5 ms·

A bit of a tangent, but I'll try to make it relevant to the topic. My third-year Cosmology course was the single most confusing and hand-wavey course I've ever

by cyphar 9y ago

A bit of a tangent, but I'll try to make it relevant to the topic. My third-year Cosmology course was the single most confusing and hand-wavey course I've ever taken during my Physics degree. It seemed to be mostly a grab-bag of "clever insights" (the cosmological constant being "wrong", but then later turning out to be "right" being the most flagrant). Maybe I just had a bad lecturer, but at the time I was working in an astrophysics research group, and I asked a few of the post-docs about cosmology and they didn't have any convincing answers either (to be fair, our research topic was asteroseismology not cosmology). It just felt as though there was a lot less rigor in cosmology than in other sciences that I was more familiar with. They definitely had some evidence for the theories presented, but most of the actual derivations felt deliberately vague and confusing. All of that being said, it feels like the author has some axe to grind and is herself engaging in unscientific dogma. Even I can tell that some of her comments on cosmology (the point about the curvature density parameter specifically) are not quite right, and I'm a little confused about her tirade on "parameters close to 1" -- not to mention that she just shrugs it off with "I have no idea" rather than following up. The points she made about choosing parameters are very odd, bordering on incorrect, as she appears to be trying to claim that unless you have a theory to derive initial parameters then the initial parameters are useless. But that's ridiculous, it's like saying (using her analogy) you need to have a theory for the position and speed of a particular arrow in order for Newton's laws to be correct. Fitting a model to data, by searching for an optimum parameter, is an incredibly common practice and is one of the most common ways of doing data science. However, saying that you need to know the probability distribution of some parameter is definitely true. It also appears that the author's views were challenged by another astrophysicist in her previous blog post, and she doesn't address this at all (in fact, she doubles down).

