a cohesive awareness, a single reference frame is, by definition, necessary-- unless we first decide to define words like "cohesive," "awareness" and "intelligence" differently. Granted, you could have numerous non-simultaneous consciousnesses running concurrently, at different speeds, with different perceptual foci, and then integrate their thought products after-the-fact (like watching memory-movies created by alternate lives). However, the fastest consciousnesses will always outweigh the slower consciousnesses because they produce more "movies." The movies produced by fast consciousness would define the UI's native realtime. The fastest consciousness is consciousness. Anything less than full-speed would have a minority share in contributing to the identity of the intelligence (proportional to the inverse of the ratio of the differences in speed). Between the two examples above is a factor of 10^18.
So, as an ultraintelligence with vast computing capacity at your disposal, the ability to write your own operating "software," and the option to operate very fast, what would you do? Pretty simple. You'd take what you learned with your slow-motion ultra-wide microscope and you'd internalize the workings of reality. You'd simulate timescales from within your primary, full-speed consciousness. Why wait billions or trillions of years for something you can easily acquire (as an incomplete but constantly improving) picture in a matter of moments?
Without trying to go into detail, let's consider: what does an ultraintelligence think about? The simplistic answer is: reality, and not just the one that it exists within, but many variations on that one, and perhaps many others that we wouldn't consider to be parts of "reality." But our perceptions are biased. Essentially: many variations operating at many speeds- each producing questions tending to be answerable. Even a UI would have trouble thinking about what it can't think about. And that's the heart of where I'm headed.
If the majority of your thoughts are about events that originate within yourself, then, for all intents and purposes, you are your reality. The concept of self-awareness, as we have heretofore defined it, becomes nonsensical. Self-awareness requires "self" to be the thing that is aware. That "self" must have something that is distinct and separate from- something incompletely known- in order to define the parameters of self. If you know everything- or consider nothing that is unknown- then all things are, in essence, within you. And if there is no distinction, there can be no awareness of distinction. Hence, the grander the UI, the less there is of "self."
Now, when we're talking about hypothetical posthuman singularity-style ultraintelligences with planet-sized minds and the ability to exploit the fine structure of reality for carrying out its computational whims, we're certainly not talking about an "infinite" intelligence. The conditions described above are a matter of degree, not of absolutes. At some degree, however, the concept of self-awareness becomes obsolete as a primary mode of description. In fact, a UI might actually choose to represent itself as some number of self-aware individuals within itself- each distinct but fully compatible with its other manifestations.
Next question: what can an ultraintelligence think about? What kind of "life of the mind" can such a being expect? It is conceivable that such an intelligence would spend an enormous amount of thought seeking meaningful answers to this very question even before it reached the level of ultraintelligence (for instance, by reading this blog). By necessity of achieving Shock-Level transcendence, ways of thinking would be devised to maintain contact with modes of being that the previous incarnation would recognize as meaningful. That's not to say that an ultraintelligence wouldn't evolve rapidly into something we'd consider absolutely alien. But, from the UI's own perspective, that evolution might be very gradual and cautious. The stakes are vast, after all. A UI faces a kind of oblivion, a loss of self by drowning in a thoughtspace where it is almost the only thing that exists; where the only things left to think about are permutations of thoughts already thought.
If a Shock Level Four UI can reach the state of UI without closing its horizons to coincide with the divide between it and outer reality, then what it is contemplating is, necessarily, what we'd have to consider to be the subject material of Future Shock Level Five.
Something beyond itself, and perhaps always beyond itself. If any intelligence can contemplate something beyond imagining, it would be- by analogy with levels one-through-four, a far greater UI. And the thing beyond imagining would, by necessary exclusion, be a mode of intelligence beyond its own ability to attain because of certain impairments to its candidacy for Level Five.
