But where does this dynamic nature come from? It turns out that the artificial neural network, a mathematical toy simpler than its biological counterparts, is enough to help us understand. An artificial neural network, simply put, is a web of nodes that stimulate each other, where a specific node activates if enough other nodes stimulate it. It won't necessarily take every other node's contribution equally but might be more susceptible to certain stimuli. Thus, an artificial neural network is uniquely described by which nodes are linked to which nodes and how much each link weights in the activation of another node. Something like this:
So consider this simple network that connects the activity at many end points to the activity at another end. It doesn't really matter how (for now) but let's say the network relaxes to a specific configuration, meaning that the weight of each link, thanks to some dynamical mechanism, reaches a stable value. Imagine each input end reacts to the ink of a black-and-white painting. Then the whole system answers a yes-or-no question about the image. The first two nodes will react to two different features of the image and the final node will react to a specific combination of these two features. For example, it might activate only if the image contains both features (the weight of each is not enough) or if it contains at least one of them (both weights are high). These familiar rules of logic emerge from the simplicity of the two-noded system.
In larger networks it becomes entirely hopeless to figure out accurately what each node does. One can however use a vast array of inputs, collect those which activate a given node and build from it an average over the set of inputs which the cell is sensitive to. That will tell us roughly which specific feature a specific node is encoding. This technique is one of the basic tools of neuroscience used to figure out the role of each neurone in our experience.
According to our own introspective conclusions, a web of connected features is commonly called a concept. Any active web-like structure will share this general trait of our subjective experience. It will perceive the world in connected happenings, every one of them bringing about the next. It doesn't even matter what mechanism brought it to its present configuration, whether it is according to our human opinion or to an environmental constraint a smart one or a dumb one: for as long as the universe lets it be the way it is, it will experience according to the concepts encoded in its network.
Fortunately, the universe is full of such systems, which means we won't run short of acquaintances to make too soon. Trees in a forest, chemicals in a cell, commercial exchanges across the atlantic are all examples of such connected webs which experience concepts like we do. The everywhere-presence of rich subjective experiences reminds us how our own is a naturally occurring phenomenon. The only reason the world appears to be far more complex where we stand is because most of the things our own networks react to are human-related. The same way I can’t tell the difference between two brutal death metal songs where others hear entirely different “melodies”, we humans can’t see order, reason, intent and sophistication in most of the world's phenomena, even though we try very hard with science, spirituality, art and other amazing human endeavours. Every creative action we take is an additional attempt to relate to some new feature of the universe, by creating a new structure in our own network that reacts to it. This cognitive movement is a very common one in nature. In fact, all it needs is interaction, and there’s plenty of that to go around.
No comments:
Post a Comment