Computational Intelligence: are you crazy? Since when has intelligence become computational?

TL;DR

This paper critiques the concept of computational intelligence, arguing that current approaches are fundamentally flawed due to a misunderstanding of data, information, and semantics, and highlights the limitations of Shannon's Information theory.

Contribution

It challenges the validity of computational intelligence by analyzing the disconnect between data processing and semantic understanding, proposing a need to reconsider foundational assumptions.

Findings

Shannon's Information theory is limited to data communication.

Current AI approaches overlook semantic information.

Misunderstanding of intelligence hampers progress in computational AI.

Abstract

Computational Intelligence is a dead-end attempt to recreate human-like intelligence in a computing machine. The goal is unattainable because the means chosen for its accomplishment are mutually inconsistent and contradictory: "Computational" implies data processing ability while "Intelligence" implies the ability to process information. In the research community, there is a lack of interest in data versus information divergence. The cause of this indifference is the Shannon's Information theory, which has dominated the scientific community since the early 1950s. However, today it is clear that Shannon's theory is applicable only to a specific case of data communication and is inapplicable to the majority of other occasions, where information about semantic properties of a message must be taken into account. The paper will try to explain the devastating results of overlooking some of these very important issues - what is intelligence, what is semantic information, how they are interrelated and what happens when the relationship is disregarded.