the model versus the "reality".

[geordief]

For myself, I liken modelling reality to trying to describe an object in a black bag with holes in for your hands as the only means of extracting information about it. From that you build a model which is necessarily incomplete.

Is that a model of a model ?

Edited June 24, 2016 by geordief

[swansont]

On the topic, one might want to read David Mermin's "What’s bad about this habit" from the May 2009 Physics Today. I think it's paywalled, so I don't have a useful link, but I've seen bootleg copies out there if one wants to Google it.

[StringJunky]

Is that a model of a model ?

No, it's an analogy.

On the topic, one might want to read David Mermin's "What’s bad about this habit" from the May 2009 Physics Today. I think it's paywalled, so I don't have a useful link, but I've seen bootleg copies out there if one wants to Google it.

Here's a PDF link . You've referred to it before, it's illuminating.

[Prometheus]

We take our machines and produce arrays of numbers - but to make sense of any of that we need a theory.

 

How do you think machine learning fits into this? As i understand it prediction is all that counts - it just attempts to find whichever model produces the best predictions, but this model itself doesn't aid our understanding of the phenomenon under investigation. This contrasts with traditional models which are predictive and provide a framework to understand something.

 

Could it then be said that, even though machine learning has uses, traditional modelling is better in that it also provides a framework which we can understand?

[Strange]

 

 

I think that, within the context of physics ( as opposed to other domains of enquiry ), the numerical outcome of measurements taken is precisely what constitutes objective reality, and that premise is what the scientific method is based on.

 

 

Exactly. Even if one thinks there is some sort of different reality behind that (or none, in the case of solipsism) it makes no difference. All we have is the measurements.

[ajb]

How do you think machine learning fits into this?

No idea, that seems to be another subject.

 

 

 

As long as the model matches nature well - and we have some freedom in 'well' - then the theory is 'good'. If not, then it is 'bad'

[studiot]

Concerning models, this extract from

 

Discovering Modern Algebra

 

by K L Gardner - Oxford University Press

 

may serve the whet the appetite of some to read more.

 

It really is a delightfully clear and easy to follow book.

 

[disarray]

So is there any reasonably instructive visual model of relativity somewhere that illustrates the principles of Relativity, e.g., an applet. From what little I know, it seems that time can be illustrated in a 3D graph by treating it as the 4th dimension of space, and depicting it as the growth of a graph in motion as it expands around a central point, which represents the big bang. (Perhaps asking for the applet to also shows spacetime bending back on itself is a bit much to ask).

 

I am particularly interested in whether such a graph could (like the trampoline illustration of spacetime with different sized spheres of mass rolling around on it) could show where a particular cube, located at x,y,z, might be distorted because of the presence of mass (e.g., the cube shrinks a bit, as well as, perhaps surrounding cubes). And are not the spheres or cubes of mass in such illustrations just bits of compressed energy, so to speak? Excuse my ignorance, just trying to make some sense of the theory.

Edited June 24, 2016 by disarray

[Thorham]

This is a purely philosophical question.

 

Is it? What if the curvature is there and can be measured somehow?

[Strange]

 

Is it? What if the curvature is there and can be measured somehow?

 

 

Curvature can be measured (e.g. the Pound-Reba experiment mentioned in post 9). The question is whether that reflects some underlying reality or not.

[Prometheus]

No idea, that seems to be another subject.

 

 

 

As long as the model matches nature well - and we have some freedom in 'well' - then the theory is 'good'. If not, then it is 'bad'

 

But with machine learning there is no attempt to match the model to nature; there is no underlying theory to motivate the use of any particular mathematical techniques: you just grab a big a data set as possible, split it into two subsets, 'train' the data on one subset and see how well it performs in predicting the other subset.

 

Maybe it is another subject, i just thought it might be pertinent because it seems to be a method entirely focused on prediction with no regard for any underlying theory why the method might make good predictions - it seems to be quite a fundamental shift away from traditional modelling approaches and one that negates the question of whether the model is reality.

[disarray]

Dissary said:

 

Indeed, a bit of philosophical clarification might be of use not only to scientists in that they are less likely to go down cul de sacs of reasoning (or to be mislead by their own use of everyday language), but also to avoid the sort of haphazard popularization of things such as quantum theory in the hands of the general public, e.g., the idea that quantum theory proves that free will exists, or that quantum theory proves that your dreams can come true because you just need to visualize something and it will come true, because our minds can turn light into particles or waves depending upon what we expect to see, etc.

 

Do you have any evidence that scientists are misled by such things?

 

Perhaps the word "misled" is a little strong...my point is that they may too adamantly maintain or have an excessive amount of confidence in conclusions that are philosophical in nature.

 

1. The most pertinent example with regards to this thread might arguably be Einstein, with regards to his debates with Bohr, in that scientists generally consider that Bohr was more right than Einstein, who apparently could not accept the idea, as i understand it, that causality was different (e.g., nonexistent or a matter of statistics) in certain areas of Quantum Theory. Also, he did not like the idea of superposition, as I understand it, and said something in general with regards to this that (I paraphrase) he could refused to believe that the moon ceased to exist when no one was looking at it.
2. I read at least one passage in which a scientist claims that some scientists misinterpret the philosophical intent behind Schrodinger's Cat dilemma.
3. Then there is R. Penrose's claim that quantum theory demonstrates the idea of free will and explains the nature of consciousness (controversial if not dubious claims)
4. There are also experiments done to show that lab subjects make choices subconsciously before they consciously make the same choices, with the two main people conducting the tests concluding that this proves that there is not free will.
5. Of course, going back in history a little, I could make quite a long list, e.g., the long held assumption that germs arise spontaneously (pre Leeuwenhoek).
6. Of course, there is the refusal of Church to accept Galileo's heliocentrism.
7. Also, I recall that Aristotle had a lot of theories that nowadays are considered something of a hoot.

