The objective of this thesis was to study the concept and methods of artificial intelligence creativity. The methodology consisted of several steps. Firstly, the basic structure of artificial neural networks was examined and compared to that of biological neural networks. The neuroscience of creativity was researched, and connections were drawn to the respective structures of different types of artificial neural networks. It was concluded that are the neuroscientific construction of creativity in the brain is most closely resembled by the recurrent neural network and the generative adversarial neural network. Examples of visual and literary artificial intelligence creativity were researched and presented.

The executional goal of the thesis was implementing different neural networks and observing how they compare to one another when given a creative task. The two types of models were a long short-term memory recurrent neural network and a generative adversarial neural network. They were fed a dataset of fictional book summaries and instructed to output sentences. Overall, the LSTM RNN performed better in terms of coherence and structure of the sentences. Both networks produced vivid imagery through their words on several occasions.