Many attempts have been made to build an artificial brain. This paper aims to contribute to the conceptualization of an artificial learning system that functionally resembles an organic brain in a number of important neuropsychological aspects. Probably the techniques (algorithms) required are already available in various fields of artificial intelligence. However, the question is how to combine those techniques. The combination of truly autonomous learning, in which "accidental" findings (serendipity) can be used without supervision, with supervised learning from both the surrounding and previous knowledge, is still very challenging. In the event of changed circumstances, network models that can not utilize previously acquired knowledge must be completely reset, while in representation-driven networks, new formation will remain outside the scope, as we will argue. In this paper considerations to make artificial learning functionally similar to organic learning, and the type of algorithm that is necessary in the different hierarchical layers of the brain are discussed. To this end, algorithms are divided into two types: conditional algorithms (CA) and completely unsupervised learning. It is argued that in a conceptualisation of an artificial device that is functional similar to an organic learning system, both conditional learning (by applying CA’s), and non-conditional (supervised) learning must be applied.