As our tribe wanders through the grasslands, searching for fruit trees and prey, it happens every now and then that water pours down from the sky.

    “Why does water sometimes fall from the sky?” I ask the bearded wise man of our tribe.

    He thinks for a moment, this question having never occurred to him before, and then says, “From time to time, the sky spirits battle, and when they do, their blood drips from the sky.”

    “Where do the sky spirits come from?” I ask.

    His voice drops to a whisper. “From the before time. From the long long ago.”

    When it rains, and you don’t know why, you have several options. First, you could simply not ask why—not follow up on the question, or never think of the question in the first place. This is the Ignore command, which the bearded wise man originally selected. Second, you could try to devise some sort of explanation, the Explain command, as the bearded man did in response to your first question. Third, you could enjoy the sensation of mysteriousness—the Worship command.

    Now, as you are bound to notice from this story, each time you select Explain, the best-case scenario is that you get an explanation, such as “sky spirits.” But then this explanation itself is subject to the same dilemma—Explain, Worship, or Ignore? Each time you hit Explain, science grinds for a while, returns an explanation, and then another dialog box pops up. As good rationalists, we feel duty-bound to keep hitting Explain, but it seems like a road that has no end.

    You hit Explain for life, and get chemistry; you hit Explain for chemistry, and get atoms; you hit Explain for atoms, and get electrons and nuclei; you hit Explain for nuclei, and get quantum chromodynamics and quarks; you hit Explain for how the quarks got there, and get back the Big Bang . . .

    We can hit Explain for the Big Bang, and wait while science grinds through its process, and maybe someday it will return a perfectly good explanation. But then that will just bring up another dialog box. So, if we continue long enough, we must come to a special dialog box, a new option, an Explanation That Needs No Explanation, a place where the chain ends—and this, maybe, is the only explanation worth knowing.

    There—I just hit Worship.

    Never forget that there are many more ways to worship something than lighting candles around an altar.

    If I’d said, “Huh, that does seem paradoxical. I wonder how the apparent paradox is resolved?” then I would have hit Explain, which does sometimes take a while to produce an answer.

    And if the whole issue seems to you unimportant, or irrelevant, or if you’d rather put off thinking about it until tomorrow, than you have hit Ignore.

    Select your option wisely.

    New Comment
    88 comments, sorted by Click to highlight new comments since: Today at 7:46 PM
    Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings

    Haha, that's a pretty good analogy. Unfortunately I think most people (myself in the past included and probably even still now) by default have their mouse cursor hovering over wherever the Ignore or Worship buttons appear when such a dialog shows up. And they click it in much the same way my grandparents would click a popup that installs malware on their computer, without thinking or paying attention. Clicking the Explain button requires effort (moving your cursor to a different spot and then waiting for an explanation), and knowing that it will bring up another dialog sooner or later makes it easier for people to just press Ignore or Worship.

    I'm sorry, but why can't there simply be an infinite amount of explanations, why can't the regress just go on infinitely? (You say "must")

    Agreed but the question is still how big of an infinity? An infinity of one? An infinity of five or eight or none?
    It's part of the worship option, I would say. As he would otherwise be contradicting himself when he outlines the actual Explain option. Edit: Whoops, the next comment is by Eliezer addressing this question but I'm going to leave this here for ease of reading.

    I say "must" in the Worship option. It is irony.

    But if there is an infinite regress of causality, I should find that highly curious, and would like to hear Explained why it is allowed, and why this infinite regress exists rather than some other one.

