By Julia Galef
* Does the language you speak shape the way you think? Not in the way we originally believed, but there's still more than a grain of truth to the old chestnut. New York Times magazine elaborates.
* This post does a nice job of explaining why scientists are loath to accept the Many Worlds Hypothesis, and about a deep underlying tension between Bayesian reasoning and the scientific method.
* Crows are going to take over the world.
* Ergo, the card game about logic in which you try to disprove your opponent's existence. Not sure what gameplay’s like, but I love the idea. Instructions here.
* Can it really be this hard to find a scientific rebuttal of cryonics?
* A brilliant example of presenting data viscerally: housing prices, 1890-2007, graphed as a simulated roller coaster ride.
* This is a letter which has been cut out of paper and folded once. It is not L. What is it?
Seems to me like the letter could be an F, the lower vertical stem having been folded so that the bottom horizontal stem is covered.ReplyDelete
Did you happen to read "Sizing up Consciousness by Its Bits" by Carl Zimmer from yesterday's NY Times? Some of the concepts seemed really up your alley - I'd be curious to hear your thoughts on it.
Also - I still want to hear about the Transhumanist conference you went to! I think I'm more into it from the angle of a science fiction fan than anything else, but I'm just curious what one of those events is even like.
You used loath in place of unwilling or reluctant. Wow. You might want to add "Politics and the English Language" by George Orwell to your picks.ReplyDelete
Oh dear, the LW post just reminds me how little I know about quantum physics and Bayesian statistics. (But well, we all have to set priorities.) I can understand both approaches: accept only that which can be tested, or accept that which is the most elegant explanation - also a form of parsimony. The most interesting part of the argument presented is the claim that the current model of science wins out by default if a new one cannot be tested. Not so sure about that - if the new one is actually more parsimonious, all else being equal, it should still be preferred.ReplyDelete
As I know so little about all this, it would probably behoove me to remain agnostic as far as a decision between many worlds and one world goes. But then, as with certain other things we have discussed here, the thing about something not being testable is that it does not make any practical difference whatsoever whether it exists or not (if it did, that would be the test), so that assuming the many worlds do not exist seems more pragmatic.
About the cryonics, an important difference to global warming denial seems to be that somebody who takes the wager and sinks a fortune into cryopreserving their corpse does not really do any harm (except, it could be argued, to their inheritors). So there is much more motivation for scientists to fight other things they see as bunk, those that actually do harm to people and society as a whole, like the aforementioned denial or religion.
It is also possible that because, as the author of that piece also mentions, we just know so little about the details of memory and brain function today, we are simply in no position to go beyond a gut feeling on the chances of reviving people. (As for the alternative of brain emulation, I have already mentioned before what I find silly about this: it is not you. Even if it is possible at all, it would be just a copy.)
One question that I always ask myself is: what do the advocates of cryopreservation think would motivate the people of the future to sink a fortune into thawing up some random elderly and, not to be harsh, but economically essentially useless people from our time? One or two maybe, for the curiosity value, but hundreds?
Eliezer Yudkowsky writes a lot of good and interesting stuff. But quantum mechanics is one example of a problem that he has not really understood. Read the comments by Mitchell Porter! (Both on the cited post and on Less Wrong in general.) The most important of the weaknesses of the "MWI" is that it does not make any predictions at all of what the probabilities for measurement outcomes are (save for the very special case simple case when the probabilities happens to be 0 or 1 (and where also the "orthodox" interpretation is unproblematic (no collapse))), the "problem of the Born probabilities" (or at least no one has been able to show how it could make such predictions). This has to do with the "problem of preferred basis" which is still unsolved. So-called decoherence indicates a possible solution but so far it seems not to have lead to more than hand-waving type of suggestions. And the Born probabilities _are_ the predictions of quantum mechanics: if you do not have them, you have nothing. Maybe someone will be able to solve this problem and then MWI would be at least as strong a candidate as other proposed "interpretations" of quantum mechanics (which are maybe sometimes interpretations, but maybe sometimes actually theories).ReplyDelete
Probably there are also problems in combining MWI with the special theory of relativity, but in this regard MWI is probably not worse of than its present competitors (this could be a possible serious problem for many different "interpretations" of quantum mechanics, yet another open question). But this takes us to another weakness in Yudkowsky's presentation: He present the schism in quantum mechanics as between the von Neumann formulation (if I use that name) and the "MWI" only. In reality, of course, there are several candidates. Both "von Neumann" ("orthodoxy"), bohmian mechanics, and "GRW" type of "spontaneous collapse" theories does predict the Born probabilities (and the two latter clearly and non-mysteriously, although there area lot of other problems with those two interpretations/theories).
