No, that’s the explicit starting point of t’Hooft; he assumes that at the bottom, physics is described by a cellular automaton.
The way I see it is that he proposes a definition for creativity, produces a hypothesis for its origin, and develops a theory of it based on that—which is, of course, ultimately subject to falsification (say, by creating an agent implementing it, and seeing whether he is capable of the same feats of creativity as we are—something which has already been done with some surprising successes in the fields of computational music, art, and, yes, science), upon which then a new theory has to be generated. Your approach just seems to be to assume that this won’t work from the start, as something unknown might be doing something we don’t/can’t understand.
What Schmidhuber describes is a process of learning aided by creativity; in this sense, creativity is a by-product of learning, but that doesn’t mean that creativity can’t become its own end. The idea is simply to propose a plausible origin for creativity, from which its ‘large scale’-application may be derived; if that turns out not to work this way (as it well may), then another approach will have to be tried.
That’s not really what I’m saying at all. Rather, I suggest following the usual falsificationist methodology: propose a hypothesis based on current knowledge, and extrapolate from that; you seem to be saying that current knowledge may be incomplete, and hence, that’s useless. Which may well be the case, but based on that attitude, no theory-building is possible.
Those are the ones one should be most suspicious about, I’ve found.
Well, think about the possibility of all of us living in the matrix: certainly, it’s possible to have different NPCs in a computer game react differently to the same stimuli, while nevertheless, the behaviour of the computer remains fully describable by known physics (even if such description is practically infeasible). So this means we have evidence for the claim that this description is possible, but none against it (we have certainly never met with a case where we had to just pack our things and leave, epistemically speaking), and I find it difficult to even conceive of such a case.
That also used to be the case for the distinction between the sub- and supra-lunar spheres, electric and magnetic phenomena, motion and spacetime, etc. etc.—science tends to unify what was previously thought to be distinct phenomena, and doing so is in fact its greatest strength. It’s not that physics is blind towards the phenomenon of life, it’s that living matter is perfectly well described by the same physical laws as nonliving matter is, which just speaks to the generality of those laws. The world does not decompose into non-overlapping magisteria, but is instead a unified whole. The notion of elan vital to set apart life from non-life has proven to be as superfluous as the notion of phlogiston or caloric fluid: life refers to a way of arranging matter, not to something that has to be added to it. Frankly, I don’t think that the idea of vitalism you seem to propose really has any adherents anywhere today.
There are many physical correlates of life that one can detect—one I’ve mentioned you seem to have overlooked is the capacity of locally decreasing entropy, thus keeping homeostasis.
Not explicitly, no. But since understanding is a series of steps of reasoning that logically follow one another, and no such steps exist for the non-computable, then it’s pretty much exactly what you’re saying: I could never explain to you the reason for something that happened in a non-computable manner (as it wouldn’t be expressible in a finite number of symbols), nor could I ever recapitulate the steps leading to it myself; I could perhaps ‘just know’ that it is right, doing something akin to asking an oracle within myself, but that would not clear up any mysteries.
Do you mean between living and dead, or between living things and non-living things?
It’s all matter and energy in any case. I don’t believe physics makes any distinction at all between a molecule in a rock and a similar molecule in a person.
Living things are systems - well, I guess anything is - but living things are systems that display certain behaviours such as acting to sustain their own activity, self-replication, etc.
We can, but I don’t think it’s strictly a physics question.
It comes from the notion of cause and effect - that effects have an explanation. Obviously in the case of a human organism, we’re talking about a horrifically complex web of causes and effects - so complex that it defies our current capacity to measure and describe it.
But the alternative to cause and effect is random un-caused-ness. How does that help?
He was a BC cocker mix. Then we started raising guide dogs, and BCs are way too smart to be guide dogs. They’d be bored.
Oh, I think he was well beyond automated responses. Think about how your subconscious works. No sentience down there, but clearly not simple automation.
Again it depends on your definition of sentience. Self-awareness would seem to be a fundamental part of it. Under that definition neither are sentient, but I’d agree that dogs are more intelligent, as we measure it anyway.
But the blueprints for the structures that enable sentience (whatever they are) are genetic, just like the capability of speech - something else little babies don’t have. And I agree with you about very little babies not being sentient. It takes them a while to even find their toes.
I totally agree. If we get the modeling and the interconnect right we should be able to reproduce our brain - though we’d be reproducing non-sentient ones first.
