The World As Information - Part 1

The history of philosophy is full of attempts at theories to describe reality. These range from theories that can’t even be simulated to those that simulate as worlds much like the game Minecraft. The Process Philosophy of Alfred Whitehead was almost good enough to make fully mathematical. His “almost successful” attempt to do so contributed to the formation of 20th Century mathematics. And when we use the equations of modern mathematics to represent the World, we typically measure things like matter and motions of matter. So the corresponding philosophy is called materialism or even materialistic reductionism.

Now, something exciting, or at least interesting is is happening. New concepts from information theory and computer science are coming into focus. These are concepts that would be nearly impossible for the pre-computer philosophers to access. Yet today it is obvious to every kid that information structures inside a computer can be made into complex 3D worlds.

A question arises, In materialism we store information in the position or other states of matter. The material is considered real and the information stored there is considered a construct. Could there be benefits to trying it the other way round? That is, could everything be information and when an information structure observes another it can appear to be matter in motion? This philosophy is called information realism.

In this series of articles we look at how modeling information structure can tell us a lot about the world, including some things that materialism cannot.

In this first article we briefly discuss such a world view in philosophical terms. But philosophy without math is woo-woo. So the next few articles will develop a concrete way to talk about information structure. In this series I will try not to use math-speak but a little work through the details will be surprisingly useful.

I find that math without code is impractical. So we will end the series with some practical articles that introduce actual models and code that makes them work.

The question “Is everything information?” Is a question in philosophy about something called an ontology. We can think of an ontology as a list of concepts that we consider to be “complete” in some way. For example, the website Yahoo has a hierarchical list of categories that they use to categorize every web-page. They call this their ontology. Similarly, the military has an ontology they maintain to describe military situations. In fact, everyone has an ontology in their head that they use to categorize their experiences.

A Metaphysical Ontology is an ontology where completeness means it can represent everything. And I really mean everything. Not just things like chickens and chairs, walking and beauty, but also things like numbers and math, consciousness and choice.

When we are young, before we learn science, our ontology often consists of a long list of object types, action types etc. Like bird and flying and flying fast. Soon we learn that these concepts, or at least many of them are relative or that the categories don't always apply. For example, the concept of an iPhone didn't exist in the past.

An ontology from the past held that everything was a substance. Think of a substance as being like cheese. If you cut cheese in half, both halves are still cheese. That is unlike a bike or a television where if you cut them in half, both sides are not bikes or televisions. In fact, the whole thing will no longer be a functioning bike or television. In the substance ontology things made out of substances are not fundamental and don't belong in the ontology. The philosopher Descartes theorized that there were two kinds of substances: material substances like cheese and the substance God is made of.

Today, a common ontology is that everything is made of a list of subatomic particles such as electrons and photons. So cheese isn't a substance because as you repeatedly cut it in half you eventually get to atoms, not cheese. Interestingly, newer descriptions of subatomic particles involve how they transmit information to each other.

Another formulation given today's physics would hold that there are about 18 things in reality and they are all fields that span the whole universe. Subatomic particles arise as waves in these fields. Again, it's interesting that these waves are often represented as *quantum information*.

These physics based ontologies can be called material reductionism. They have the problem that they cannot account for such things as numbers, which philosophers have had to classify as “abstract entities”. Nor can material reductionism account for consciousness or meaning and value. Actually, it cannot really even account for macroscopic objects like cheese or chickens. Much less love or emotions.

There is actually a third ontology that physicists use. They use it to describe both particles and fields, and even bigger things like the masses and positions of projectiles. This is the ontology of states.

Everything in physics — everything, and indeed in other scientific fields, is represented in terms of states. Even space and time are considered states.

Now there is a tiny equation in computers science that relates states to information. In a non-general form we have:

2 states = 1 bit of information

Or:

256 states = 1 byte of information

Also, 1 byte = 8 bits. So the conversion from states to information isn't linear.

So states and information are the same thing but under a different unit of measure.

When our brain is using a naive or a simple material ontology we try to classify information in terms of objects. That is, objects exist and their position (or other state) can store information. But when we can wrap our head around it, it works better the other way around.

Parts of this series will be a little technical and you might find yourself wondering why you should wade through them. So I want to mention a few benefits that you can remind yourself of when you're in the middle of it.

In current philosophy, namely material reductionism, there is a problem accounting for things like numbers. If everything is made of matter (or energy), what are numbers made of? You can wade into that debate by googling “Abstract entities”. But with information realism, it’s not a mystery. Numbers are just a type of information. And as we will see, so are other kinds of abstract entity such as concepts and words.

Interestingly, this also solves another philosophy problem which is the divide between math or logic statements and the world. More on this in a future article.

Many have noticed, in the last few decades, that traditional logic isn't that effective for real world argumentation. Logic that takes statements that are either true or false and deduces other statements that are true or false is extremely unwieldy. As easily as such logic can be used to both prove and disprove the existence of a god, it can be used to bolster any side of political arguments.

With the theory of information realism we can instead take any set of information pieces and deduce other information. For example, from the information in a particular photograph, perhaps one could deduce that Bob was in Hawaii with his family over the holidays. Imagine trying to use traditional logic to explain how to correctly interpret photographs. It’s not worth doing.