One of the ways an ultraintelligence of any amount of power is irrevocably impaired is by the fact that it has evolved from a lower starting place- a place from which its primitive memories are still available. A hypothetical SL4 UI knows, and cannot help but know, the full story of its existence. Therefore, it knows- if only by analogy- that its existence is (hopefully) one of ever-increasing vibrancy in the forward temporal direction, and ever-decreasing vibrancy in the negative (back in time). Impairments in time equate to impairments in space- you can't see beyond your lightcone. And time travel, if it were possible at all, would make causality, and therefore, "knowability" meaningless. It would destroy the timeline that gave it rise and would, therefore, happen a maximum of once per universe. Therefore, even if a UI was capable of time travel, it couldn't affect the limitations of its personal past.
Any UI capable of knowing itself at all would know that it is impaired on certain space-time vectors and that the absence of these impairments would lead to a horse of a different existence- an alien sort of intelligent identity. The UI would, therefore, consider the ever-transcendent Shock Level Five intelligence to be one without any impairment vectors. One with all the benefits of time travel without collapsing causality. One that can consider variables and alternatives within itself without losing, or even damaging the concept of "self."
At Shock Level Zero, many intellectually rigorous individuals tend to consider the idea of a Level 5 UI to be an unassailable unlikelihood. Such a UI would have outside solutions to achieving cohesiveness of consciousness. To be plain about it, it would indistinguishable, at least for those lower than it, from the definition of God (God: a being above which no greater being can exist, for if one did, it would be defined as God). While certain SL0's may have a problem going there, it is a legitimately unassailable unlikelihood that a Level Four UI wouldn't consider such an entity to be worthy of contemplation: a UI without impairment on any space-time vector; without temporal-ratio impairment; without blindspots in time or space. A UI embodying a native solution to the self-awareness vs. internalized reality paradox. A UI capable of writing not only its own code, but pre-coding everything we think of as reality (and probably much more) to meet its own requirements.
Is there a Shock Level Six? Is there a Seven? The answer is twofold, and mundane. First, the answer is no. We SLO's left the imaginable behind at Level Four. For us, there is no Level Five unless a UI at Shock Five wants there to be. And if so, there is. And so on with higher levels- all of which belong to the definition of "God," and which become inevitable to that Level Five UI, but still entirely unknowable to us, for such things are not on the continuum of immortality, technology, transhumanism, or posthumanism. They are on the continuum of creation- of time, space, mass, and energy. They are on the continuum of intractible mysteries spanning the edge of even our theoretical universe, and deep into a place where no mind can- however grand- can imagine. If we take seriously the possibility of Level Zero leading to One, and One to Two, and Two to Three, and Three to Four, and from Four, the serious contemplation of Five, then Level Six is also a recursion. Knowing what we imagine-we-know about every other Shock Level, we can suppose that Level Six is actually no different from Level Zero- except that, this trip around, we are not alone and we never were. Level Six is relational. It's about a deep strange future, designed by ANOTHER, but into which we may travel.
Humans are not smarter than humans. They tend to fall prey to the attribution fallacy even when they have the best intentions. What I should have said was "We tend to fall prey" but I was actually falling prey to that very fallacy. Our personal reasons for believing what we believe are always too difficult for others to understand, so we tend to ascribe less-developed reasons to others. Atheists call deists self-deceiving idiots. Deists call atheists self-worshiping materialists. But we're all at Shock Level Zero when it comes to what is actually going to happen next. All parties rely on faith whenever they make assertions about the shape of the future. And that's not to say that faith is also a fallacy. This is already a strange universe, and faith may be the strangest thing we can "know" about it.