nonbel 9y ago
>"and I'm a little confused about her tirade on "parameters close to 1" -- not to mention that she just shrugs it off with "I have no idea" rather than following up." Why do cosmologists prefer numbers close to one? >"The points she made about choosing parameters are very odd, bordering on incorrect, as she appears to be trying to claim that unless you have a theory to derive initial parameters then the initial parameters are useless. But that's ridiculous, it's like saying (using her analogy) you need to have a theory for the position and speed of a particular arrow in order for Newton's laws to be correct." I'm not seeing it: >"The initial conditions are either designed by the experimenter or inferred from observation. Either way, they’re not predictions. They can not be predicted. That would be a logical absurdity. You can’t use a differential equation to predict its own initial conditions. If you want to speak about the probability of initial conditions you need another theory." She is just saying initial conditions are not predictions of the same equation you plug them into.
- cyphar 9y ago
  > Why do cosmologists prefer numbers close to one? Omega=+1 is preferred, because: a) We measure Omega to be approximately 1 in the modern day. This indicates that the density of our universe is approximately the critical density (for a flat universe). Really, Omega=+1 means that our universe is flat. b) In order for Omega to be approximately 1 _today_ in an expanding universe, it must have been even more close to 1 at t=0. The only example of a parameter that she gave is omega. She didn't give any other examples, so it's quite a stretch to say that "cosmologists prefer numbers closer to one). Here's the wikipedia article on the problem: https://en.wikipedia.org/wiki/Flatness_problem https://en.wikipedia.org/wiki/Flatness_problem. All of the points in that article were things we discussed in third-year cosmology as possible solutions to this fine-tuning problem. > She is just saying initial conditions are not predictions of the same equation you plug them into. The initial conditions are deduced by looking at modern observations and "winding back the clock" to see what set of parameters match the observed facts. I don't know what she's referring to "using a differential equation to predict its own initial conditions". And you can create a probability distribution of possible initial conditions for omega, it's just very small (of the order 1e-62). Inflation is the proposed physical explanation for why that is the case.
  - nonbel 9y ago
    Thanks, So for the first issue you claim her premise that cosmologists arbitrarily prefer values "near" to 1 is incorrect. Ok, I have no idea. I guess that would make for a good review article or even just some science journalism where all the bigshots get asked. For the second issue you seem to be in agreement with her for the most part: >"What happens if you ignore this and go with the belief that the likely initial value for the curvature density should be about 1? Well, then you do have a problem indeed, because that’s incompatible with data to a high level of significance. Inflation then “solves” this supposed problem by taking the initial value and shrinking it by, I dunno, 100 or so orders of magnitude. This has the consequence that if you start with something of order 1 and add inflation, the result today is compatible with observation. But of course if you start with some very large value, say 1060, then the result will still be incompatible with data. That is, you really need the assumption that the initial values are likely to be of order 1. Or, to put it differently, you are not allowed to ask why the initial value was not larger than some other number." She is claiming that the inflation theory presupposes the initial value should be near 1, this value is not derived from anywhere.
wyager 9y ago
Cosmology was the only physics class I ever dropped. Very interesting, but I got the impression that it is very disorganized as a field and (I felt) not worth the time given how little I was learning after spending all my time trying to put together piecemeal conjectures into a coherent understanding. Its bounds for acceptable error levels were also much wider than other fields in physics, which made it hard to ground your understanding in something precise.
mickronome 9y ago
Nah, I don't think she has an axe to grind, except maybe to chop down bad arguments in general. Anyway, her argument about 1 in that particular case is afaiu simply that it's rather preposterous to assume a specific (essentially impossible to know) iv. and then massage the theory around it until the results fit the observations. In data science it would be like determining the model parameters by gut feeling, and then start searching for some kind of model with those parameters that actually fit the data. Almost exactly the opposite of how an parameter search/optimization is usually done. I tend to start with some model and see what parameters that fit instead of randomly deciding one parameter is 1 and then go look for a model that might fit my data with that arbitrarily fixed parameter. I'm no astrophysicist, or anything even remotely close to the authors competence, so I'll refrain from commenting on those points, but I really see no reason to critique her arguments in general because somebody pointed out some possible weakness of one a sub argument. Rather the opposite, if you would only have one counter comment on a physics related blog, then you are probably correct, nobody has read it yet, or the comment function is broken.
- cyphar 9y ago
  Her tirade about omega=+1 is odd because she talks about how it's incredibly anti-science, but doesn't even try to explain what the opposing view is (going so far as to say that she hasn't the foggiest). The first question I would want to know when evaluating the claim that omega=+1 is bad science is to hear the argument for it. From my brief touches with cosmology, omega=+1 comes from measurements that the universe appears to have a flat topology (both from the size of structure in the CMBR and from a variety of other measurements). The flatness problem effectively is a fine-tuning problem that states if omega has slight deviations from +1 (of the order 1e-62) then the universe of today would not be able to sustain galaxies or other large structures. The measurement of _today's_ omega (omega_0) is also very close to +1 (which is what forces the initial condition limit). From that, it was concluded that if we accept the model of an accelerating universe, then the initial condition must be close to +1. Inflation is the proposed physical solution to explain why that is the case. To use the Newton analogy, you are backtracking the path of an arrow to find where it was fired from. To be clear, I personally don't like how hand-wavey cosmology is as a science, but pretending as though modern cosmologists just came up with a random "cult-like" belief that omega=1 is just ludicrous. [The author calls omega 'k'. While omega and k are related (k is the curvature, and can be used to deduce the critical density -- which omega is a fraction of) I'm not sure if that's quite right. But I'm speaking as a senior undergrad here, I might've missed something.] As for the "possible weakness of a sub-argument", you can go to the reddit discussion of this particular post (others have linked it here) and see for yourself. It's quite a bit more than a "possible weakness of a sub-argument". Effectively the analysis is on her entire thesis (that the Bullet-cluster can be explained with modified gravity perfectly fine). The rest of the complains about modern cosmology are sub-arguments to justify why modified gravity should be considered.

by cyphar 9y ago

nonbel 9y ago

>"and I'm a little confused about her tirade on "parameters close to 1" -- not to mention that she just shrugs it off with "I have no idea" rather than following up." Why do cosmologists prefer numbers close to one? >"The points she made about choosing parameters are very odd, bordering on incorrect, as she appears to be trying to claim that unless you have a theory to derive initial parameters then the initial parameters are useless. But that's ridiculous, it's like saying (using her analogy) you need to have a theory for the position and speed of a particular arrow in order for Newton's laws to be correct." I'm not seeing it: >"The initial conditions are either designed by the experimenter or inferred from observation. Either way, they’re not predictions. They can not be predicted. That would be a logical absurdity. You can’t use a differential equation to predict its own initial conditions. If you want to speak about the probability of initial conditions you need another theory." She is just saying initial conditions are not predictions of the same equation you plug them into.