On the other hand, specifying how information flows from a scene, via photons, into a camera lens and into a jpg file, along with some shape information about our world (such as that encoded in 3d video games), and a few other things such as who took the picture, and so on, such an inference about Bob would be computationally trivial.

Part of the reason information based logic works better is because in 20th Century logic, all the relevant ‘if’ statements have to be spelled out. And there can be an unbounded number of them. But if you know the information structure of a situation, a possibly unbounded number of ‘if’ statements can be deduced. So it doesn’t take a huge number of rules to do complex things because a small structural specification can generate the huge number of rules.

Problems that involve inferring what will happen when we put a bunch of parts together can easily be solved by thinking about them. If I describe how to construct a bicycle from some wheels, tube's, a chain, etc. Almost everyone could figure out that it’s a bike and that it would work as a bike. Or not. If I told you that this bike does not have a chain or anything else connecting the pedals to a wheel you would instantly know that it won't work. Its not even a hard calculation. But 20th century math and logic cannot make such inferences. Material reductionism, the idea that we reduce things to their component parts and so on until we reach atoms or subatomic particles doesn't work.

But our brains must be using information reductionism internally because with information reductionism it’s pretty easy.

Suppose I have a bike on a rack and you are nearby but can only see the rear wheel. If it starts to spin you know something about the pedals even though you cannot see them. In other words, information about the pedal state of motion flows through the gears, into the chain and so on into the rear wheel. You receive the information from light that reflected off the wheel and thus about the chain and about the pedals and perhaps even that I am standing there turning the pedals.

Now suppose I had told that same story but instead of talking about pedals and chains I referred to the materials they are made of. For example, I am moving some plastic wrapped over metal (the pedals) while cause atoms in a metal lattice to send that information from atom to atoms through the lattice and so on though the metal links in the chain, through the aluminum and rubber of the wheel and so on.

Notice that from the information perspective it doesn't change the picture. The same information is flowing in the same structure and can be used to make the same inferences. A lower level-of-being is just a more detailed example of how the information was structured but the overall flow is the same. In fact, as long as the information from seeing the wheels is linked to the pedals correctly, the description does not even need to refer to the chain. Perhaps the bike doesn't even use a chain. Maybe it uses gears or something else. It’s still a bike because the information flow has the “bike” structure.

With what is to come we will be able to define and calculate this kind of thing with software.

Before embarking on this journey I want to give a brief description of what we are constructing so that you can recognize what we are making and won't wonder why the heck are we talking about this.

Think about how programming languages represent classes. A class in a programming language can represent and simulate essentially anything. Physical objects, numbers, a 3d game world. Anything. There are 3 aspects of this we are interested in. 1) how they represent the state of the system at an instant, 2) how they represent the way that state changes over time, and 3) how they facilitate interfacing the objects of the class to other objects.

The first aspect: classes represent the state of an object by listing its sub-parts as member variables. Those sub-parts have sub-parts as well with the hierarchy terminating in numbers like an int or in characters or even bits.

A good example of storing the state of something at an instant is the JavaScript spin-off JSON. JSON is a notation that can represent the state at an instant of small systems.

There are a few things we will need to fix with JSON. One is that we need to be able to represent large lists of states using expressions. In JSON if you wanted to refer to the states of a trillion atoms you would have to list each one. We will need to be able to say something like There are a trillion atoms that are roughly organized like <some expression>. It obviously should not try to allocate all trillion of them.

JSON can store the state of an object at an instant but it cannot represent how the state will change over time.

The second aspect: representing how states change over time. Typical programming languages explicitly spell out how the object will change using member functions that encapsulate state updates. The functions have a list of steps the computer should take to calculate the next state.

The problem here is that state evolution is given imperatively. We need to modify this so that state changes are represented declarativly in such a way that we can traverse changes backwards and forwards to make inferences.

In addition to representing state evolution declarativly, we need to be able to represent objects that have many things going on at the same time. For example, a bike may be being pulled down by gravity, held upright by balancing movements and pushed forward by the rear wheel’s interaction with the ground. All the while, it’s parts are holding themselves together and their atoms are doing atom things. So the objects need to be more like 3d video games objects.

The third aspect: programming language classes use public member functions to specify how they can interface to other objects.

It might be easy to hack JSON to make a language that met the above descriptions. But how would we know that the set of inferences that could be made from such models is complete? In fact it certainly would not be. For example, it wouldn't be able to refactor models to obtain different perspectives. For example, in that language everything is an object. But in reality, not everything is an object.

We want the features of what we described but it should be smaller and more mathematical (at least at the low level) to ensure we get the theory right.

What we want here is a description of very low level state systems that can be combined through operators that are similar to multiplication and addition. And through such combinations, we can build up to descriptions of anything.

By having the way state systems connect and interact be described like arithmetic we can do algebraic like manipulations to make inferences about what will happen, or what did happen given evidence.

In the next article I will show how to develop the low level theory to describe information structures. In future articles we will use the low level structure as a foundation to describe more and more complex systems.

If you have questions feel free to post them!