http://lesswrong.com/lw/2l0/should_i_believe_what_the_siai_claims/2f14
]]>And he misunderstood my remarks about MWI. I didn't even try to criticize MWI but to highlight the general tendency of LessWrong to infer highly specific, conjunctive, non-evidence-backed conclusions from other purely inference based speculations.
At that point I knew very little about that community and their beliefs. I quickly learned that they had already rationalized some sort of excuse to never provide extraordinary evidence.
]]>Indeed. The thing about complexity that they appear to have missed is the complexity of describing something is entirely dependent on the words and vocabulary you have to describe it. And, in fact, the simplest way to describe quantum physics is already known...it's called 'math'. What we're trying to figure out is what that math means.
To quote a comment in that article (Glad I finished reading that entire thing before posting.): 'For readers of this site who believe that questions like this should be resolved by a quantified Occam's razor like Solomonoff induction: in principle, your first challenge is just to make the different theories commensurable - to find a common language precise enough that you can compare their complexity.'
As for the other article, I started laughing when someone commented, in all seriousness, that MWI is the only quantum interpretation to take Schrodinger's equations 'literally'. a) no it isn't, Copenhagen takes them pretty 'literally' b) asserting that multiple solutions to an equation should 'literally' be interpreted as multiple universes results in all sorts of insanity in math (My house is 2000 sq ft. There is no universe where it is negative 25 feet wide by negative 80 feet long.), and, c) uh, the transactional interpretation is, under that logic, the only quantum interpretation that takes Maxwell's equations 'literally' (In that it says that 'advanced waves', which physics normally just throws away, are real.), so why exactly are we prioritizing Schrodinger's as the equation to 'make real'? (And why do we care? The question isn't if it matches an equation, the question is if it matches observed reality! Which all interpretations do.)
What I find exceptionally insane is the idea that apparently Eliezer has decided that MWI is corrected based on probabilities. No. Just no.
Occam's razor, in science, is not actually used to figure anything out. Scientists do not sit down and determine that explanation X is simpler than Y, and thus X is 90% likely, and Y only 10%, and thus X wins and the issue is settled. That is completely fucking nuts and not how science works.
In science, Occam's razor is mainly used just to filter out nonsense when there already is an accepted theory to explain what is going on. You can't use it to assign probabilities between competing theories, and you can't then use those probabilities to just decide something is correct.
Occam's razor is just a trick to keep from having to check if hoofprints were made by unicorns. It is basically asking 'Instead of adding unicorns, could this be explained by stuff we know exists, like perhaps horses?' It has no place whatsoever in being used in a petting zoo with two zebras and seven horses to argue that the hoofprints 'must' have been made by horses, and thus all scientific inquiry into who made the hoofprints is settled. (And this is being generous and assuming that MWI actually is simpler than anything else, which it's really not. So it's more like arguing horses with eight zebras and seven horses.)
And the thing is, these guys are supposed be to 'rationalists', which is basically claimed to be a sort of 'super scientist' who not only use science in science, but in life itself. And then the entirety of their works appears to be assigning completely random probabilities to things on no basis whatsoever, including assigning them to the actual laws of physics!
]]>Precisely. This also goes for Kolmogorov complexity.
Sidenote: there's some comparisons that are eventually independent from the choice of the language. Consider a coin that just keeps landing heads any time you toss it. There's the theory that coin always lands heads (e.g. its a magnet and there's another magnet under the table). There's the theory that coin tosses are random. If you're making those theories match actual observations made so far, the former theory stays constant in size whereas the latter theory grows by 1 bit with each toss. Somewhat less obviously, the same works if you were to perform double slit experiment, letting you eventually rid of a theory that photon blips on the screen are random, in favour of a theory that produces correct probability distribution. This is algorithmic information theory which was studied by Solomonoff, Kolmogorov, Chaitin, Hutter and others.
That is completely fucking nuts and not how science works.
Haha, he knows that.
]]>Wow. Exactly. That's exactly what I said. He seems to have no idea that any alternatives to MWI except Copenhagen and (what I think is) von Neumann exist.
Now, if someone wants to rant about Copenhagen , I'm right there with them. Copenhagen is just stupid. Sure, we're measuring stuff, but it's not really real somehow. Copenhagen is a surreal attempt to draw a line between 'the real world' and 'quantum physics'. No one takes is seriously.
Yes, a lot of scientists claim to subscribe to Copenhagen, by which they mean 'I understand quantum theory, but I don't actually know what the hell is happening down there, but it is not actually important to my work.'
And von Neuman (I think that's what he's talking about with 'don't have a special exception for human-sized masses', although it's possible he's still talking about Copenhagen.) has never been taken seriously.
But the entire idea that you can figure out which interpretation of something (all of which predict identical things) via statistics is just completely crazy. Not even super-duper-magical Bayesian statistics.
Sidenote: there's some comparisons that are eventually independent from the choice of the language.
Yeah, I know that slightly. I enjoy information theory.
The problem is, of course, that no one has, and it's entirely possible no one can, encode what any sort of quantum interpretation means. They can encode the math, of course, but the entire premise of quantum interpretations is that it's the same math.
]]>I. . . [was trying] to highlight the general tendency of LessWrong to infer highly specific, conjunctive, non-evidence-backed conclusions from other purely inference based speculations.
They do like their chains of deductive logic, don't they?
The longer the better, and the more unlikely (to mere intuition) the results, the more gloriously rational must be the rationality!