Many people believe in probability- that there is a 50-50 chance that a flipped coin will come up heads. But there's actually no such thing as probability on the level of actual, individual events. If, on a particular flip, it does come up heads, then it means that heads was a 100% certainty even before we knew it to be so; tails was 0%. Probability is just a sophisticated way of managing ignorance about the future, or sometimes, the past. And faith works the same way- except that, if there is a Level Five, Six, Seven, etc UI, then for it, knowing how the coins will land is a matter not of simulation, calculation, or even pre-destination. It's a matter, potentially, of anything it chooses it to be, including active decision. In other words, who's Shock Level Zero we're in is a matter of faith, nothing else. And faith is of the same substance as decision, as intelligence interacting with reality. It can only act so wrong before the truth imposes itself. And we don't choose whether or not we'll fill the gaps between knowable things with beliefs-pertaining-to-meaning. We just do it. That much is human nature.
So if you know- by whatever means, whatever reason, whatever bridge your intelligence can build- that what we imagine now as possible is not all there is to be imagined- there's such a thing as Future Shock Level Five, Six, Seven, or beyond- and in this reality- in this already-surpassingly strange universe- then I'm in no position to contradict you.
Just in case you didn't, here's an excerpt (begin quote):
Future Shock Levels:
SL0: The legendary average person is comfortable with modern technology - not so much the frontiers of modern technology, but the technology used in everyday life. Most people, TV anchors, journalists, politicians.
SL1: Virtual reality, living to be a hundred, "The Road Ahead", "To Renew America", "Future Shock", the frontiers of modern technology as seen by Wired magazine. Scientists, novelty-seekers, early-adopters, programmers, technophiles.
SL2: Medical immortality, interplanetary exploration, major genetic engineering, and new ("alien") cultures. The average SF fan.
SL3: Nanotechnology, human-equivalent AI, minor intelligence enhancement, uploading, total body revision, intergalactic exploration. Extropians and transhumanists.
SL4: The Singularity, Jupiter Brains, Powers, complete mental revision, ultraintelligence, posthumanity, Alpha-Point computing, Apotheosis, the total evaporation of "life as we know it." Singularitarians and not much else.
If there's a Shock Level Five, I'm not sure I want to know about it!
(end quote)
Catch that?
"If there's a shock level five I'm not sure I want to know about it."
If.
Which is to say, "there might not be." That it's possible, on a conceptual level, that nothing beyond what we can presently imagine will ever be imaginable. That there's not necessarily anything beyond "ultraintelligence."
Now remember, we're talking about levels of Future Shock. Shock is an very human concept: the feeling awe in the face of the ramifications of future possibilities. We're not talking about the future directly. We're not making statements about inevitability. Future Shock is a symptom of no-holds-barred extrapolation that are, nevertheless, firmly rooted in our present existence. As much as some people would like to pretend otherwise, there is no way to avoid experiencing future shock in the face of the future. Isaac Asimov was shocked by the emergence of personal computers and the internet. And yet, he wrote about godlike AIs. Arthur C. Clarke helped to pioneer the digital age. He actually invented the idea of the communication satallite. And yet, even he was shocked at the one-way cultural implications of what emerged. Shock occurs in countless ways. Microsoft's Encarta- written by paid professionals- was a massive failure compared to the shocking emergence of the volunteer-driven Wikipedia. Future shock is unavoidable. The only person immune to future shock is a dead person. It would take a knowledgeable intelligence of a higher Shock Level to not be shocked by its own level.
To question the possibility of a Shock Level Five is an honest question that, in some ways, is far more serious than what I'm about to undertake.
I'm going to ask what would a Shock Level Four ultraintelligence think about this question? We exist, with our slow, natural, organic intelligence, at SL0. Zero shock. That means that all levels of future shock are beyond us. But what would happen to the question if we were to find ourselves at SL1? Would we still consider SL4 to be the point-beyond-which nothing is imaginable?
And if a posthuman ultraintelligence (UI) with unlimited computing ability were to pose the question, how many levels of shock would there necessarily be?
What I can say for sure- since what we're talking about is shock levels- is that there would be at least one more level. It would be characterized by that which only a SL4 UI could imagine to be "unimaginable."
