From entertainment to medicine and engineering, artificial intelligence (AI) is now being used in a wide range of fields, yet the extent to which AI can be effectively applied to the creative arts remains to be seen. In this research, a neural algorithm of artistic style was used to generate six AI paintings and these were compared with six paintings on the same theme by an amateur painter. Two sets of paintings were compared by 380 participants, 70 percent of whom had previous painting experience. Results indicate that color and line are the key elements of aesthetic appreciation. Additionally, the style transfer had a marked effect on the viewer when there was a close correspondence between the painting and the style transfer but not when there was little correspondence, indicating that AI is of limited effectiveness in modifying an existing style. Although the use of neural networks simulating human learning has come a long way in narrowing the gap between paintings produced by AI and those produced in the traditional fashion, there remains a fundamental difference in terms of aesthetic appreciation since paintings generated by AI are based on technology, while those produced by humans are based on emotion.

In producing a painting on the theme of a happy family, the human artist draws upon his related emotions and memories to form an overall artistic concept expressing joy andstability and then applies his technical skills to manifest the theme through a combination of color, lines, and brushstrokes. By contrast, an AI painting is limited to the level of technique, since a computer is unable to operate on the levels of semantics and effects. Whereas a human painting originates in the heart and mind of the artist, an AI painting is a purely technical process, such that an AI painting is more of a simulation than a genuine work of art. Whereas human art is a function of the mind, AI art is a function of technique. It should be pointed out that as AI becomes more mature this study believes that AI artists are likely to create like human artists in the future. Then, for human artists how to maintain their own advantages will be another issue worth exploring.

Acknowledgement:  The authors gratefully acknowledge the support for this research provided by the National Science Council under Grants No. MOST 110-2410-H-144-006. This video is one of the phased research results of this plan. The authors would like to appreciate the experts and participants who took part in the experiments. The author would also like to thank Emeritus Prof. John G. Kreifeldt from Tufts University for his valuable advice on data analysis and semantic accuracy. The insights of three anonymous reviewers and academic editors also made the study as comprehensive as possible.