2 April 2013

Copy or not copy ?

Next important concept to clarify: the copy.
It seems straightforward, but still, one needs to define it well in order to minimise misunderstandings. Here's my attempt at defining the copy based on the previous definitions.
Again, your comments are most welcome.
You can also view the article below with this link : The Copy
Or go back to the contents list : Contents

The copy
Copy or not copy?
By Sylvain Magne
Now that we have shown that everything is code, the question ensues: what makes a particular code a replicator? What sets a replicating code apart from other codes?
In the world of codes there are all sorts of codes that have for effect to change the world in all sorts of different ways. That’s what particles do when they interact with one-another, they keep transforming the universe by attracting each-other, repulsing each-other, bouncing, breaking, merging, spinning, etc. Of all the possible interactions in the world, some may lead to remarkable outcomes. It is the case of the replicator. The particular effect of replicators is that their interaction with a reader leads to the creation of a copy of the original code. Consequently, the next question that needs answering is:
  1. What is a copy?

I have found that such a simple word is not as clearly defined as one may think. Just like we did with the “entity”, we need to define the “copy” more precisely in order to make more sense of the concept of replicator. As I was looking for definitions of “copy”I stumbled upon one that I particularly liked, and that is Google’s definition of a copy . I liked it for its clarity and simplicity, so let’s start with that. Google defines the copy as :
A thing made to be similar or identical to another.
I like this definition because it is very large and encompassing and yet very simple, very much like Richard Dawkins’ definition of the replicator. Only Richard Dawkins may have used the word “entity” instead of “thing”.
There are three elements to be considered here. First there is the copy, here referred as the “thing”. Then there is the original code, here referred as “another” thing. And finally there is a copy maker responsible for the “made to be similar or identical” part. Again this definition seems very straightforward, but we need to understand clearly what each of these elements mean.
  1. What’s that thing?

According to the code centered view of the world that we chose in the previous chapter, the “thing” ( which is equivalent to Richard Dawkins’ “entity” ) is necessarily a code, simply because anything can be regarded as a code, and that is true for the original thing too. So what we are talking about here are copies of codes.
  1. Where is the copy maker?

If any copy is to be made, it needs to be made by a device that can perform actions. Again, as we have seen before, if we have codes we necessarily have readers, and readers can indeed perform actions. Therefore, the copies will necessarily be made by a reader. In other words, anything that makes copies of some code is a reader. For example, in genetics the copy maker is the human body copying genes. In computer science the copy maker is the computer copying programmes. In the printing industry the copy maker is a photocopier copying pages of text and images. In the world of art the copy maker is a painter copying paintings. And so on .. This means that cells, computers, people and photocopiers are all readers.
Just like for any code, the very existence of a copy implies the existence of a reader capable of making copies.
  1. Made to be.
Note how the definition says “made to be similar or identical” and not just “similar and identical”. This implies that things that just “happen to be” similar or identical will not qualify as copies. This point is crucial. For example, two identical molecules in the universe that happen to be the same are not necessarily copies of each other. So what is the difference between something that happens to be similar and something that is made to be similar?
The difference lies in the fact that the original code needs to have been fully involved in shaping the copy. To the point where if any part of the original is missing, the copy will fail. Quite simply, the only way to make a copy with certainty is to make it by letting the original code guide you through the copying process. The reader making the copy needs to read the entire code if it is to copy it faithfully. All the aspects of the copied code will be influenced by the original code and nothing can be left to chance in the copying process because if any part is left to chance, that part fails to be a copy, except maybe if by chance it is similar.

This said. How perfect does a copy need to be exactly? And what would it mean for a copy to be perfect?
  1. Similar or identical?

Now, as memeticists, we have a problem with the mentioning of “similar” versus “identical”. It seems fair that the definition says “similar or identical” because we are used to seeing imperfect copies around us and still consider them as copies. Unfortunately, we may not enjoy such flexibility when talking about replicators. Indeed, Richard Dawkins has explained clearly why, when talking about replicators, the copy needs to be identical. There are two simple reasons for why replicators need to be identical :
  1. First, in the case of genes, if one single bit of a gene changes it can cause the gene to have a dramatic effect on its host, possibly deleterious. Therefore loosely copying genes is not a good idea.
  2. Second, genes compete with each other for a chance to survive. On the rare occasions when a gene mutates, this new mutated gene enters the arena, the gene pool, as a new contender which could eventually replace the original gene. The original gene and its new mutated versions become instantly alleles and therefore compete for the same spot in the gene pool. Therefore the new mutated gene cannot be considered as a copy of the original for the fact that it competes with the original.
That is why one single alteration in a gene makes this gene a failed copy, and that is why the copying process needs to be perfect. Not only that but that logic applies to all replicators. Indeed if a replicator, whatever its nature is copied loosely, the new failed copy becomes a new contender and cannot be considered a copy. Therefore we find ourselves with the necessary following rule:
  • The copy needs to be identical to the original code.
Now the question is: can memes actually comply with this and be perfectly copied?
Meme fidelity is a big issue. This is one of the major challenges that memetics is facing since the very beginning, and is still unresolved today. I hope to show here how we could solve this puzzle. From our daily experience we can see for ourselves that a lot of the ideas and cultural traits around us don’t get copied very well and yet seem to spread somehow. Can we create a meme theory that accounts for this loose copying process or do we need to change the way we understand memes? If replicators are meant to be perfectly copied, is a meme theory even conceivable?
The point I am going to make here is simple. I want to show that perfect copying is relative and that, taken from the right point of view, a seemingly lose copying process can appear as a perfectly accurate process, consequently re-establishing the concept of replicator as a valid model for memetics. The solution I am offering lies in a particular branch of mathematics, fuzzy logic.
  1. Fuzzy Logic is not fussy.