Before I continue, I need to diverge for a few paragraphs. I need to ask whether intelligence- the way we define intelligence- is actually scalable beyond a certain point. Is there a point at which a higher IQ just gets you the same answer that a lower IQ could have gotten, only faster? Are ultraintelligent minds able to think in transcendent ways? We know that the human mind is transcendent compared to computers- computers are our inventions. It's conceivable that a single human being, given sufficient time, freedom from senility, and opportunities-of-necessity, could invent everything that has been invented by individuals. Indeed, many inventions arrive from prolific individuals. Are UI's likely to be much better than a great number of normally-intelligent minds operating at a higher speed?
This is not a Penrosian critique of the possibility of AI, but I'm asking: what if the only kind of ultraintelligence that's possible is incapable of being self-aware because the internal vibrancy of thought causes external reality to become too slow, too dark, too uninteresting to have a self concept that exists relative to it? And what if self-awareness is necessary for intelligent intention- for acting in ways we would define as intelligent?
Let me pose the question this way. This is a thought experiment, and isn't meant to be taken as a conclusive argument, even if it draws conclusions.
Suppose you're a UI studying the goings-on of all the parts of the universe that "isn't you." Everything outside what you think of as "yourself." Let's say, for the sake of defining what we mean by "ultra," that your "mind" operates a billion times faster than a normal human mind, with billions of times more bandwidth- which we'll define as the ability to think about different salient details concurrently. Bigger minds take longer to operate, but if we use light-speed post-human "neurons" in place of 25mph neurochemical signals, we can get a truly huge brain operating at fantastic speeds.
Let's say that, with just a few of your innumerable "eyes," you stare right at the surface of the sun. Because you think a billion times faster, it takes you over a hundred thousand years to observe one hour of solar evolution. You might learn an unimaginable number of things about the sun in that one hour of "objective" time. Perhaps more than what a billion scientists could learn given a billion lifetimes. But you couldn't learn more than was available to know. And you wouldn't be able to separate what a thing is from how it's experienced. For a billion-X UI, the universe holds a different meaning. Subatomic activity ceases to be either a statistical firehose or a needle-wide window. Instead, you are able to watch a million square miles in simultaneous microscopic slow motion. And if that's the way you see the universe, then that becomes the normal way of seeing things. The universe becomes a horse of a different color. And that kind of difference would lead to insights unanticipated by normal-speed minds.
Same thing if you were to operate more slowly. An ultraintelligence able to operate a billion times more slowly would have another set of advantages: the ability to perceive over two million years of solar activity per personal "day." The more natural place to put one's attention would be on the macroscale- where big things are happening constantly. Dozens of supernova a minute. Galactic gravity unfolding before your eyes in realtime. Forests changing to deserts faster than a cloud moves across the land. Again, the universe would take on an entirely different meaning derived from a different mode of perception and interaction.
Can you have both kinds of thinking? The most straightforward answer is simply, "No." For a single ultraintelligence to have a cohesive awareness, a single reference frame is, by definition, necessary-- unless we first decide to define words like "cohesive," "awareness" and "intelligence" differently. Granted, you could have numerous non-simultaneous consciousnesses running concurrently, at different speeds, with different perceptual foci, and then integrate their thought products after-the-fact (like watching memory-movies created by alternate lives). However, the fastest consciousnesses will always outweigh the slower consciousnesses because they produce more "movies." The movies produced by fast consciousness would define the UI's native realtime. The fastest consciousness is consciousness. Anything less than full-speed would have a minority share in contributing to the identity of the intelligence (proportional to the inverse of the ratio of the differences in speed). Between the two examples above is a factor of 10^18.
So, as an ultraintelligence with vast computing capacity at your disposal, the ability to write your own operating "software," and the option to operate very fast, what would you do? Pretty simple. You'd take what you learned with your slow-motion ultra-wide microscope and you'd internalize the workings of reality. You'd simulate timescales from within your primary, full-speed consciousness. Why wait billions or trillions of years for something you can easily acquire (as an incomplete but constantly improving) picture in a matter of moments?
