Logical model of the physical world

[miksu]

Hi,
are there any attempts to make a machine readable definition of most common rules and actions (or a causal model) of physical world? I'm looking more a logical level of modelling, at the same level that human common sense works.

I have some experience on semantic tecnologies and ontologies, and they usually define classes and individuals with properties, for example machine readable version of Wikipedia. What I'm looking for, is a set of descriptions of what physical properties these instances can have and how to manipulate them.

Let's make an artificial example of cake. So we have a class https://en.wikipedia.org/wiki/Cake. We would then add some physical properties to is, such as size and weight (range of) and a the fact that it permanently deforms on impact (dropping it will mash it, unlike dropping a ball).
Then we can have other data, independent of the cake, such as defined action "drop object" which causes an object to change location in free fall until it hits some other object and deforms, or bounces, depending of the defined properties. With this we can then run a simulation what happens to an instance of cake if it is dropped.

So with this model it would be possible to infer what happens to objects on physical events on some logical, common sense level.

 

[russ_watters]

It isn't clear to me what the scale/scope and use would be for what you are asking. But...

Simulations tend to be highly specific and limited in scope due to the computing power needed to run them. So they tend to be tailor-made or copied and modified from previous simulations.

 

[anorlunda]

I understand the appeal. I was once teased by the chance to make an object model of the whole electric power industry. Technical, finance, admin, social, political, including every job description. It was tremendously appealing to an analyst But it was shot down because:

1. There is no way to limit the scope. It could be a forever project. And the bigger the model, the more cumbersome it becomes.
2. In your example, the method "drop" might apply to many objects, but other programmers might choose "shock loading", or "shear" as their preference. Wrong choice of method could poison the whole project. We have no standards to clearly separate right from wrong.
3. There is no objective evidence that such a model would be useful compared to existing software models and even common sense.

I believe those are the biggest reasons that most simulations tend to be highly specific as @russ_watters said.

 

[Klystron]

Your "cake" model reminds me of a series of lectures and books by my combinatorics / computer science teacher Rudolf von Bitter Rucker. The wikipedia entry is incomplete and since this is a science thread we should only source his mathematics texts, but Prof Rucker develops fine semantic and CS arguments based on terms such as 'cake', 'rake' and, with a nod to Charles Dodgson, 'teapot' as reality descriptors; i.e., symbols.

'Teapot symbology' helped me understand topological constructs like Klein bottles and 'cake rake membranes' algebraic constructs in n-dimensions such as Calabi-Yau manifolds.

 

[miksu]

Thank you for the replies. I was afraid this would impossible task, or huge atleast. There are few bigger knowledge base projects such as Cyc and ConceptNet which try to model actions too, for example http://conceptnet.io/c/en/cake. But they are too random to be usable for any simulation.