    At some point, the answer may become "we cannot know". For example, in quantum mechanics, the uncertainty principle tells us that there is a limit to the accuracy of our measurements, and once we hit that limit, attaining more accuracy is impossible. The big bang is similar -- if time makes no sense in a singularity, perhaps we can't know what happens before that point. Maybe at some point we will find a way around these limitations, in which case it was just another instance of hitting Explain and letting science grind along, but it could be that we have already reached the ultimate limit, and no more explanations will ever come.
    That is not what the uncertainty principle says. The uncertainty principle says that you can't measure two complementary observables such as position and momentum or energy and time to arbitrary accuracy at the same time. However it does not say that you can't measure any one observable such as position or momentum to an arbitrary degree of accuracy.
    If you have a set of entangled particles, would it not be possible to measure one aspect of each particle in the set, and thus achieve a fully accurate observation?
    I'm not a physicist, and I couldn't give a technical explanation of why that won't work (although I feel like I can grasp an intuitive idea based on how the Uncertainty Principle works to begin with). However, remember the Litany of a Bright Dilettante. You're not going to spot a trivial means of bypassing a fundamental theory in a field like physics after thinking for five minutes on a blog. Incidentally, the Uncertainty Principle doesn't talk about the precision of our possible measurements, per se, but about the actual amplitude distribution for the observable. As you get arbitrarily precise along one of the pair you get arbitrarily spread out along the other, so that the second value is indeterminate even in principle.
    I didn't come up with it. It's called the EPR Paradox.
    Neat. Consider my objection retracted. Although I suspect someone with more knowledge of the material could give a better explanation.
    I'm going to read the QM sequence now. I have always been confused by descriptions of QM.
    Do you know calculus? If so, it will be very easy to explain what the uncertainty principle actually means quantitatively, which should reduce any qualitative confusion.
    I know calculus. Not enough to enjoy looking at the harmonic equation though. It's a shame I never took a class on Quantum Mechanics. Most descriptions I've heard of it, even from professors, are indistinguishable from magical thinking.
    Ok. Here's the brief sketch with slightly simplified details: In classical mechanics, "position" and "momentum" are different features, and so can be totally independent. In quantum mechanics, "position" and "momentum" are both derived from the same source (the wavefunction), and thus are dependent. In QM, reality is the wavefunction. This is a complex-valued continuous function over the spatial dimensions of the universe which integrates to a certain amount. Let's consider a universe with only one particle in it: If you want to find out something classically recognizable about that particle, you use an operator on the wavefunction. (The classical values now come with probabilities attached to them, and in realistic situations it only makes sense to ascribe probabilities to position and momentum ranges, even though energy is restricted to particular values.) For the position of the particle, this corresponds to integrating the magnitude of the wavefunction across the part of space that you're interested in. For the momentum of the particle, the operator is the derivative, which cashes out as taking its Fourier transform. The more localized a particle is in location-space, the more spread out it is in momentum-space, because the Fourier transform of something narrow is broad, and the Fourier transform of something broad is narrow. Now, what about entanglement? Let's add some more particles to our universe; now, the wavefunction is defined over three spatial dimensions per particle. In typical situations, we can factor the wavefunction of the universe into independent wavefunctions for each particle, which are then multiplied together. When particles are entangled, that means we can't factor the universe's wavefunction when it comes to the set of entangled particles- they're dependent on each other / unified in some way. This doesn't alter where position and momentum come from- they're both still the same functions of the wavefunction, with the same fundamental re
    erm...not technically run into "knowledge about each element of a system versus knowledge about the entire system" tradeoffs. CAN partially bypass "no quantum Xerox" if you have a large sample. It's the principle used in error-correction for quantum computers. Take a laser. point it at a perfect polarizer of unknown orientation, and fire a pulse. Send the photons that get through one by one through filters of known orientation as you hone in convergently (hehe) on the orientation of the first polarizer. There is a tiny chance that you won't have a remotely correct value, and you never get exact with a finite sample, but you can probably do well enough to satisfy the typical engineer with only a "couple hundred" photons.
    I don't believe that for a second though. Everything we know is likely as wrong a phlogiston, though our predictions are surely getting more accuate. "We cannot know" is just hitting the worship button - which I'm fine with if you are talking about "what's the meaning of life." However, this is the mechanics of the universe, so we should probably stay away from that particular button in this case. Don't forget a singularity is Not an anomaly in reality itself, it is an anomaly in our models' ability to predict was will happen in reality. So time makes no sense in a singularity - that means the model for time will need to be changed. That's not the same as there is no answer.
    I don't understand why you assign a lower probability to the possibility of an infinite regress of causality, than to the possibility of a non casual event or casual loop.
    1Idan Arye4y
    To question the infinite chain of explanations you must first observe that it is indeed infinite. If the terminal explanation is always "just around the corner" you'll never reach that point.