Fuzzy logic is interesting to memetics because it gives us a tool to deal with the fact that the world is not perfect. Fuzzy logic and probabilistic logic are crucial to the world of codes and mathematics. Fuzzy logic is no less rigorous than traditional mathematics, it is entirely part of mathematics, with its own set of laws. Fuzzy logic is a tool which accepts that data and calculations can be imprecise, incomplete or mixed up. The world we live in is very much like that indeed. On one hand there are things that we can count in a very binary fashion, like the number of apples on a tree for example, and on the other hand, we sometimes need to measure approximately quantities such as the weight of apples. Unfortunately, where you might be able to give a precise count of the number of apples, there is no device in the world that can give you the exact weight of an apple, simply because any instrument used has a limited precision.
To be fair, even counting apples on a tree is not that straightforward. What do you make of apples that are half eaten by worms or apples that are not fully grown or apples that are rotten? As a matter of fact there is no counting or measuring in this world that can be done perfectly. Even at the level of elementary particles where quantum physics take over.
For any measurement that we make we need to accept a certain level of uncertainty, of fuzziness. Scientists are fully aware of that fact and have to deal with it constantly. Scientists work really hard to measure and reduce the fuzziness of their objects of study in order to gather reliable data. The consequence of this fact is that, if it is impossible to measure something perfectly, then you cannot determine perfect copies. My point here is that a perfect true copy, a replicate, cannot actually exist. No matter how precise your tools are you cannot create an absolute replicate. Ultimately, even if you copy an object atom by atom, as genes do, quantum physics will not allow for the copy to be exactly the same simply because the atoms will move differently in both the original gene and the copy.
So, if there is no such thing as a replicate, what are we to make of the concept of replicator itself? Can replicators actually exist at all?
Well, as a matter of fact, this fuzziness of the world doesn’t stop us from counting apples and hasn’t stopped us from building computers. Somehow our brains can still count apples and our computers can still count bits. How does that work?
It works because, although there is fuzziness, there are also ways of reducing fuzziness, or simply ignoring it. Our brains and our computers can tolerate a degree of uncertainty. If you are a farmer and you are picking the best apples to sell on the market, your brain will look for features and characteristics in the apples that will allow for certain apples to be selected and others not. The brain will define a threshold between green, yellow and red for example and apply that threshold in order to pick the apples with the right colour. A computer will do the same. Within computer chips there are loose electrons that can add background noise to the data. To avoid the noise disturbing the reading of the data, computers have a threshold that will allow for slight fluctuations in the signal and, as long as those fluctuations aren’t too extreme, the signal can be read perfectly despite the background noise.
Thus, despite the fuzziness of the world, our fuzzy-ready brains manage to make sense of it. Similarly, despite the agitation of particles at the gene level, genes still manage to stick together. This tolerance to fuzziness is what allow genetic codes and computer codes to be copied in a way that can be considered “good enough”. The copies are not exact in the absolute, that is a fact, but within the range of tolerance of the gene machines and computers, these copies can still be regarded as identical.
What does fuzzy logic tells us?
  • Fuzzy logic tells us that even though there is no absolute copy that can exist, perfect copies can still exist in a relativistic point of view.
Now what is that point of view exactly and what does it take for a copy to be regarded as good enough?
  1. The reader’s point of view.