Without trying to go into detail, let's consider: what does an ultraintelligence think about? The simplistic answer is: reality, and not just the one that it exists within, but many variations on that one, and perhaps many others that we wouldn't consider to be parts of "reality." But our perceptions are biased. Essentially: many variations operating at many speeds- each producing questions tending to be answerable. Even a UI would have trouble thinking about what it can't think about. And that's the heart of where I'm headed.
If the majority of your thoughts are about events that originate within yourself, then, for all intents and purposes, you are your reality. The concept of self-awareness, as we have heretofore defined it, becomes nonsensical. Self-awareness requires "self" to be the thing that is aware. That "self" must have something that is distinct and separate from- something incompletely known- in order to define the parameters of self. If you know everything- or consider nothing that is unknown- then all things are, in essence, within you. And if there is no distinction, there can be no awareness of distinction. Hence, the grander the UI, the less there is of "self."
Now, when we're talking about hypothetical posthuman singularity-style ultraintelligences with planet-sized minds and the ability to exploit the fine structure of reality for carrying out its computational whims, we're certainly not talking about an "infinite" intelligence. The conditions described above are a matter of degree, not of absolutes. At some degree, however, the concept of self-awareness becomes obsolete as a primary mode of description. In fact, a UI might actually choose to represent itself as some number of self-aware individuals within itself- each distinct but fully compatible with its other manifestations.
Next question: what can an ultraintelligence think about? What kind of "life of the mind" can such a being expect? It is conceivable that such an intelligence would spend an enormous amount of thought seeking meaningful answers to this very question even before it reached the level of ultraintelligence (for instance, by reading this blog). By necessity of achieving Shock-Level transcendence, ways of thinking would be devised to maintain contact with modes of being that the previous incarnation would recognize as meaningful. That's not to say that an ultraintelligence wouldn't evolve rapidly into something we'd consider absolutely alien. But, from the UI's own perspective, that evolution might be very gradual and cautious. The stakes are vast, after all. A UI faces a kind of oblivion, a loss of self by drowning in a thoughtspace where it is almost the only thing that exists; where the only things left to think about are permutations of thoughts already thought.
If a Shock Level Four UI can reach the state of UI without closing its horizons to coincide with the divide between it and outer reality, then what it is contemplating is, necessarily, what we'd have to consider to be the subject material of Future Shock Level Five.
Something beyond itself, and perhaps always beyond itself. If any intelligence can contemplate something beyond imagining, it would be- by analogy with levels one-through-four, a far greater UI. And the thing beyond imagining would, by necessary exclusion, be a mode of intelligence beyond its own ability to attain because of certain impairments to its candidacy for Level Five.
One of the ways an ultraintelligence of any amount of power is irrevocably impaired is by the fact that it has evolved from a lower starting place- a place from which its primitive memories are still available. A hypothetical SL4 UI knows, and cannot help but know, the full story of its existence. Therefore, it knows- if only by analogy- that its existence is (hopefully) one of ever-increasing vibrancy in the forward temporal direction, and ever-decreasing vibrancy in the negative (back in time). Impairments in time equate to impairments in space- you can't see beyond your lightcone. And time travel, if it were possible at all, would make causality, and therefore, "knowability" meaningless. It would destroy the timeline that gave it rise and would, therefore, happen a maximum of once per universe. Therefore, even if a UI was capable of time travel, it couldn't affect the limitations of its personal past.
Any UI capable of knowing itself at all would know that it is impaired on certain space-time vectors and that the absence of these impairments would lead to a horse of a different existence- an alien sort of intelligent identity. The UI would, therefore, consider the ever-transcendent Shock Level Five intelligence to be one without any impairment vectors. One with all the benefits of time travel without collapsing causality. One that can consider variables and alternatives within itself without losing, or even damaging the concept of "self."