    I like the Ignore/Explain/Worship scenario for roughly describing our epistemological options. I will note that in this particular fable you do not distinguish between different approaches to the Explain option. Mythological and scientific explanations are produced by different methods and have different qualities. I would especially note that scientific explanations have the quality of being predictive where mythological ones are not.

    My other note is that "Worship" is a loaded word. For you apparently it can mean contemplating mystery. For some the word worship could only imply one thing - the 'G" word, and you know where people go with that.

    Most of those people would 'go' "worshipping is for God only; we shouldn't be worshipping other stuff" and then hit "Explain" or "Ignore".

    Mr. Rozendaal, should we reexamine the notion of "Explain"? Perhaps the ultimate goal (from a value perspective) is power, not knowledge as such. (This would obviously constitute a testability criterion.) Or, with Bacon, we could similarly say that Knowledge is Power. Either way, the sky-spirits answer is substantively different from, for instance, Lavoisier's explanation of combustion.

    Perhaps "Explain" should be split into "delay" and "scientific answer"?

    I will note that in this particular fable you do not distinguish between different approaches to the Explain option. Mythological and scientific explanations are produced by different methods and have different qualities. I would especially note that scientific explanations have the quality of being predictive where mythological ones are not.
    It doesn't have to. You request enough explanations and you start getting answers that make sense as they probe for the shortcomings of the answers you were given. Thorough investigation was not always the norm.

    Suppose that rain actually was blood shed by large sky-going creatures? Only now, in later years, and with the conventional mistaken belief that religion is non-disprovable, do we think of "sky spirits" as a non-explanation. Back in the old days, it was a reasonable hypothesis. It's just that later it was found to be wrong.

    On the other hand, it's not clear how to test "From the before time. From the long long ago." Even in the days when people actually believed their religions, this counts as hitting Worship.

    Interpreting "Spirits", or "Gods" as physical creatures is completely missing the point, which is to attempt to describe natural phenomena in terms of human personality. Personality is more understandable to people in general than the numerical measurements and relational formulae that are the currently trendy ways of describing nature. Complaining that there are no observable physical creatures out there making rain, or whatever, is like complaining that there are no actual physical numbers or physical laws in nature, just observations, and that therefore science is nonsense.

    I've found that hitting either (E) or (I) entails a bit of (W). If you're running regressions on some enormous dataset creating some elasticity estimates, and you're pretty sure that the estimate should be positive and not negative, and you find it's negative you can either hit (E) - systematize the anomalous result: what's driving it and why is this set of datapoints not what the theory would predict - which I suppose is joined by the sentiment toward God that's either (W) God, why the f--- did you make this universe so f----- complicated or (W) thanks b... (read more)

    We can hit Explain for the Big Bang, and wait while science grinds through its process, and maybe someday it will return a perfectly good explanation. But then that will just bring up another dialog box.

    I was reminded of "can the second law of thermodynamics be reversed?", here.

    But why bother "worshipping" something entirely unlike and completely indifferent to yourself? Doesn't the "personality" of the creator in play matter a great deal in our choice of worship? You need a far more detailed argument to prove that whatever exists at the end of the recursion is worth our consideration, let alone our admiration. I see no need for anything remotely concious to end it. Unless, of course, you are just using the word "worship" as some hippy feel-good term for anything you can't explain and want to pretend not to ignore.

    I think he's using worship to mean that you believe it's beyond explanation. If you consider something to be the end of the recursion, you are worshipping it. Ignoring it would be if you hit that button before considering whether or not there's any explanation.

    You know, modern computer science gives us lots of examples of questions that we can't ever know the answer to even though they have mundane answers. These could require halting oracles to answer, but could also simply need physically unrealizable computing power due to their complexity class. Maybe science ends when the next step in the causal chain is simply provably not answerable with realistic resources.

    "God and the gods were apparitions of observation, judgment and punishment. Other sentiments towards them were secondary. The human organism always worships. First, it was the gods, then it was fame (the observation and judgment of others), next it will be self-aware systems you have built to realize truly omnipresent observation and judgment. The individual desires judgment. Without that desire, the cohesion of groups is impossible, and so is civilization."