If we could read a working brain well enough to do this, we might be able to tell the difference. Think of the shock if we find that, just like our DNA is very close to chimps, our brains are also. I wonder how integrated our sentience is - from the perspective of someone who has built simulators and knows that it would be more efficient to do really detailed simulation of the sentient part and higher level simulation of the other parts.
Maybe we are non-sentient beings in the main with a small but distributed sentience module added on quite late in evolutionary history.
Exactly. I first encountered this idea in “1266: A Distant Mirror,” a history set in the Dark Ages, in which she speculates that the reason for the incredible cruelty people displayed to one another in those times began with parents having to think of their own children as not-quite-human until they got out of the age where infant and child mortality was likely. As I recall she said that many families did not even name their children until they turned six.
It’s a profound insight, not one that guys nowadays are likely to see unaided.
How did it take the dinosaurs to achieve full flight? How long did it take us?
100 billion used to be a big number, but I generated a terabyte worth of data in the past year without breathing hard. As for why, there is a Nobel Prize and fame waiting for anyone who can make this happen. Not to mention the desire to understand.
We must be talking past each other, then, because that’s just not what I’ve been saying at all.
Ditto vitalism. I looked it up, and frankly, I’m not an adherent either. I don’t suggest that life is a mysterious add-on, let alone a force of some kind operating outside the normal bounds of science. To the contrary, actually, I’ve been surprised to learn that physics can’t describe the pertinent “way of arranging matter.” I think being able to categorize life-based attributes would be helpful when it comes to both analyzing the operators in human intel/behavior and sorting out the necessary/unnecessary attributes of an artificial intelligence. Perhaps it would have pressed fewer hot buttons, if I had stuck to terms like organism vs non-organism, instead of life.
I’m not sure how you end up imputing irreducible mystery to me thus, when you, yourself, stated that, “even in a computable world, some questions cannot be computationally decided, but only through exploration of the possibilities.” Maybe it depends on how we define possible. One of the first things you learn in creative problem solving is that an idea which seems unworkable on its face is often the very novelty that triggers an otherwise unlikely solution. Nothing stalls out the process faster than narrowing down the possibilities too soon. Some people find the necessary suspension of judgment involved more difficult than others, whether working groups or on their own.
That is a logically problematic statement.
I’m running out of ways to say that’s not the case. I am questioning Schmidhuber’s premise and talking about circularity.
IF we define creativity as an outgrowth of the intrinsic desire to build a better predictive model of world, and develop an algorithm with which to equip an agent who subsequently produces a better model, then we have accurately defined creativity insofar as it pertains to building better models. To put it another way, creativity, when directed at building better models, will build better models. Ironically, if the algorithmic agent couldn’t produce anything of merit, we would say Schmidhuber was wrong, even though creative thinking does, in fact, often fail to produce any desirable results. Unfortunately, I see no practical way to factor that in without destroying the metrics for falsification, even if it has a lot of ramifications.
The main thing I take issue with is Schmidhuber’s premise in re the idea that he is addressing the universal origins of creativity, rather than the application of creativity to a targeted task. As “a hypothesis based on current knowledge” I believe Schmidhuber’s theory is far more narrowly tailored than you apparently do.
I’m a successful artist, and I’m telling you that my creative work and the intrinsic incentives and rewards which generate it do not derive from a desire to build a better predictive model of the world, or produce one. What to do? We can stretch the meaning of “predictive” and “model” and “world” beyond recognition to lever me into Schmidhuber’s formulation semantically. We can assume that the intrinsic incentives and rewards of my work are actually opaque to me, not Schmidhuber. We can agree that I don’t fall within the parameters of Schmidhuber’s premise. We can say that if I don’t turn in my falsification papers, I’m a scientific nihilist.
But if it helps, you can just think of me as a creative, novelty seeking thinker who will eventually contribute to a better predictive model of the universe, and we can leave it at that.
Might it be valid to say that your artistic creativity derives from a desire to create a more communicative depiction of the world? Many artists are trying to convey a sense, a mood, a feel: they are seeking to create a channel by which to communicate things that are shared in common by most people, but which are difficult to put into words. You’re still trying to “make a better model” – you want to paint a portrait, say, of a clown that will actually make people feel like laughing.
Living vs non-living things. Oddly enough, I actually did start out pondering the ramifications of a human brain’s co-dependencies in an integrated organism, in terms of how that might affect simulating a brain in isolation. And here we are.