cyphar 9y ago
> Why do cosmologists prefer numbers close to one? Omega=+1 is preferred, because: a) We measure Omega to be approximately 1 in the modern day. This indicates that the density of our universe is approximately the critical density (for a flat universe). Really, Omega=+1 means that our universe is flat. b) In order for Omega to be approximately 1 _today_ in an expanding universe, it must have been even more close to 1 at t=0. The only example of a parameter that she gave is omega. She didn't give any other examples, so it's quite a stretch to say that "cosmologists prefer numbers closer to one). Here's the wikipedia article on the problem: https://en.wikipedia.org/wiki/Flatness_problem https://en.wikipedia.org/wiki/Flatness_problem. All of the points in that article were things we discussed in third-year cosmology as possible solutions to this fine-tuning problem. > She is just saying initial conditions are not predictions of the same equation you plug them into. The initial conditions are deduced by looking at modern observations and "winding back the clock" to see what set of parameters match the observed facts. I don't know what she's referring to "using a differential equation to predict its own initial conditions". And you can create a probability distribution of possible initial conditions for omega, it's just very small (of the order 1e-62). Inflation is the proposed physical explanation for why that is the case.
- nonbel 9y ago
  Thanks, So for the first issue you claim her premise that cosmologists arbitrarily prefer values "near" to 1 is incorrect. Ok, I have no idea. I guess that would make for a good review article or even just some science journalism where all the bigshots get asked. For the second issue you seem to be in agreement with her for the most part: >"What happens if you ignore this and go with the belief that the likely initial value for the curvature density should be about 1? Well, then you do have a problem indeed, because that’s incompatible with data to a high level of significance. Inflation then “solves” this supposed problem by taking the initial value and shrinking it by, I dunno, 100 or so orders of magnitude. This has the consequence that if you start with something of order 1 and add inflation, the result today is compatible with observation. But of course if you start with some very large value, say 1060, then the result will still be incompatible with data. That is, you really need the assumption that the initial values are likely to be of order 1. Or, to put it differently, you are not allowed to ask why the initial value was not larger than some other number." She is claiming that the inflation theory presupposes the initial value should be near 1, this value is not derived from anywhere.

wyager 9y ago

Cosmology was the only physics class I ever dropped. Very interesting, but I got the impression that it is very disorganized as a field and (I felt) not worth the time given how little I was learning after spending all my time trying to put together piecemeal conjectures into a coherent understanding. Its bounds for acceptable error levels were also much wider than other fields in physics, which made it hard to ground your understanding in something precise.

mickronome 9y ago

Nah, I don't think she has an axe to grind, except maybe to chop down bad arguments in general. Anyway, her argument about 1 in that particular case is afaiu simply that it's rather preposterous to assume a specific (essentially impossible to know) iv. and then massage the theory around it until the results fit the observations. In data science it would be like determining the model parameters by gut feeling, and then start searching for some kind of model with those parameters that actually fit the data. Almost exactly the opposite of how an parameter search/optimization is usually done. I tend to start with some model and see what parameters that fit instead of randomly deciding one parameter is 1 and then go look for a model that might fit my data with that arbitrarily fixed parameter. I'm no astrophysicist, or anything even remotely close to the authors competence, so I'll refrain from commenting on those points, but I really see no reason to critique her arguments in general because somebody pointed out some possible weakness of one a sub argument. Rather the opposite, if you would only have one counter comment on a physics related blog, then you are probably correct, nobody has read it yet, or the comment function is broken.

cyphar 9y ago
Her tirade about omega=+1 is odd because she talks about how it's incredibly anti-science, but doesn't even try to explain what the opposing view is (going so far as to say that she hasn't the foggiest). The first question I would want to know when evaluating the claim that omega=+1 is bad science is to hear the argument for it. From my brief touches with cosmology, omega=+1 comes from measurements that the universe appears to have a flat topology (both from the size of structure in the CMBR and from a variety of other measurements). The flatness problem effectively is a fine-tuning problem that states if omega has slight deviations from +1 (of the order 1e-62) then the universe of today would not be able to sustain galaxies or other large structures. The measurement of _today's_ omega (omega_0) is also very close to +1 (which is what forces the initial condition limit). From that, it was concluded that if we accept the model of an accelerating universe, then the initial condition must be close to +1. Inflation is the proposed physical solution to explain why that is the case. To use the Newton analogy, you are backtracking the path of an arrow to find where it was fired from. To be clear, I personally don't like how hand-wavey cosmology is as a science, but pretending as though modern cosmologists just came up with a random "cult-like" belief that omega=1 is just ludicrous. [The author calls omega 'k'. While omega and k are related (k is the curvature, and can be used to deduce the critical density -- which omega is a fraction of) I'm not sure if that's quite right. But I'm speaking as a senior undergrad here, I might've missed something.] As for the "possible weakness of a sub-argument", you can go to the reddit discussion of this particular post (others have linked it here) and see for yourself. It's quite a bit more than a "possible weakness of a sub-argument". Effectively the analysis is on her entire thesis (that the Bullet-cluster can be explained with modified gravity perfectly fine). The rest of the complains about modern cosmology are sub-arguments to justify why modified gravity should be considered.