Predictive Learning About the World from Actional Consequences
Jun Tani · Oxford University Press eBooks · 2016
Based on the thoughts described in chapter 6, new challenges discussed in chapters 7 through 10 concern the reconstruction of various cognitive or psychological behaviours in a set of synthetic neuro-robotics experiments. The first half of chapter 7 investigates the development of compositionality of grounded by reviewing a robotics experiment on predictive navigation learning using a simple RNN model. The second half of chapter 7, on the other hand, explores a sense of groundlessness (a sense of not to be grounded completely) through the analysis of another navigation experiment. It was shown that the developmental learning process during the exploration switched spontaneously between coherent phases and incoherent phases. The author attempts to provide an account for autonomy of self-consciousness through the analysis of such experimental observation by making possible correspondences to phenomenology.