Neural nets vs. semantic nets
Neural nets
Neural nets are creators of labels. By attaching labels to inputs, a mind makes decisions quickly and decisively. It can recognize and classify inputs and decide what actions to trigger.
A neural net can become very good at answering a question, but it can only answer one (and only that one) question.
Neural nets are lopsided voting machines. They have many inputs which are boiled down to an output. Each output is either a label (e.g. food/poison, friend/enemy), or its corresponding action (e.g. eat/avoid, help/fight).
Since a neural net’s purpose is to make decisions quickly and decisively, its design makes it possible to turn inputs into outputs instantly.
Each neuron gathers votes from its inputs, combines them unevenly, and passes the result onward. The vote is on a sliding scale, typically represented as a fraction between 0 and 1. Each neuron gathers votes but not all the votes are added equally. The weight of a neuron’s vote can be changed to increase or decrease the impact of that vote on the final outcome.
The only output from a neural net is the final result. There is no useful way to investigate why it arrived at any particular decision, or to show any kind of justification or reason. There are no inner workings or partial answers that are worth looking at.
Semantic nets
Semantic nets are storage for labels. By storing labels and the relations among them, a mind can carefully reflect on what it knows. It can critically evaluate its labels and decide to rearrange them or create new ones.
A semantic net can be used to answer many different questions. It can also be used to ask new questions.
Semantic nets are road maps. They don’t have inputs or outputs. Each node is the label representing a single idea. Each link describes the relationship between one node and another (e.g. “the friendship between Jeff Bezos and Elon Musk”)
Since a semantic net’s purpose is to support reflection and understanding, its design makes it possible to represent complexity, exceptions to rules, and subtle shades of meaning.
A subject’s elements are a description which can be used to reason about or understand that subject.
When you use a semantic net, you apply reason as a series of steps. These steps both describe and justify the conclusion you arrive at.