4 ms·

> But multiverses are entirely out of keeping with the Ockham's Razor approach to doing science, and constitute more of a popular interpretation of quantum mech

> But multiverses are entirely out of keeping with the Ockham's Razor approach to doing science, and constitute more of a popular interpretation of quantum mechanics than a serious physical theory. The multiverse is not a theory - it's a logical consequence of a simple theory. This claim is logically akin to claiming that the universe is identical to the observable universe, because to propose that matter exists outside of direct observation is a violation of Ockam's Razor.

dnautics 12y ago
remember, though that ockham's razor is not an axiom, it's a heuristic.
- spacehome 12y ago
  Yea, but I feel it's a good heuristic, and I'd like to hold onto it. It seems like a crude, non-mathematical formulation of Solomonoff theory. Also, Ockham's razor is the best argument in support of the Multiverse view.
  - tjradcliffe 12y ago
    Ockham's razor is a terrible heuristic, and one of the reasons the Multiverse view gets so much skepticism is that Ockham's razor is the best argument for it. Consider: Ockham himself believed that "not multiplying entities beyond necessity" was an argument for the existence of God and an argument against the existence of everything else. That is, once you had an omnipotent god, everything else was just a product of that god's imagination. This idea wasn't necessarily original to Ockham, but he was one of it's more notable proponents. Furthermore, "simplicity" is an extremely wobbly concept. To a circularist, elliptical orbits are horribly complex, as they have to be approximated by an infinite series of epicycles. Acknowledge ellipses as primary and they become simple: adding just one parameter over circles. There are of course fairly objective measures of complexity (Shannon entropy and Kolmogorov complexity) but neither of them apply very nicely to the kind of conceptual arguments where Ockham's razor is typically introduce. Finally, we know that there is exactly one consistent way to update our beliefs in the face of new evidence: Bayes' rule. Since it is very simple, we don't really need to invoke any additional heuristics like Ockham's razor. And we find, in fact, that in the few cases where Ockham's razor makes sense it is nothing but a consequence of Bayes' rule, and can be expressed as follows: In cases where two causes would give rise to the same evidence with roughly equal probability, and one cause has a much higher prior plausibility, the presence of the evidence means that cause is the more likely of the two. That is, "When you hear hoofbeats think horses not zebras" is a good idea in most cases because the prior plausibility of "There are horses around" is a whole lot higher in most cases than that of "There are zebras around". The presence of hoofbeats raises the plausibility of both propositions by exactly the same amount, but the ratio between them remains the same, so the evidence is most likely evidence of horses, not zebras.
    - spacehome 12y ago
      I agree with some of what you said, but Bayes' Rule by itself will never give you priors. Regardless of it's unfortunate origins, the way that Ockham's razor is used now in common language is a way to give approximate qualitative priors. Also, if you don't have enough mathematics to understand ellipses, then ellipses are quite complicated. I'm comfortable with ellipse-based theories needing additional evidence in this case.
- eli_gottlieb 12y ago
  No, it's a necessary consequence of the fact that the observable macrostates we care about can be composed of many, many ontologically-basic microstates in any reductionist ontology. If you live in a world where large things are composed of small things that run according to fixed rules, with causality being more fundamental the further into the small you go, then any observable "question" will have low Kolmogorov complexity compared to the total information necessary to write down the exact micro-state of the system you're asking the question about, which means simpler "questions" or "concepts" will capture larger portions of the fundamental particles' configuration-space.
Strilanc 12y ago
A better response is that it's completely in keeping with Ockham's Razor, just in terms of number of rules instead of number of things. For example, in an expanding universe we expect lots of galaxies that we see now to end up beyond the observable horizon in the future. Should we assume they stop existing when they cross that horizon (minimizing # of things) or should we avoid adding an extra rule that removes things that cross out of our horizon? Many worlds is just quantum physics without collapse. It replaces a rule with a whole lot of things. It is both simpler and more complex than Copenhagen, depending on how you weigh simplicity.

dnautics 12y ago

remember, though that ockham's razor is not an axiom, it's a heuristic.

spacehome 12y ago
Yea, but I feel it's a good heuristic, and I'd like to hold onto it. It seems like a crude, non-mathematical formulation of Solomonoff theory. Also, Ockham's razor is the best argument in support of the Multiverse view.
- tjradcliffe 12y ago
  Ockham's razor is a terrible heuristic, and one of the reasons the Multiverse view gets so much skepticism is that Ockham's razor is the best argument for it. Consider: Ockham himself believed that "not multiplying entities beyond necessity" was an argument for the existence of God and an argument against the existence of everything else. That is, once you had an omnipotent god, everything else was just a product of that god's imagination. This idea wasn't necessarily original to Ockham, but he was one of it's more notable proponents. Furthermore, "simplicity" is an extremely wobbly concept. To a circularist, elliptical orbits are horribly complex, as they have to be approximated by an infinite series of epicycles. Acknowledge ellipses as primary and they become simple: adding just one parameter over circles. There are of course fairly objective measures of complexity (Shannon entropy and Kolmogorov complexity) but neither of them apply very nicely to the kind of conceptual arguments where Ockham's razor is typically introduce. Finally, we know that there is exactly one consistent way to update our beliefs in the face of new evidence: Bayes' rule. Since it is very simple, we don't really need to invoke any additional heuristics like Ockham's razor. And we find, in fact, that in the few cases where Ockham's razor makes sense it is nothing but a consequence of Bayes' rule, and can be expressed as follows: In cases where two causes would give rise to the same evidence with roughly equal probability, and one cause has a much higher prior plausibility, the presence of the evidence means that cause is the more likely of the two. That is, "When you hear hoofbeats think horses not zebras" is a good idea in most cases because the prior plausibility of "There are horses around" is a whole lot higher in most cases than that of "There are zebras around". The presence of hoofbeats raises the plausibility of both propositions by exactly the same amount, but the ratio between them remains the same, so the evidence is most likely evidence of horses, not zebras.
  - spacehome 12y ago
    I agree with some of what you said, but Bayes' Rule by itself will never give you priors. Regardless of it's unfortunate origins, the way that Ockham's razor is used now in common language is a way to give approximate qualitative priors. Also, if you don't have enough mathematics to understand ellipses, then ellipses are quite complicated. I'm comfortable with ellipse-based theories needing additional evidence in this case.
eli_gottlieb 12y ago
No, it's a necessary consequence of the fact that the observable macrostates we care about can be composed of many, many ontologically-basic microstates in any reductionist ontology. If you live in a world where large things are composed of small things that run according to fixed rules, with causality being more fundamental the further into the small you go, then any observable "question" will have low Kolmogorov complexity compared to the total information necessary to write down the exact micro-state of the system you're asking the question about, which means simpler "questions" or "concepts" will capture larger portions of the fundamental particles' configuration-space.

Strilanc 12y ago

A better response is that it's completely in keeping with Ockham's Razor, just in terms of number of rules instead of number of things. For example, in an expanding universe we expect lots of galaxies that we see now to end up beyond the observable horizon in the future. Should we assume they stop existing when they cross that horizon (minimizing # of things) or should we avoid adding an extra rule that removes things that cross out of our horizon? Many worlds is just quantum physics without collapse. It replaces a rule with a whole lot of things. It is both simpler and more complex than Copenhagen, depending on how you weigh simplicity.