The statement: ""There is currently no theoretical reason to use neural networks with any more than two hidden layers" was made by Heaton in his 2008 work, and reflects the theoretical perspective at that time: that multilayer perceptrons (MLPs) with more than two hidden layers had no guaranteed theoretical advantage over shallower networks. However, as is often the case in a fast-evolving field like deep learning, this claim has since been overtaken by both empirical evidence and new theoretical insights.