AI and logic.
According to the Oxford Dictionary definition, logic is reasoning conducted or evaluated according to strict principles and validity. In the case of artificial intelligence, this definition has about the same meaning. Logic can be defined as the proof or confirmation of any reason given.
It was incredibly important for scientists to incorporate logic into artificial intelligence, because the main task was to teach the system to think and act humanly, and for this, it must be able to make any decision based on the current situation. If we talk about normal human behavior, then to make decisions we choose the best option out of many available. And there are certain reasons for choosing a particular option or rejecting options. So artificial intelligence should also work similarly.
In making any decision, the AI must be guided by the specific reasons based on which the decision is made. And this reasoning is only possible if neural networks can understand the logic.
Artificial intelligence uses 2 types of logic:
Deductive. Deductive logic provides complete proof of the truth of the conclusion made. Here artificial intelligence uses specific premises that lead to a particular conclusion. For example, this logic guides an expert system designed to select medications for a patient. The AI provides complete evidence about the medications offered — certain medications are offered to people when certain symptoms are present.
Inductive. In inductive logic, the reasoning is done using a different bottom-up approach. This means that here the AI takes specific information and then generalizes it for complete understanding. A prime example is neural network processing of natural language, during which the computer summarizes words according to their categories (verbs, nouns, prepositions, etc.) and then infers the meaning of that sentence.
Of course, the logic of artificial intelligence is somewhat different from that of humans, if only in terms of how many algorithm processes are used to achieve a result. Nevertheless, the possibility to use logic constructions in neural network operation opens new perspectives for AI in different spheres of human life.