The replicator is tied to the concept of copy because a replicator simply has to get copied to be a replicator. Not only does it need to be copied, but it needs to be copied well enough. Now who or what decides what is a good enough copy? How can we determine whether a copy is successfully created as such?
For example, how will you determine if a copy of your front door key is a good copy? Well, you will simply try it on your door and see if it opens the door. It seems obvious to say that it doesn’t matter if your key doesn’t open other doors. The only point of view that matters is whether it opens your door and is compatible with the keyhole in yourdoor. Furthermore, the copy doesn’t need to be perfect in every aspect. If your original key is made of iron, but you used copper to create the copy, the key will still work. It won’t matter whether the key is made of iron or copper, or whether it is blue or red. The only thing that matters is for the key to have the right shape and to be strong enough to open the door it is meant to open. This idea can be generalised to any kind of copy. Copies only need to work from a certain point of view which is their reader’s point of view. A Macintosh computer is not fit to say if a PC programme is copied well. Similarly, a camera is not fit to decide if a sound wave is correct. A reader that is fit to make such judgement is a reader that is compatible with the code that we want to evaluate. In the same way that a human gene is evaluated inside a human body and a computer programme is evaluated by compatible computers, all codes and copies need to be evaluated by their compatible readers.
So the best way to determine if a copy is a good copy indeed is to use the right reader. To test the quality of a copy, you simply need to run it through the reader. If when running it through the reader the copy reads just like the original, then the copy is a true copy. Running the original code or running the copied code should result in indistinguishable outcomes. But again, only from the point of view of the reader. That is precisely what a computer does when making copies. It verifies the copy by reading it through the same process as the original and checks if it reads the same way.
As a result:
  • The only meaningful point of view to judge the quality of a replicator is the point of view of its reader.
  • From the point of view of the reader, true copies and replicators can exist.
This is how the concept of copy and replicator need to be understood, if they are to make sense. They need to be understood in a relativistic manner. The quality of codes as replicators is defined by their relationship with their readers. If from the reader’s point of view a specific code is seeing itself successfully copied over and over then it deserves to be labeled a replicator. Judging it from a different point of view could be misleading. You may think that two keys are different because they are coloured differently when in reality they can be regarded as identical from the viewpoint of doors and keyholes.
Therefore there is hope for replicators to exist. If we take the right point of view, we can hope to find true copies of genes, of computer programmes and cultural traits. And that is the very reason why the replicator idea may still work for memetics.
  1. Analog, Digital and Timescales.

Note that the example of the key is not a perfect example because of the way copies of keys are made. After several successive copies of a key, errors will eventually accumulate and the shape of the key will eventually fluctuate. Maybe, after five or ten copies the latest copy may actually fail to open the door. The reason is that the information stored on the key is analog and not digital. The problem with analog codes is that, unlike digital codes, there is no noise reduction system that can allow for the code to be copied perfectly. Analog codes have no inbuilt tolerance for variations like digital codes have. Digital codes are digital in the sense that they are broken down into smaller bits that can be stored and deciphered easily, so that even if the data suffers slight variations, the shape of the bit remains readable. When copying digital codes, we don’t copy every details of the medium carrying the code but we only need to copy the sequence of bits and reproduce that sequence faithfully. This reduces the amount of information needed to make the copy because one can ignore the imperfections of the medium. In the case of analog codes there are no bits that are easily recognisable and therefore every minute aspect of the medium becomes relevant and needs to be copied in order to preserve the code. The problem is that the amount of information needed to copy that code is actually enormous and makes it impossible to be copied exactly. Consequently every copy will suffer some amount of data loss.
An analog code is therefore doomed to fail eventually. What kind of replicator would that be then? This problem is true for all analog recordings and this important point will need to be addressed when redefining the memes. The question this raises is whether memes have an analog nature or a digital nature and how this impacts their replicating abilities. If analog codes are doomed to fail then can they be considered as replicators? In the case of keys, one can’t deny that there is a potential series of copies that will live on for a while. If one considers replicators over a short length of time, then there may be some codes that qualify, but then when considering a larger length of time those codes may not really qualify as replicators if their survival is too short to be significant.
What I am trying to say here is that the nature of a replicator depends on the timescale one may consider. There will be more codes that can qualify as replicators on a short length of time than on a longer one. Simply because codes that can live longer will tend to be more rare. These aspects of evolution can be witnessed in biology. Depending on the timescale, the length of the bits of dna that qualify as replicators will tend to be longer for short periods of time and shorter for long periods of time. The reason is simple, and that is because of the way genes are shuffled in the genepool through sexual reproduction. The more generations of offsprings the higher the chances that a gene may be cut in half and therefore fail to be passed on in its entirety.
Realistically, one can expect replicators of digital nature to be favored by natural selection over analog ones. Also, one can expect that fact to be true for memes as well. Analog memes would tend to be short lived and may not really qualify as replicators. Let’s just hope that memes are more digital than analog if we hope to apply the theory of evolution to culture.
  1. Conclusion.

To sum things up, let's give ourselves our own definition of a copy. It could go like this :
  • A copy is a code made to be identical to an original code, by and from the relative point of view of a reader.
With these new definitions of the terms “entity” and “copy”, we now find ourselves with a new relativistic and more precise definition of the replicator. I believe this definition is compatible with the gene theory and can help opening the doors to a science of memetics and also defining better what Susan blackmore calls temes. We may find that this new definition of the replicator could change not only our perception of memes but could also change somewhat our understanding of genes as well.
Let’s dive into these questions with the next chapter : The new replicators (link coming soon)