You are here: THE PROJECT
THE PROJECT
Insight – Darwinian Neurodynamics
Our targeted breakthrough is to provide a biologically plausible basis for human problem solving and thus a bio-inspired robots for search by ICT artefacts. Open-ended human problem solving and open-ended learning remain far superior to what can currently be achieved by machines, despite considerable progress in modelling and implementing routine problem solving. We are particularly lacking adequate algorithms for insight problem solving, so important in human understanding. Insight is a process of creating new and more useful (“fitter”) representations of a problem that in many cases enhances the understanding of cause and effect, and guides future search. It has become increasingly clear that genetic evolution is a process of insightful search, because it is also able to learn from past environments to structure and improve future search operators.
These remarkable and deep similarities between thought and evolution enticed us to propose a hypothesis called the Neuronal Replicator Hypothesis (NRH) that states that a Darwinian process of production of cognitive adaptations by natural selection can run in real-time in the neuronal network of the human brain during its lifetime. Subsequently our belief in the NRH has been strengthened by the unexpectedly high levels of parsimony the hypothesis has permitted. Therefore, there is a growing desire to test the claims of the NRH. At the functional level this is done by building and examining the improvements that result by including Darwinian neurodynamics in the neural network controllers of robots for spatial learning, creative sensorimotor exploration, and language learning tasks. These models are informed by comparing the performance of Darwinian neurodynamic models with human performance in insight problem solving and language learning tasks. We predict that Darwinian neurodynamics will improve the flexibility, robustness and open-endedness of learning in a wide range of behavioural and cognitive domains, and permit bio-inspired robots to act more creatively than has previously been possible.
INSIGHT’s workplan unfold over six workpackages
These remarkable and deep similarities between thought and evolution enticed us to propose a hypothesis called the Neuronal Replicator Hypothesis (NRH) that states that a Darwinian process of production of cognitive adaptations by natural selection can run in real-time in the neuronal network of the human brain during its lifetime. Subsequently our belief in the NRH has been strengthened by the unexpectedly high levels of parsimony the hypothesis has permitted. Therefore, there is a growing desire to test the claims of the NRH. At the functional level this is done by building and examining the improvements that result by including Darwinian neurodynamics in the neural network controllers of robots for spatial learning, creative sensorimotor exploration, and language learning tasks. These models are informed by comparing the performance of Darwinian neurodynamic models with human performance in insight problem solving and language learning tasks. We predict that Darwinian neurodynamics will improve the flexibility, robustness and open-endedness of learning in a wide range of behavioural and cognitive domains, and permit bio-inspired robots to act more creatively than has previously been possible.
INSIGHT’s workplan unfold over six workpackages