I think I stirred up the hornet’s nest when I suggested that there might a level of complexity which exceeded our theoretical ability to comprehend it. I probably won’t do that again.
Does true randomness exist as a concept in physics? Would I be right in assuming that it doesn’t?
I would have to say no. It would take an enormous amount of twisting & turning & fudging definitions to come up with a description that meets the “desire to creative a better predictive model” standard, even if you leave out “predictive.” There’s really no way to do that without crossing over the descriptive line into abstraction and metaphor – and the more you do, the harder things become to quantify or extrapolate from in any meaningful or useful way.
I suspect that’s already problematic, whenever you’re trying to translate conceptual language into quantitive/mathematically symbolic form, before you add semantic somersaults to the mix.
I’d be content with creative novelty seeking thinker who makes stuff. I just threw the last part in for the brain simulation.
Then tell me how I should read your arguments that attempt to refute the extrapolation of current knowledge via reference to the possibility of unknown things?
Well, if you tell me exactly what life is, then of course one can in principle come up with physical characteristics shared by living systems—simply by virtue of them being distinguishable from non-living ones. But that’s more difficult than you might think: is, for instance, a virus alive? A self-replicating strand of RNA?
Our concept of life is an intrinsically fuzzy one; thus, expecting physics to come up with a hard-and-fast delineation puts the cart before the horse. Besides, I have (twice now, at least) proposed a characteristic that seems to be more or less universal to life: the local reduction of entropy in order to stem off its decaying effects. But not all of what we would consider ‘life’ might fit within that envelope.
Again, it’s like the distinction between sublunar and supralunar physics; there is no operational way to distinguish between both in modern physics simply because there is no hard-and-fast distinction between both: the planets follow the same laws as rocks thrown on Earth.
One such question (the most famous one) is the halting problem: you can’t computationally decide whether or not a given program will halt. But of course, for any halting program, you will eventually be able to find out whether it does, in fact, halt.
The process of finding this out has important parallels with the creative process in science, by the way: you can’t ever verify the thesis that a program does not halt, you can only falsify it. We ‘chip away’ the wrong, and accept what remains—at least provisionally—as true, until it likewise becomes chipped away. This illustrates the futility of arguing from the unknown: we simply can’t say what will be chipped away before it does, and thus, can’t base any conclusions on the possibility of any given item of knowledge being so chipped away.
Enlighten me? The idea is that you need creativity to produce novel patterns, which (since there is no computable method to do so) is the only way to achieve a complexity reduction for a given task. Of course, creativity has since become co-opted by our social and cultural evolution towards other ends, but this is—according to Schmidhuber—where our original drive for and pleasure in creative pursuits stems from. This is a common process, also in biological evolution: faculties developed in response to certain stimuli are adopted towards other problems.
For instance, it’s easy to explain the fun in games using this theory: game-playing is a sort of ‘safe’ exploration of a ‘toy world’, it allows an agent to test out and develop its faculties—physical as well as mental—in a relatively controlled environment, and adapt the problem-solving strategies thus earned to real-world problems at some point. And of course, art has often be analyzed as a ‘game people play’, in transactional psychology for instance.
That doesn’t mean that this is the reason why you create art, just as babys don’t play with the express intention of training their abilities to recognize and interact with the objects in the world. But it’s at least possible to create a narrative starting with the adaptive learning agent in the world utilizing creativity to attack novel problems and ending with the art-producing modern man, who might take himself to have entirely different, more noble and not so reducible motives.
But that’s not at all what Schmidhuber does. Certainly, he defines creativity a certain way—though I would not agree with your paraphrase—, but then he proposes a method by which this creativity could lead towards the multiplicity of creative expression we see humans expressing. This leads to a theory that basically states, if we take a community of these learning agents in the world, then it is likely that they will produce expressions of creativity similar to those we propose; if this is not the case, then his theory will be falsified. There’s nothing circular about it; it’s positing a hypothesis and examining its consequences.