At Shock Level Zero, many intellectually rigorous individuals tend to consider the idea of a Level 5 UI to be an unassailable unlikelihood. Such a UI would have outside solutions to achieving cohesiveness of consciousness. To be plain about it, it would indistinguishable, at least for those lower than it, from the definition of God (God: a being above which no greater being can exist, for if one did, it would be defined as God). While certain SL0's may have a problem going there, it is a legitimately unassailable unlikelihood that a Level Four UI wouldn't consider such an entity to be worthy of contemplation: a UI without impairment on any space-time vector; without temporal-ratio impairment; without blindspots in time or space. A UI embodying a native solution to the self-awareness vs. internalized reality paradox. A UI capable of writing not only its own code, but pre-coding everything we think of as reality (and probably much more) to meet its own requirements.
Is there a Shock Level Six? Is there a Seven? The answer is twofold, and mundane. First, the answer is no. We SLO's left the imaginable behind at Level Four. For us, there is no Level Five unless a UI at Shock Five wants there to be. And if so, there is. And so on with higher levels- all of which belong to the definition of "God," and which become inevitable to that Level Five UI, but still entirely unknowable to us, for such things are not on the continuum of immortality, technology, transhumanism, or posthumanism. They are on the continuum of creation- of time, space, mass, and energy. They are on the continuum of intractible mysteries spanning the edge of even our theoretical universe, and deep into a place where no mind can- however grand- can imagine. If we take seriously the possibility of Level Zero leading to One, and One to Two, and Two to Three, and Three to Four, and from Four, the serious contemplation of Five, then Level Six is also a recursion. Knowing what we imagine-we-know about every other Shock Level, we can suppose that Level Six is actually no different from Level Zero- except that, this trip around, we are not alone and we never were. Level Six is relational. It's about a deep strange future, designed by ANOTHER, but into which we may travel.
Humans are not smarter than humans. They tend to fall prey to the attribution fallacy even when they have the best intentions. What I should have said was "We tend to fall prey" but I was actually falling prey to that very fallacy. Our personal reasons for believing what we believe are always too difficult for others to understand, so we tend to ascribe less-developed reasons to others. Atheists call deists self-deceiving idiots. Deists call atheists self-worshiping materialists. But we're all at Shock Level Zero when it comes to what is actually going to happen next. All parties rely on faith whenever they make assertions about the shape of the future. And that's not to say that faith is also a fallacy. This is already a strange universe, and faith may be the strangest thing we can "know" about it.
Many people believe in probability- that there is a 50-50 chance that a flipped coin will come up heads. But there's actually no such thing as probability on the level of actual, individual events. If, on a particular flip, it does come up heads, then it means that heads was a 100% certainty even before we knew it to be so; tails was 0%. Probability is just a sophisticated way of managing ignorance about the future, or sometimes, the past. And faith works the same way- except that, if there is a Level Five, Six, Seven, etc UI, then for it, knowing how the coins will land is a matter not of simulation, calculation, or even pre-destination. It's a matter, potentially, of anything it chooses it to be, including active decision. In other words, who's Shock Level Zero we're in is a matter of faith, nothing else. And faith is of the same substance as decision, as intelligence interacting with reality. It can only act so wrong before the truth imposes itself. And we don't choose whether or not we'll fill the gaps between knowable things with beliefs-pertaining-to-meaning. We just do it. That much is human nature.
So if you know- by whatever means, whatever reason, whatever bridge your intelligence can build- that what we imagine now as possible is not all there is to be imagined- there's such a thing as Future Shock Level Five, Six, Seven, or beyond- and in this reality- in this already-surpassingly strange universe- then I'm in no position to contradict you.
Wednesday, February 10, 2010
Inflatable Soldiers
I saw the Hurt Locker recently and the most riveting scene, in my opinion, was the sniper battle in which Voldemort was killed by an Iraqi insurgent operating a Dragunov from a small concrete shack off up the hill. Sanborn (Anthony Mackie) takes over his abandoned M107 and takes off his helmet. Granted, if he gets hit in the head, helmet doesn't help. But his head is actually more visible without the camouflage the helmet provided. William James (Jeremy Renner) takes over as his spotter and leaves his helmet on. That, the spit-and-polish lesson, and the juice box, were all significant imo.
