First, we explain why Genetic Algorithms (GAs), inspired by the Modern Synthesis, do not accurately model biological evolution, being rather an artificial version of artificial, rather than natural selection. Being focused on optimisation, we propose two improvements of GAs, with the aim to successfully generate adapted, desired behaviour. The first one concerns phylogenetic grounding of meaning, a way to avoid the Symbol Grounding Problem. We give a definition of Phylogenetically Acquired Representations, based on a parallel between the notions of representation and of adaptation. In the second part of the paper, we propose a hybrid version of genetic algorithms, differently organizing the flow of genetic information by introducing inheritance of acquired traits and Horizontal Gene Transfer, a good tool for handle a cumulative directional process of artificial selection.
