Artificial neural networks (ANN) nowadays can match human performance in challenging complex tasks, but it remains unknown whether ANNs achieve human-like performance through human-like processes. Here we used approaches and methodologies from cognitive neuroscience to pry open the black box of ANNs and to design brain-like ANNs with first principles of brain and cognitive sciences. First, we applied a reverse-correlation method of cognitive neuroscience to make explicit representations of DCNNs and humans to demonstrate that ANNs and humans can use a similar and implementation-independent representation to achieve the same computation goal. Second, equipped with the wiring cost minimization principle constrained by the wiring length of neurons in human temporal lobe, we constructed a hybrid self-organizing map model as an artificial visual cortex to explain how the abstract and complex object space is faithfully implemented in the brain to recognize occluded objects. With these two studies, we showed the possibility of bringing together research efforts from AI and cognitive neuroscience toward a new field of cognitive neurointelligence.