((Using stats collected up through 2007, I'm estimating that Improvised Explosive Devices (IEDs, aka roadside bombs) cause around 40% of U.S. fatalities in Iraq. The current number may be higher. Obviously, anything that can be done to address IEDs directly will save lives. My best idea is to create a robotic conversion kit for a Humvee allowing it to be remotely controlled by a local operator. The Humvee would carry an array of pulsed magnetic detection equipment that would identify metalic objects near the road being traveled. A camera on a boom would allow a higher vantage point for visually identifying suspicious metallic mass. Anything suspicious would be fired upon using anti-materiel rifles by U.S. troops operating from a safe remove. The remote Humvee could also "toss" detonation charges before retreating or continuing down the road.
Another idea would be to create a series of quickly deployable surveillance poles. Video data would be delivered via wireless data connection to local hubs. Whenever a convoy is about to travel down a specific length of road, the data would be remotely reviewed (by non-military subcontractors). Poles would be equipped with night vision. Software would extract footage featuring motion. Traffic that doesn't stop or significantly vary its speed would be detected by matching it to a template. Any tampering with surveillance equipment would result in immediate live attentiveness and possibly fast-moving (for instance, a helicopter gunship) response.
But this is an idea specifically for the kind of scenario featured in the sniper battle in "The Hurt Locker."))
As soon as U.S. troops come under fire from whence they know not, they seek cover. At the same time, inflatable soldiers are (somehow) deployed- carefully crafted inflatable decoys that self-inflate under the cover of a temporary smoke screen using compressed CO2. These inflatable soldiers would move slightly using a combination of robotic actuators, internal puppet marionette strings, and swiveling. To be effective, they would need to swivel to face the enemy. Also, they'd appear in a crouched or semi-crouched firing position holding an inflatable rifle.
Now, the sniper(s) firing on the friendly position aren't going to be fooled by the decoys within a certain range. However, beyond a certain range, the enemy will have no choice but to second guess whether the inflatable soldiers are the real thing. Meanwhile, actual soldiers are prone, moving to a more protected position, or simply crouched in the open right among the decoys. Every time the enemy fires on an inflatable soldier instead of the real thing, friendly forces have an increased chance of surviving the encounter.
Inflatable soldiers would weight about ten pounds each and would look like landmines in their undeployed state. They'd be tossed around like Frisbees or launched like clay pigeons. The first stage of their inflation process would turn them upright if they happened to land upside down. They'd be reusable- just open a valve and return to manufacturer to be repacked. They'd be usable in both urban and open conditions. They could be deployed remotely using a radio-controlled vehicle using off-the-shelf RC parts. Their directional microphone and swiveling ability would provide a cue to actual soldiers which direction the enemy fire was coming from. By putting some sort of fine powder inside the inflatable, a penetrating bullet could indicate the direction of origin by creating a plume.
Using modular inflation chambers, it might be possible to make an inflatable soldier stand up to being shot. Sequenced inflation might also make it possible to raise the newly inflated soldier in a way that crudely mimics the natural motion of rising from prone to a crouched position.
The appearance of more soldiers than the sniper originally anticipated might also give cause to break off the attack.
Inflatable soldiers could be used in lots of tactically creative ways. For instance, a small force could initiate an attack on a fixed position by inflating an entire platoon of inflatable soldiers away from their point of attack as a distraction. A single sniper could deploy a half-dozen inflatable soldiers around, but not-too-close-to his own position. After each shot, he could then inflate another soldier as a distraction. The enemy would have no choice but to dedicate some of its attention to the inflatable, even knowing that it was a decoy.