Then Yudkowsky's picture of "ideal science" is, I think, not accurate enough. That the von Neumann collapse postulate is somewhat problematic is rather generally (although not universally) acknowledged. Any quantum theory/(interpretation) making without that postulate and without introducing some new problems of the same or larger severity, and still predicting the Born probabilities, would (I think) quite quickly be accepted by the physics community as the standard formulation of quantum mechanics. No new predictions would be needed. (Although, of course, it would be even better if also new predictions were made. And another good thing would be a better compatibility between quantum mechanics and the special theory of relativity. Not to speak about general relativity!) Both "simplicity"/"occam's razor" and "unambiguous formulation" are important criteria in science. And von Neumann's collapse postulate fails unambiguousity and would have been inferior on occam's razor to MWI if MWI would have actually worked technically.
And here comes one of the real weaknesses with bayesianism, I suspect. Bayesianism claims to be able to give the probabilities for various hypothesis as evidence is collected. Forget for now about whether that claim holds up to scrutiny. Because there is a bigger problem: how to come up with the hypothesis in the first place. One could maybe say that one should treat every sentence as a hypothesis and calculate the probabilty of those according to the bayesian rules, but that would make an infinite number of hypothesis. Of course, how to come up with hypothesis/theory candidates is not a part of the scientific method. That is just how we should evaluate them once we see them. But it is a very important part of scientific work and we have no really good explanation for how it works. As for quantum mechanics, not at all unprobable would be if it turns out that a theory no one has so far imagined turns out to be the correct one. How does a bayesian calculate the probability for such an unimagined theory?
But then comes also, of course, how the bayesians get all the needed input to calculate their probabilities, even after knowing for what alternatives they should calculate them. E.g., on Less Wrong one see a lot of talk about bayesian probailities, but very few examples there they are actually calculated. (And for the few examples (see e.g. Amanda Knox case) the actual numbers seems to "be taken just out of the air")?) Now my comment starts to get off-topic because I am not anymore discussing science versus bayesian rationality, but just bayesian rationality itself. But still it is very interesting, and maybe you could write a post here at Rationally Speaking that discusses this "bayesian rationality" thing in some more detail?
I can't resist the urge to post what I think the answer to the puzzle. I say it is a capital J.ReplyDelete
It's either a capital "F" or a capital "J", or a lower case "n" or "u" depending on the meaning of "is".ReplyDelete
Aw man, Ritchie, couldn't you at least have pretended it was hard? Y'know, to spare my feelings?ReplyDelete
@hmmm -- Yes, that Orwell essay's an old favorite of mine!ReplyDelete
@Adam -- The consciousness article's interesting, thanks for the link! Seems like this guy's theory is substrate-independent (i.e., it's about information theory and relies on patterns of interactions between parts, but not on the physical nature of those parts). I would be willing to bet Chalmers had some misgivings about that which did not make it into the final article...ReplyDelete
The Summit gave me food for thought that I've been chewing on for a while. It's definitely not going to fit in a single post, but I plan to write several posts, coming up soon, about topics related to the Summit in various ways. Looking forward to hearing your thoughts about 'em.
@tonyf -- I'm no physicist, but my understanding from speaking with physicists is that the MWI does give you probabilities, and that it does not conflict with special relativity. So I'd be curious where you're getting your info on those points.ReplyDelete
... But as to your last comment, yes, Bayesian rationality is a huge and excellent topic, and one that I have considered for possible blog posts, but haven't yet found exactly right angle to write about it.
I read the article on language from the NY Times magazine you posted. It was interesting but when I got to the third page, paragraph 3, where he says, "They simply feel where north, south, west and east are, just as people with perfect pitch feel what each note is without having to calculate intervals." it made no sense. You don't "feel" north or south just like people with perfect pitch don't "feel" the note. People don't say, "I feel a color" or "I feel an odor". I can "feel" an emotion like sadness in which case I might say, "I feel blue" but that's entirely different than saying, "I feel the color blue". Unless they had Synesthesia what could that possibly mean?ReplyDelete
I have to admit that I have only the weak argument for that the MWI does not predict the Born probabilities that I find the argument for that it does unconvincing. But see e.g. David Wallace:
and judge for yourself.
OK, I have also the argument that "I have heard" this to be a problem! The bit curious thing here is that Yudkowsky himself seems to agree on this but still seems not only to favor MWI but to find it even obviously true. "Wallace, Deutsch, and others try to derive Born's Rule from decision theory. I am rather suspicious of this,":
Exactly how strong tension there is between quantum mechanics and the special theory of relativity is a controversial subject. The (possible) problem for special relativity being that quantum experiments find correlations that most simply are explained by non-local interactions. A rather good popular introduction is given by Huw Price:
Tim Maudlin has made important studies of this: "Quantum Non-Locality and Relativity" Blackwell Publishing 1994, 2002
This comment has been removed by the author.ReplyDelete