Edited June 24, 2016 by disarray

[Strange]

Ah, I see. You are talking about personal opinions. Not science. I agree that scientists are just as likely as anyone else to have irrational opinions.

[disarray]

Strange...what? The arguments between Einstein were tooth and nail about fundamental issues of science. I fail how you interpret my post as listing personal opinion..as far as scientists are concerned, they were making very definite claims. Yes I mentioned someone has an opinion about other scientists re schrodinger, but even that is with reference to matters of scientific judgment. I don't see what distinction you are trying to make.

[Markus Hanke]

So is there any reasonably instructive visual model of relativity somewhere that illustrates the principles of Relativity, e.g., an applet. From what little I know, it seems that time can be illustrated in a 3D graph by treating it as the 4th dimension of space, and depicting it as the growth of a graph in motion as it expands around a central point, which represents the big bang. (Perhaps asking for the applet to also shows spacetime bending back on itself is a bit much to ask).

 

I am particularly interested in whether such a graph could (like the trampoline illustration of spacetime with different sized spheres of mass rolling around on it) could show where a particular cube, located at x,y,z, might be distorted because of the presence of mass (e.g., the cube shrinks a bit, as well as, perhaps surrounding cubes). And are not the spheres or cubes of mass in such illustrations just bits of compressed energy, so to speak? Excuse my ignorance, just trying to make some sense of the theory.

 

 

I don't know if you would find it instructive, but I have ( some time ago ) written a "First Primer" type article on General Relativity on my own website :

 

http://www.markushanke.net/general-relativity-for-laypeople-a-first-primer/

 

It's largely non-mathematical, and focused more on diagrams and explanations. Perhaps it can be of help to you. If you want to delve into it deeper, there are also more mathematical articles on there.

[ajb]

But with machine learning there is no attempt to match the model to nature; there is no underlying theory to motivate the use of any particular mathematical techniques: you just grab a big a data set as possible, split it into two subsets, 'train' the data on one subset and see how well it performs in predicting the other subset.

I don't see how this is really indpendent of a model. Collecting the data usually relies on models in the first place. Also, physics is all about matching models to nature, no shame in that.

[TakenItSeriously]

I once heard Dr. Michio Kaku say, in a documentary, that Einstein was never comfortable with curved spacetime and said that it was just a model to demonstrate the effects of gravity but that it shouldnt be taken literally.

[ajb]

... In my opinion the answer to this question is, briefly, this: as far as the propositions of mathematics refer to reality, they are not certain; and as far as they are certain, they do not refer to reality. ...

Albert Einstein, "Geometry and Experience", 1921

[Thorham]

Curvature can be measured (e.g. the Pound-Reba experiment mentioned in post 9). The question is whether that reflects some underlying reality or not.

 

If you can still ask that question, then you can't say that you measured the curvature. Basically, I'm asking if it's impossible in principle to perform experiments that show whether or not the curvature physically exists.

[ajb]

Basically, I'm asking if it's impossible in principle to perform experiments that show whether or not the curvature physically exists.

It is impossible. The best you can do is see if the mathematical models agree well or not with nature.

[Strange]

 

If you can still ask that question, then you can't say that you measured the curvature. Basically, I'm asking if it's impossible in principle to perform experiments that show whether or not the curvature physically exists.

 

 

The only access to what "physically exists" is through our measurements and observations. If our measurements and observations are consistent with curvature, then that is the closest we can get to saying curvature is "real".

 

To get any further than that, you would have to define more precisely what you mean by "physically exists" or "real" if they don't mean "what we observe and measure".

[geordief]

It is impossible. The best you can do is see if the mathematical models agree well or not with nature.

Suppose an observer on the Earth had access to physical graph paper that was large enough to itself be affected by the curvature of spacetime , would the observer / gaph maker need to make that graph paper with curved lines ? Or would the curvature of spacetime do that for him and so he he "attempt" to draw them straight ?

 

Perhaps my actual question makes no sense and I will accept that judgement....

[ajb]

You want to assume that the paper is light and so does not act as a source of gravity. Then you are asking if a big enough piece of paper would bend? Yes it would, there would be tidal forces that warp and bend the paper. I am not sure exactly how to answer your question, locally we have straight lines but globally on the paper the straightest lines would be curved geodesics.

[Strange]

I was going to say you could use a laser and mirrors to create a triangle in an area of high gravity and then measure the angles of the triangle to more directly observe the curvature.

 

But it is even simpler than that: gravitational lensing.

[geordief]

You want to assume that the paper is light and so does not act as a source of gravity. Then you are asking if a big enough piece of paper would bend? Yes it would, there would be tidal forces that warp and bend the paper. I am not sure exactly how to answer your question, locally we have straight lines but globally on the paper the straightest lines would be curved geodesics.

So my question makes sense? Will it help me if I use it as a (theoretical ,obviously) example for clarifying my understanding ?

 

To my mind it seems like it might be a way of mentally melding the model with the modeled

 

The paper would bend but would the lines on it bend with respect to the paper?

 

Could the parallel lines be drawn (on the paper) as never meeting but the curvature imposed on the paper itself meant that they did actually meet with respect to the wider environment?

Edited June 25, 2016 by geordief