An inflatable soldier would be much easier to deploy in a rapid, believable fashion than an inflatable tank for instance. The organic curves of a human body would also be a more natural shape to attempt using small-scale inflatable architecture.
((Using stats collected up through 2007, I'm estimating that Improvised Explosive Devices (IEDs, aka roadside bombs) cause around 40% of U.S. fatalities in Iraq. The current number may be higher. Obviously, anything that can be done to address IEDs directly will save lives. My best idea is to create a robotic conversion kit for a Humvee allowing it to be remotely controlled by a local operator. The Humvee would carry an array of pulsed magnetic detection equipment that would identify metalic objects near the road being traveled. A camera on a boom would allow a higher vantage point for visually identifying suspicious metallic mass. Anything suspicious would be fired upon using anti-materiel rifles by U.S. troops operating from a safe remove. The remote Humvee could also "toss" detonation charges before retreating or continuing down the road.
Another idea would be to create a series of quickly deployable surveillance poles. Video data would be delivered via wireless data connection to local hubs. Whenever a convoy is about to travel down a specific length of road, the data would be remotely reviewed (by non-military subcontractors). Poles would be equipped with night vision. Software would extract footage featuring motion. Traffic that doesn't stop or significantly vary its speed would be detected by matching it to a template. Any tampering with surveillance equipment would result in immediate live attentiveness and possibly fast-moving (for instance, a helicopter gunship) response.
But this is an idea specifically for the kind of scenario featured in the sniper battle in "The Hurt Locker."))
As soon as U.S. troops come under fire from whence they know not, they seek cover. At the same time, inflatable soldiers are (somehow) deployed- carefully crafted inflatable decoys that self-inflate under the cover of a temporary smoke screen using compressed CO2. These inflatable soldiers would move slightly using a combination of robotic actuators, internal puppet marionette strings, and swiveling. To be effective, they would need to swivel to face the enemy. Also, they'd appear in a crouched or semi-crouched firing position holding an inflatable rifle.
Now, the sniper(s) firing on the friendly position aren't going to be fooled by the decoys within a certain range. However, beyond a certain range, the enemy will have no choice but to second guess whether the inflatable soldiers are the real thing. Meanwhile, actual soldiers are prone, moving to a more protected position, or simply crouched in the open right among the decoys. Every time the enemy fires on an inflatable soldier instead of the real thing, friendly forces have an increased chance of surviving the encounter.
Inflatable soldiers would weight about ten pounds each and would look like landmines in their undeployed state. They'd be tossed around like Frisbees or launched like clay pigeons. The first stage of their inflation process would turn them upright if they happened to land upside down. They'd be reusable- just open a valve and return to manufacturer to be repacked. They'd be usable in both urban and open conditions. They could be deployed remotely using a radio-controlled vehicle using off-the-shelf RC parts. Their directional microphone and swiveling ability would provide a cue to actual soldiers which direction the enemy fire was coming from. By putting some sort of fine powder inside the inflatable, a penetrating bullet could indicate the direction of origin by creating a plume.
Using modular inflation chambers, it might be possible to make an inflatable soldier stand up to being shot. Sequenced inflation might also make it possible to raise the newly inflated soldier in a way that crudely mimics the natural motion of rising from prone to a crouched position.
The appearance of more soldiers than the sniper originally anticipated might also give cause to break off the attack.
Inflatable soldiers could be used in lots of tactically creative ways. For instance, a small force could initiate an attack on a fixed position by inflating an entire platoon of inflatable soldiers away from their point of attack as a distraction. A single sniper could deploy a half-dozen inflatable soldiers around, but not-too-close-to his own position. After each shot, he could then inflate another soldier as a distraction. The enemy would have no choice but to dedicate some of its attention to the inflatable, even knowing that it was a decoy.
An inflatable soldier would be much easier to deploy in a rapid, believable fashion than an inflatable tank for instance. The organic curves of a human body would also be a more natural shape to attempt using small-scale inflatable architecture.
Subscribe to:
Posts (Atom)