As we mentioned, when discovering new facts and more details, not only do they need to be inserted into the existing set, but the validity of all assumptions and approximations must be revisited. Maybe the discovery supersedes one or more of the old items, provides a new hypothesis in place of an old one, turns a hypothesis into fact, or question marks the overall validity (at least part) of the set of approximations and abstractions. Alternatively, it turns a hypothesis into a fallacy.
Neuroanatomy provided an unbelievable wealth of details about the CNS, its structure, components, their infinite variety of implementation, connection, chemical/enzymatical composition, and so on. A vast amount of data is collected and available, and uncountable attempts (mathematical models) have been made to describe the actual phenomena. However, focusing on too many details prevents understanding that ”the nervous systems adopts a number of basic principles” [2]. The illusion of having an imposing knowledge base inspired undertakings such as simulating the entire human brain.
The brain must be studied ”from Inside Out” [151]. First of all, understanding how the known and established physical laws underpin the operation of single neurons (the interface of non-living matter and living matter) instead of hypothesizing additional laws and phenomena complementing/overwriting them, led to creating a fictive nature, which in some points resemblant to the real one. The abstraction, the discrete components connected by ideal wires that can describe electric phenomena, is successful in electronics. However, it is not valid for neurons (see mainly the electrotonic models), although some rough resemblance indeed exists. The basic differences are that the structure of living matter is different and that many interactions with drastically different interaction speeds are behind the phenomena, as contrasted with the (mostly) single interaction with a single speed of electronics.
Nature is based on the collective operation of single neurons and is prepared to consider the finite operating and transmitting times and uncertainties/failures of operation, unlike (most) technical networks. However, the usual neural network models [118] do not consider those differences.