The Response of Artificial Neural Networks to Hyper-Normal Stimuli
Artificial neural networks are models of biological neural systems--they are made of individuals nodes or "neurons" that can form connections. These connections can be used to transmit and alter inputs to ultimately convey an output. Such capabilities allow them to learn patterns and evolve to become more efficacious. In our case, the neural networks were trained to categorize colors and recognize their patterns when given as stimuli. When presented with a performance-altering condition that slows down transmission--akin to a depressant--artificial the neural networks were able to gradually adjust and return to their baseline level of transmission. Networks containing very few nodes were capable of building up tolerance to depressive conditions. Additionally, neural networks were able to predict categorization patterns when presented with a range of color stimuli that were beyond the range that they were trained to respond to. In theory, the networks should have no response. What could these findings mean in terms of the complexity of artificial intelligence and how it can resemble biological neural networks? What could the implication of these findings mean for the future of animal models?