I found a more detailed paper here which actually explains how Spaun works. What they’re doing is really pretty far from WBE or even neuromorphic AI. They’re wrote down specific algorithms for solving the tasks they wanted Spaun to perform, and then analytically solved for neuron connection weights that would cause their model to implement those algorithms, so essentially just using the simulated brain as a very slow programmable computer.
I found a more detailed paper here which actually explains how Spaun works. What they’re doing is really pretty far from WBE or even neuromorphic AI. They’re wrote down specific algorithms for solving the tasks they wanted Spaun to perform, and then analytically solved for neuron connection weights that would cause their model to implement those algorithms, so essentially just using the simulated brain as a very slow programmable computer.