If we examine anything in life, we can summarize what we deduce out of it to two modes: generic and specific. The generic mode is general rules that are extrapolated from specific rules by repeated observations. The only exception that I can think of are both philosophy and maths. These two sciences have fixed some axioms as general rules and then started extrapolating from it. Natural sceinces use these tools as a way to provide general rules from specific observations (experiments).
We are always moving back and forth between the two: generic and specific rules in a way we don’t always think about. This is our System 1 and System 2 mode of thinking formalized by Daniel Kahneman. Generic rules are goverened by System 1 where specific rules are governed by System 2.
Most problems I can observe in the world are related to cognitive bias which is related to either trying to deploy general rules or getting stuck on specific ones. For example, people with racial bias usually go with their intuitive and generic extrpolation from reality (System 1) not on specific treatments of indviduals according to their merits. They only observe specific rules that then supports this generic conclusion (that is why we call it also generalization).
This is very pervasive in scientific and engineering fields as well. You see people getting stuck doing somehting specific for a very long time. Software engineers have been building specific rules into programs for years now. When Machine Learning came out and with it Large Language Models (LLMs), they were shocked and couldn’t cope with it (me being one of them). Because we were stuck deploying specific rules. We were stuck in System 2 for so long that we forget to extrapolate to System 1 or generic rules. LLMs are very good at extrpolating everything into generic rules and the people stuck on the specifics became threatneded and are potentially replaceable.
In my own observation, this all comes down to Entropy. Entropy or Chaos is the enemy of life. We are always aiming at either freezing this chaos or decreasing it in a closed system if we can. That is why we always seek to develop models of reality that are usually generic and an extrapolation of it. Because it puts everything in a closed system. We can then ensure that chaos doesn’t enter into it. Chaos can come in the form of fear or danger or a violation of this model of reality.
Now are generic and specific rules a binary, meaning it’s either generic or specific? No, they are not. Generic rules can account for only some of the specific rules and not all of them. This model of reality is flawed. We cannot account for all the chaos in the world. Back to my point previously, we always aim at developing closed systems that then control this chaos. In our quest to develop these closed systems or models, we can only encode specific rules that are missing some information from the environment. That is why, in my opinion, these models are dangerous. If they are taken too seriously of course.
Science came to encourage and push people to continously adjust their models of reality based on observation. Even though this updating of the model is always flawed. The new version of your model of reality can still miss or drop some specific rules. The only hope we have is with a large enough experiements and peer reviews, we can arrive at a close approxmation of reality. This is why LLMs are very good at some tasks and not at others. This is why LLMs need a massive amount of data to reach a good point at approximating reality.
Now, let’s summarize: we have 3 main branches of knowledge: Natural, Social and Formal. Formal sciences, like Maths or Philosophy, are concerned with fixing some axioms from logic itself that were an extrapolation of centuries of thinking and observing nature. Numbers aren’t a physical tangible thing, they are a generic representation of many things. So you zoom out and you see generic representation of numbers and then when you apply it to counting money you go to the specific rules. Nautral sciences, observe reality, they go to speicific first, so they observe the motion of an object, then they try to zoom out and formalize some generic rules using Maths as their tool. Because Maths has developed the generic rules that allows you to extrapolate from specific ones in any science. Finally, social sciences like arts or languages are specific expressions of system 1 in an uncontrolled manner, then there is an attempt to extrapolate them to generic rules, for example, inguistics for languages. Linguistics attempts to observe all languages and then extrpolate to generic rules. Is the linguist’s model accurate? it can never be. Because, again, the specifics are too much. But the specifics are chaotic.
Can we then say that science is an attempt to control? even if it’s not an intentional one? Isn’t science then a tool to gain power and influence over large swaths of specific rules. These specific rules are manifested in people. People are chaotic and you can never extrapolate a model that represents everyone’s behavior or future for that matter. Is this nature’s way of protecting itself from a false god? I am saying god not in the religious sense, but in the sense that a god in generic terms would be something that can know both the generic and specific rules of anything. Because entropy isn’t a hindrance to this entity. Humans and in turn and human systems can never achieve that because they are attempting to control something that is random. The best they can get is a very good approximation or perhaps a massive deployment of intelligent agents that will allow them to map out all the specific rules as well (look at all the data centers getting built). But still, you can never predict what will happen in all the chaos and specific rules we have. I highly recommend a book by Carissa Veliz called Prophecy which goes deeper into this.
In natural sciences you can see this attempt in the “theory of everything” or the unificaiton of the standard model of particles. These are all attempts to extrapolate to generic rules. They are facing some challenges though. If you look at the physics of large objects and the quantum physics of small ones, you can clearly see that generic rules always fail at specific points. Even in Maths, axioms always assume specific ranges of input or specific cases.
I have been deliberately thinking about this in my own project, a programming language I have built, and when I applied this generic vs specific rules, interesting results came out of it. The generic rules reduced my burden of maintaining many specific rules, but again, I had to leave room for specific rules because there are always unforseen and unknown cases.
The best design that I was inspired by was Atomic Design. In Atomic Design, you break things up into atoms, molecules and organisms. I stuck with just atoms and organisms, organized in pairs. So atoms and orgnaisms are paired together (atoms compose organisms) then this pair is used as an atom for another pair of atoms and organisms and so on. This way you can gurantee that your system grows organically, so that if you can look at your system as an onion, the inner shell is a pair of atoms and orgnaisms that are generic and then as you go out, each layer becomes more and more specific till you arrive at a good approximation of the domain you are representing.
With this design, I can represent any domain of science, except the arts, because the arts are an expression of System 1 which is also the generic rules. So System 1 comes from a deployment of inutition which is in a sense controlled chaos. It can’t and shouldn’t be preditctable. This is why people are resisting the formalization of arts aggressively. People recognize that it’s not random. I am not sure which evolutionary part of our brain is responsible for that but I think we have something in us where we are botehred by art that isn’t intuitive, chaotic and somewhat random. Perhaps nature favors this randomness because it contributes to our survival. If we lose this, then we lose life itself, we lose who we are. This is why people are hurt and bothered by it. Everytime I see an AI generated video or image, something in me is bothered by it and I can’t even tell which one it’s.
Now, my final thought is this: is this article itself generic or specific? I am not sure. Is our categorization of System 1 and System 2 also a generalization and it’s missing something else? We can only hope that this model is a good representation and categorization of how we think. We have to stay vigilant and keep updating our models based on our observation of reality while keeping our biases in check. This article is 100% flawed and I know it’s. I just wanted to write my thoughts down because it does help me update my model of reality.