    A reply to anonymous from a fictional character

    Such a great game. Seeing this makes me want to play it again, having discovered this site and done some actual reading on transhumanism and AI. It might change the choice I'd make at the end... Of course, this goes even further than just proving the old saying about Deus Ex, considering you never even mentioned the title! I know this is a serious necro-post, but I felt compelled.

    daaaaaaamn that's a good post. sums up exactly the way i feel about things. i'm not a scientist, but i do engage in observation, more as a poet than anything else in terms of what i end up doing or creating with that observation. the things i believe are the things i've observed. it wasn't always that way for me, but it is now.

    i recently sat and listened to robert bly read lots of poetry. he talked a bit in an offhand way about writing poetry, and what he said was, if the last line you just wrote makes sense to you, cross it out.

    somehow poets go straight t... (read more)

    Each time you hit Explain, it tells how it's a special case of a more general, more accurate, and hopefully simpler problem. There are two possibilities: At some point you get a model that explains everything with perfect accuracy. When you have that in simplest form, there's no way to Explain. You have to Worship. The other possibility is that the model keeps getting slightly more accurate and slowly gets more complex. There is simply no way to explain everything perfectly with a finite model. You just have to eventually hit Ignore. That said, if you hit ... (read more)

    Why couldn't there be a explanation that needs no explanation; an axiom? Why couldn't this list of explanations end with one of those?

    Maybe it's metaphysically possible that there's some explanation that needs no explanation, but how would we know if a particular explanation qualifies?
    It would produce no 'why' questions, there would be nothing else to explain, ignore, or worship.
    Axioms give the starting assumptions for generating consequences within logical systems. If our reality is defined in terms of such a system (which is entirely possible), the rules generating it could be expected to point back toward a small set of irreducible axioms, but that doesn't mean no "why" questions would be left; it means that definitive answers to them would be unavailable within the context of the system.

    There - I just hit Worship.

    Not sure if this is of any importance, but I thought I'd mention that this sentence is potentially syntactically ambiguous in a way that originally made me misread it. Since "hit" can be past or present tense, I originally read this sentence as saying "There - I just hit[present tense] Worship", i.e. "In that case, I'd just hit Worship", as though you were endorsing that rather than just demonstrating it; whereas presumably it was meant more as "Do you see what I did there? That constituted h... (read more)

    Ignore is a perfectly fine option. Although "bookmark" might be a better option.

    But either way, thinking and understanding can be as much of a obsessive compulsive, maladaptive behavior as anything else. It's certainly one of my maladies, and I doubt I am alone on that score around here.

    "Maladaptive" is subjective though. One's utility function may be such that their utility is only increased by explaining/understanding things. Of course if you mean that this drive supercedes eating and drink (I exaggerate), so that you undermine your ability to increase this utility in the long run, then I of course agree.

    Are Worship and Explain necessarily mutually exclusive?

    In this article's sense, yes - Explain means "try to make something less mysterious". For example, Newton Explained when devising the laws of physics, and a student who learns them is also doing some Explaining. Worship, on the other hand, means (in here, not in general) "revel in something's mysteriousness" (e.g. Newton seeing an apple fall from a tree, and saying "I guess phlogiston did it".) Ignore is the boring but sometimes practical option, saying "Eh, an apple fell from a tree" and leaving it at that.

    “In many cultures, is important to understand that stories are not explanations. They are neither true nor false because they do not describe ‘factual’ events; they do not claim that they do either. “

    Any comments on the above?

    Are you looking for an explanation or opinions?
    "Neither true nor false..." Not so. We gather such stories and treasure them. But at the end of the day, we label them fiction (or mythology, if some portion of humanity believed them to be true at some point) and know better than to go looking for Hogwarts. We know fiction is not corresponding with reality, not part of the map, in other words - not true. In every sense that matter, we treat fiction as false. All that is good and proper - as long as such works don't claim to describe factual events.

    You hit Explain for life, and get chemistry; you hit Explain for chemistry, and get atoms

    Is the first one supposed to be "biology"?

    This is also a nice complement to David Foster Wallace's speech, "This is Water".  You will worship something - hitting Ignore just risks you Worshipping by default something that ends up eating you alive.

    Is the mathematical universe "explained" to the last question? And is it specifically not mentioned in the post? And then I once thought that Yudkovsky did not like something in neural networks.


    nit: 'then'

    This is so well written it's insane. Crisp, clear, and crazily simple, despite being foundational. I have lots to learn.