Series Progress


Introduction


In the previous article, we tried to explain some fundamental aspects that could help us grasp the purpose of Science and how certain characteristics of science come together to bring us knowledge. We approached the definition of Science, knowledge, and what is a representation in our context.


As far as we are concerned, we would like to understand how everything relates to each other and how as researchers, we can leverage this understanding to expand our perception and have several perspectives of things (aka phenomenon).


In this article, we are going to go a bit deeper into a notion we saw previously; representations. How do they come to life and their importance as researchers? We have defined what could be a representation and to recall it, we said it is an approximation of reality. This representation can change, meaning expand depending on the volume of knowledge we have collected.



Reflection


As human beings, and as long as we have the chance to live with all our faculties functioning correctly, we receive tons and tons of impressions about our world (inner and outer world). Just by sitting on a bench in a park, we receive tremendous impressions not only from our environment but also from what is happening inside us. Those impressions are what our eyes receive, what our skin senses, what our ears get, and so on. Such impressions let us understand that we are still alive.


Man sitting on a bench
Fig.1: Man sitting in a park.


Receiving such impressions is a passive phenomenon, nobody at least for the majority of people, makes a serious effort to receive impressions from their environments.


These are the source of representations. In Artificial intelligence, we usually call it raw data. At this stage, no representation is created yet (But there is a huge science behind the nature of what we call raw data either in AI or in general science), but the impressions represent the solid elements that will make possible the representations to come to life… but how?


To know how this is possible, we should be aware of some common knowledge that we have about the machine that we call human beings. The human being is not only able to receive impressions, but he has also the possibility to process them like any machine will do. And the way the human machine does this is through several processors. Some are more intuitive to us and others are less.


We will expand a lot about the nature of such processors, but not in this current monograph as it should focus only on representation. Briefly, such processors that we have to help us process several kinds of impressions (visual impressions, smell impressions, taste impressions, etc…). By nature, those impressions don’t have a special and direct meaning to us. They don’t directly tell something meaningful until our processors process them. When they process them, then we give them certain flavors like intellectual flavors, emotional, or even instinctive flavors. At this point, only, the creation of a representation is made.


So what happened?


When we receive impressions, either consciously or unconsciously our processors treat them to make sense of them. But how do they really make sense of them? Everything starts at the beginning of our existence. All impression we receive from the world is treated by our processors but we still can’t get something out of them, and for the majority of people observed, it comes from the labels they received since they were children. During that period of our life, we start making sense of the impressions by how our parents, friends, and surroundings understand them.


In artificial intelligence this is purely what we call supervised learning and in more sophisticated approaches, transfer learning. By transferring the meaning of how our surroundings(people) understand the impression into us, we slowly and passively have a sense of what is happening and we can give meaning to the impressions. This sum of impressions translating finally into some meaningful information are the representations. We should not confuse the representations and the models as the models are the process by which we obtain the representations and a slight change in the process or in the data(impressions) results in a huge difference in the representations.


So as human beings, we deeply live our existence most of the time from that passive learning we had since the early stages of our life. Depending on our demographic situation and our experience with our direct environments we will develop different representations. People living in extreme cold regions and people living in warmer ones for example have different ways of perceiving the world and even educating their children. This understanding helps us have a bit of what is happening with what we call representations.


Why is it a big deal?


As we pinpointed earlier, most of our representations come from the first part of our existence, either in a family setting, in a formal education setting, or just in social life. We definitely or rarely produced a representation in an active way. Things just happen to us and usually, our old and passive models produce “old-new” representations that align with the old ones. We called it that way because they are not new since they just enforced what was already there in our representations but the form it takes looks new.


As researchers this process has an interesting advantage because, from those old models and representations we have within us, we can rapidly infer certain possibilities that could be helpful. However, you already sensed that there is a problem. With old models and representations, we can’t produce new knowledge. We can surely combine several old models we have, to produce something we believe new, but it is not and this is the trouble most researchers and scientists face at a certain point: how to see the new? how to think “different”?


There are no deep breakthroughs in science in our current and modern society not because we are almost at the end of the knowledge acquisition journey, but just because we have difficulties in expanding our representations (The first article will help us understand if you are lost). That’s why we observe more and more new applications than new pure discoveries. It is just about combinations of old representations and models rather than the discovery of new ones.


To relate to artificial intelligence, that’s why some researchers firmly believe our models will not be able to reach any kind of AGI since one of the consequences of the definition of AGI infers that these types of models are able to discover something beyond their representations (probability distributions), which is not possible at this time.


So what is the relationship between representations and reality? If you have read the first monograph in this series, you surely have a sense of their main difference. From our perspective(human being), reality can be defined as the unknown, the ultimate target (some ancient scripts call it The Truth, but we will discuss about them in another monograph as mentioned in the structure of this series). Models are what create representations. And representations are our reality(approximation of Reality), we understand it at a certain point. Through science, we aim to get the full sense of Reality methodically ideally.


Methodically means that we need and want proofs that are empirical and logical. So as researchers, we feel like we are living in a moment in our society where we reached the top and the summum of knowledge compared to the past. It is quite true, that we have almost reached the summum of our representations which stand as our reality but, yet we are extremely far from Reality, and yet infinite things are waiting to be discovered. The difference with the past is they were more focused on foundational discoveries, on new representations than the combinations of those representations targeting applications. This switch came gradually and it has got more interest starting from the industrialization momentum.


Interesting fact, from representations we can derive knowledge, but knowledge is not a representation (We will discuss it in the third monograph that talks about philosophy as mentioned).


This is from our perspective the most beautiful journey a researcher or a scientist could start. The objective should not be to come up with solutions, but to discover what problems we have not yet identified and why. It is a painful journey because we have to break mostly everything down and move at a slower pace. It’s more difficult because as researchers we sometimes take as granted certain knowledge while we barely understand them and their implications, so a great deal in this journey is to face our Ego than to write formulas. (“lesser is the Ego, better are the formulas.”)



Conclusion


We tried to be extremely brief in this monograph as we wanted to not lose focus on what we wanted to convey which is what are really representations where do they come from and what are their importance in our life as researchers. I think the reader can intuitively map the representation definition we gave here with how it is defined in artificial intelligence. If not, this could be a nice exercise for the minds of our readers. Representations play a huge role in our lives as normal humans, and researchers but also play a considerable role in Science and obviously in artificial intelligence. Understanding how this concept influences us as researchers is extremely important if our goal is to push forward the general knowledge of our human race.


In the next monograph, we will be talking about, Philosophy and will find out its place and relations with Science. Hopefully, the next monography will bring more clarity to our understanding of Science.