He does not discuss an application of creativity, but how creativity originates in accomplishing a particular task. The noncomputability of finding the optimal strategy in a given situation means that you have to go a ‘generate novel patterns and see if they help’-strategy. A very simple way to implement such a strategy is via genetic algorithms: you start using a trial solution, and vary that solution along several lines; you then keep those solutions that accomplish the task better than its predecessor, and vary them again, weeding out those performing worse, and so on. That way, you hit upon novel solutions that you could not simply have derived from the problem statement. Add in some simple reward mechanism, and you have an agent driven towards tinkering with novel approaches to problems, with an intrinsic appreciation of their novelty, one that will come up with new ideas that don’t follow—in the sense of logical implication—from previous ones. And I think at least for starters, that’s a very passable definition of creativity. Whether it’ll ultimately do, of course, can only be seen by experiment.
Well, it’s an often-heard canard, but you must forgive me if I say that this sounds exactly like the special pleading that humanity has always been want to engage in whenever science threatened to remove us a little further from our ‘special place’ at the center of creation, whether literally (as in non-geocentric cosmological models) or figuratively (as in evolution). In fact, it sounds particularly close to the latter: where once people pointed out all of mankind’s great achievements in order to show the sheer absurdity of considering them just another ape, you point to similar things in order to establish that your creativity certainly can’t be explained by something as simple as a computer program optimizing some evaluation function.
And you may well be right, of course. But this is where we must wait and see: there’s no way to predict whether the theory will be able to account for all aspects of human creativity. But your dismissal of it is akin to the dismissal of evolution because humans wear clothes, and apes don’t, so how can we be apes? (I don’t want to paint you into the ‘evolution denier’ corner here, by the way. I just think your logic is flawed in a fundamentally similar way, and it’s the easiest example to come by.)
As for randomness (you asked in a different post), the majority of modern physicists would argue that there is, in fact, genuine randomness in physics, given by measurement occurrence in quantum mechanics. Some would say that this randomnes is ultimately only a function of the ‘epistemic corner’ of the world in which we happen to be, e.g. in the many-worlds interpretation, the underlying evolution is fully deterministic, but we have only access to a part of it, leading to measurement outcomes that are at least according to their distribution fully random (as opposed to pseudo-random). A minority, those holding to some form of hidden-variable theory, as e.g. Bohmian mechanics, would say that we are just barred from acquiring full knowledge about the true dynamical degrees of freedom of the theory, hence the randomness being only apparent.
Actually, I may withdraw the question and pose it differently - what’s the difference between a living person and a dead person? How is this difference distinct (except in complexity) from the difference between a working clock and a stopped clock? Or between a person and a rock?
I was under the impression that it did - that certain kinds of quantum events are truly uncaused and random (and that there’s no way to fit them to a notion of hidden causes).
Is there such a thing as an uncaused cause, or else if not, how can it be turtles all the way down?
In any case, whether or not there are random events or whether every effect has a preceding cause does not help with the notion of why people have minds.
-Universal cause-effect makes us orderly machines.
-Uncaused random causes just means there is some noise in the signal.
Unless the uncaused causes of quantum events is somehow the expression of fundamental spiritual volition, or something - but I don’t wish to put words in your mouth.
I quit expecting physics to delineate life a couple of posts ago. I still believe an operational definition would be helpful in re AI, for reasons I have already described, but unless you’re saying that none of the assorted sciences could provide that, then I don’t see the problem. Oddly enough, in editing an earlier post for length, I scratched out “Maybe decreasing entropy and homeostasis will save the day.” I’m not sure what you were expecting, when you’re not exactly piling eggs in that basket, yourself.
Gosh, why wouldn’t I forgive an assumption fueled characterization so infused with disparaging comparison that it couldn’t possibly apply to me? I could tell you that I don’t see science threatening humanity’s special little place at the center of creation, or my own (I vacated that spot when my kids were born), and that I’d be interested to see what a theory that encompasses the operative incentives and rewards which drive my own creative work might look like, but somebody apparently stamped “closet creationist” on my forehead. At this point, I’ll just suggest that perhaps you could keep your knee from jerking long enough to contemplate the possibility that I might not, in fact, fit within Schmidhuber’s parameters, and that the reason I might consider that problematic is because I don’t think the experience I represent is unique.
I’m not sure I get what you’re saying anymore. At first, you seemed to be arguing that the absence of a categorical distinction between living and nonliving matter in physics was a failing of physics, whereas I think it’s the opposite, just like the absence of a categorical distinction between sub- and supralunar physics, and the general unifying rather than fractionalizing tendencies of physics. But now, I’m honestly not sure anymore what exactly it is you are arguing for.
