Indeed. The learning / training we do today for ANN's clearly isn't what humans do. So yeah, even if we had billion "neuron" ANN's that were more biologically plausible, we'd probably still have to figure out more about how human learning works, in order to come up with the right way to train the AI.