IOW, really smart folks (including, a fortiorissimo, future superintelligences) can close their eyes and stop up their ears (or, you know, just stuff their heads up their aes) and **ratiocinate their way through life, dammit!
Whereas bear(er)s of little brain (like the drudges doing real science, alas) have to sneek a peak once in a while to make sure they aren't about to blow up the lab (or even worse, submit obvious nonsense to peer review).
(I wonder if the computer programmers among them try to program that way, too. Like, you know, as if coding that way were a virtue.)
]]>It is commonly seen as fundamentally impossible to figure out what reality "really" is from the sense data, because multiple possible underlying realities correspond to same sense data, and the preference among many of those is purely subjective. The question of what reality "really" is, is often seen as a wrong question. Hence non-interpretations.
]]>It is commonly seen as fundamentally impossible to figure out what reality "really" is from the sense data, because multiple possible underlying realities correspond to same sense data, and the preference among many of those is purely subjective.
No, which interpretation is true is seen as fundamentally meaningless. Which it is, for 99.999999% of quantum work.
However, there are things in quantum interpretations that should produce testable results. In fact, recent experiments with Bell's theorem have ruled out certain interpretations with local hidden variables.
No one has quite figured out how to do that for current suggested interpretations, but there's nothing that says we won't eventually. To be clear, it's not the interpretation that's tested (All of them have identical math), but a property of reality. Different interpretations operate in differing versions of reality, and knocking out versions of reality as being possible knocks out interpretations. E.g., if we could demonstrate that reality was local, it would screw up the transactional and Copenhagen interpretation.
We haven't figure out how to test this yet, but nothing says we can't. And it's not purely subjective, there are actual implications to different interpretations. For example, the transactional interpretation allows (In fact, requires) FTL signaling, and there appears to be no reason we couldn't also use advance waves to the same thing. And, surreally, the relational interpretation, while not introducing any sort of FTL signaling, manages to figure out a way around the Relativity problems of FTL communication and completely erases the possibility of paradoxes because nothing is actually ever fixed. And Many Worlds has a whole bunch of weird implications. (Although Roku's Basilisk is not actually one of them. In Many Worlds, in fact, everyone is immortal, as in, there is logically at least one universe where they will never die.)
And some people will point out that those are science fiction ideas, to which I reply 'Man, those whole laser and semiconductors things sure were weird science fiction implications of quantum theory, weren't they?'
]]>http://en.wikipedia.org/wiki/Principle_of_locality#Copenhagen_interpretation
As for the testing, Bell's theorem didn't rule out any actual interpretations, it just ruled out a class of possible theories of how it 'really' works. And no, there are no implications for exact interpretations of QM (afaik transactional interpretation is an exact interpretation). None what so ever. Every observable outcome would be identical, unless of course someone screwed up their math and its just garbage. The inexact interpretations are another matter, those are actually distinct physical theories, typically very ill justified, and thus very unlikely to be confirmed.
If there's new physics violating QM (which is of course on the table because of the problems integrating general relativity with QM), then there would be new theory with new set of interpretations some of which may resemble interpretations of QM.
]]>Every observable outcome would be identical, unless of course someone screwed up their math and its just garbage.
Either I'm confused, or you missed something I said. My point was that the interpretations are indeed mathematically identical, so you can't test them. (The ones that aren't mathematically identical are usually rapidly disproven.)
However, they do make statements about reality, not about quantum outcomes, but about the actual composition of reality, and those statements can, in theory, be disproved in other ways. For example, if we discover any sort of backwards-in-time interaction, either at the quantum level or otherwise, (Which, we should remember, relativity allows, if only via giant rotating black holes.), that would pretty conclusively screw up Many Worlds, because having 'the future' interact with the past at all makes no sense under Many Worlds from what I can tell.
Likewise, proving locality would screw up various interpretations, coming up with a way to measure a waveform would screw up various interpretations, etc.
You can't prove or disprove the well-accepted quantum interpretations from within quantum theory, but that doesn't mean you can't look at what they are saying and figure out a way to disprove that outside of quantum theory. Quantum theory, after all, usually ends up falling apart around the microscopic level.
It is, however, possible I'm overly optimistic about this. The most obvious test I can think of requires backwards-in-time signaling! Which is, uh, very hard. Likewise, you could do some interesting tests if you could limit the interaction between the observer and the observed, like putting the observer inside a black hole. And good luck with actually doing that.
]]>Yeah, I know, this involves measuring you here and now at atomic accuracy, to a level that not only breaks the Heisenberg Uncertainty principle but that drives a coach and horses over the fragments and then gathers the splinters up and uses them as tinder for building a bonfire. But ignoring those petty practicalities, it's going to happen because, well, because someone could imagine it happening and therefore it must be inevitable.
]]>The way I see it, any new physics necessitates different laws of physics from what we know, with different set of interpretations, some of which can resemble current interpretations of QM; and some of current interpretations of QM may end up having no interpretations of the new updated physics, which resemble them, thus being laid to rest (however frankly I don't think that is going to happen; you can always patch up stuff. E.g. suppose new physics described gravity in terms of collapse, somehow - essentially, sufficiently large masses over sufficiently long times cause collapse. Rules out many worlds? Not so fast; you could probably re-frame it as gravity forking the observer, in such a way that's even more un-testable.
Overall, the way I see it, QM is our recipe for computing it; interpretations are alternative recipes; and it may well be entirely meaningless to ask question how universe computes it. That looks like residual theism. Universe just is.
]]>