I haven’t characterized you (merely described how your statements come across), and in the part of my post you neglected to quote I’ve taken pains to point out that I am not lumping you in with creationist ideas; but if you feel you must understand my words that way in order to drum up some faux outrage, then knock yourself out.
Regarding more to the point matters, of course I don’t know what kind of artist you are, and what sort of art you produce. But I think that it’s quite easy to imagine that a lot of art that is being produced falls under the umbrella of Schmidhuber’s theory. Take music, for instance: it’s basically all pattern—harmonies, movements, repetitions, etc. Lots of it is probably due to the incidental way our auditory system is wired up (which notes sound good together, for example), but the rest is basically a poster example for highly compressible data. Similarly for poetry: rhyme schemes, meters, poetic forms like sonnets, ballads etc., all of which constrain the available variability—and to be sure, it’s quite plausible that the original reason for this, prior to written records, was to make it more easily remembered, which is a direct consequence of its compressibility; the next line is much more easily recalled if you know that it will be in iambic pentameter and end on a word that rhymes with ‘-eart’ than if it is completely unconstrained. Abstract graphic art, to a point, follows similar themes: a reduction of content to pure color, form, hue, etc., serves to detail those universal rules that underly all of our visual experience.
Of course, all of this is just at the very base of each discipline, and has been transcended in innumerable ways; but each step of this process reflects in some way what has gone on before, even in being its diametrical opposite. Indeed, much modern art can perhaps better be categorized as being meta-art, reflecting not on our experience of the world, but on our experience of art, on the rules of what makes art art, etc. (often by explicitly breaking them). To me, this is quite an intriguing, even beautiful story, though of course I have no way of knowing whether it is true (but if I were to hazard a guess, I’d suppose that while it captures an essential element of the human drive towards novelty, it’s probably not the sole pillar on which it rests; nobody says that the explanation for anything people do has to be cut out of one piece, and I can easily imagine other reasons for producing beautiful things, such as for instance once again in analogy to the peacock’s tail as simply an adornment, superfluous in principle, whose very superfluousness signals to a possible mate that one has many resources to spare). As you say, you may be a counterexample, or someone else might: there always may be black swans. But as a theory, it’s at least worth investigating, I’d say. It’s a possibility, and in your words,
Which is exactly what the assumption of a black swan does. Even though it seems inapplicable to you, again as you say,
I looked around for examples of Schmidhuberian successes in music and art and came up empty save for “Femme Fractale,” which apparently just represents the kind of art that a computational AI supposedly could create. It does, however, appear that I wasn’t wrong in thinking that the ultimate target of Schmidhuber’s work is indeed an “algorithmic theory of everything” – which certainly reinforces my impression that even his theory of creativity was strongly shaped toward that specific end.
I also bumped into Gerald Edelman, who has been taking what sounds almost exactly like the approach I’ve been talking about, with his “brain based devices,” emphasis mine:
Having been cast as a denier, Edelman’s “Neural Darwinism” seems amusingly apropos. We may not see eye to eye, but I’m hardly bouncing off the scientific walls.
I think it’s safe to say that doesn’t just sound deliberately insulting. There’s a point at which qualifiers like “sounds like” or “seems like” stop taking the edge off being lumped in with nihilists and dualists of one sort or another at every turn.
I can easily see a computational basis for music composition, although I’m not sure how you get from cataloguing combinations of tones etc. that are likely, or more or less likely, to sound “good” and “bad,” to devising the metrics for measuring success. It would be quite unlike testing for accuracy in math.
It may seem counterintuitive, but I think the writing of poetry, despite its deceptively simple schematics and original purposes, may actually be one of the creative endeavors which is least susceptible to quantitative/computational analysis, let alone predictive modeling. The dynamic between form and content and meaning may be one of the most complex in linguistics (which is probably why there’s so much bad poetry floating about). I’m not sure how you would begin to assign a value, let alone a poetic value, to the words listed in even a single thesaurus entry (where words approach synonymity but are not, in fact, synonymous). The valuation problem is pretty foundational, but you also add a whole other dimension of complexity, when your theory has to accommodate multiple languages, which differ dramatically just in terms of the basic number of available words (twice as many in English as German) or structural plasticity (turning nouns into verbs and inventing neologisms is easy in English). I would note that these are mechanical, not psychological, concerns. As with “low complexity” art, low complexity language may prove easy enough, and we’ve come a long, long way from typing monkeys, but I suspect producing a New York Times best selling novel will be one of the last AI frontiers.
All that said, however, I don’t think any one art form or another necessarily “falls under the umbrella of Schmidhuber’s theory” just because the end product can be described computationally. We could feed AI Joe every poem ever written and let him quantify the incidence of every combo of words & positions etc. with little reason to believe that thus armed, he will produce an original poem worth reading. In that context, I should think that his computationally induced novelty seeking behaviors would probably decrease his chances of a (literally) meaningful result, let alone success. If physics cannot delineate the difference between life and non-life, is it inconceivable that the difference between successful and unsuccessful art is incalculable? Or, even if better modeling is the presumed objective, predict why a Picasso shifts the paradigm and a child’s finger painting does not? The mathematical representation of Guernica would sure be something to see, though!
Incidentally, I think I may have reversed the role of white and black in my use of the swan metaphor. AI Joe would never have made that mistake. But here’s a potentially instructive irony for you. When I write poetry, the intrinsic incentive and reward is, indeed, a better/newer/different, if not predictive, modeling of the world around me. Whether I do so successfully has yet to be tested. Per the above, I have my doubts as to whether Schmidhuber’s algorithm will ultimately produce an artificial poet, but I have no problem admitting that, in my experience, my writing would fall within his parameters. So, call me two swans in one – but not someone desperately trying to defend the mysteries of creativity from ratiocination.
I was at a Barnett Newman exhibition in New York once, when a friend, watching me walk up to one of his paintings, said, “Look at what he makes you do. You’ve practically got your nose on the canvas trying to find something interesting to look at.”
I’m not very well acquainted with Edelman’s model, but I fail to see where it is in categorical conflict with Schmidhuber’s ideas. The basic idea, as far as I know, is a sort of Darwinistic pruning of neuronal connections, favoring those producing a response better adapted to certain stimuli. This is very much akin to the genetic algorithms I mentioned earlier, and the basic structure is present in Schmidhuber’s model: novel patterns are produced as a test, and only kept if they do present a possibility to reduce the complexity the data the agent is faced with, thus enabling him to become a better predictor.
It’s a fitting response to your attempt at dragging my argument down to ad hominem levels; besides, I would not consider calling somebody either a nihilist or a dualist to be insulting, but again, to each their own. In any case, if you feel that I insulted you, then I’m sorry; it wasn’t my intention.
Well, that’s what Schmidhuber’s model aims at: a pattern is a good pattern if it reduces the overall complexity of the data.
Natural language processing in general is one of the greatest hurdles for AI, nobody is contesting that. But so far, there does not seem to be any reason to suspect that this hurdle will prove impossible to scale; to the contrary, programs attacking it improve all the time (it would have seemed preposterous to suggest that a computer could beat a human, much less the all-time champion, at Jeopardy just a few years ago). But most of the poetry produced by computers these days is pretty much rubbish; but then again, so is most poetry produced by humans, so maybe computers have closed the gap there to a greater extent than it may seem.
But I want to add that poetry is actually a good example for the role of compression in art in another way. The German word for it is Dichtung, which comes from Verdichtung, literally meaning something like ‘compaction’ or ‘concentration’; the poem is a concentrated expression of its contents. And indeed, what is interpretation if not the unpacking of this content (not that this necessarily means that interpretation only gets at ‘what the author wanted to tell us’; indeed, I think it’s itself a creative process, and that’s where most of the fun in the appreciation of art comes from)?
Maybe the wikipedia article on computational creativity may be of some use, or the introduction to this collection of computer-generated poems. The idea is not to associate any word with some static valuation, but rather to generate networks of words, linked by association. Which words are likely to appear in a similar functional role as others, which occur often together, etc.? Then, you get a sort of map regarding the associations a word may bring with itself, delineating, broadly, what it evokes. On this basis, you can generate texts not merely based on the brute meaning of words, but instead, on the reactions they are likely to produce in any putative audience (of course, I am being exceedingly broad here).
It’s not Shakespeare, but I think that it’s worth reading:
[QUOTE=Racter]
More than iron, more than lead, more than gold I need electricity.
I need it more than I need lamb or pork or lettuce or cucumber.
I need it for my dreams.
[/QUOTE]
The final line, of course, is what makes it. It’s stunningly a propos to the unusual situation of the AI in a natural world, and it certainly opens up a window to engage in interpretive activities. Personally, I think I have read worse human-produced stuff. And this example is thirty years old by now.
