Artificial Intelligence and Art: Opportunities and Challenges

Abstract

Linguistic studies of art discourse are gaining more and more relevance in the context of rapidly developing artificial intelligence. This research featured the impact of art discourse on the audience and their interpretation of images generated by AI image generators based on text queries. AI image generators open up new opportunities for research and understanding of art and art discourse, which is interpretive and dualistic in nature. Anna Andrzhievskaya’s FV and Dawn from the catalogue of Heavenly Wasteland exhibition were tested using Shedevrum and Bing Image Creator. Expressive vocabulary helped to convey the original atmosphere and ideas. However, complex metaphorical images proved to be a barrier for correct interpretation. Epithets, similes, and hyperboles provided a clearer representation of the artist’s ideas, thus reducing the risk of misinterpreting by the neural networks. These devices improved the perception of artistic value and communication between the artist and her audience. The results can help art historians in verbalizing works of art, i.e., as a writing guide for art descriptions. In addition, they may improve the quality and accuracy of AI-generated images, i.e., in training neural networks to recognize expressive vocabulary and stylistic devices at the level of linguistic pragmatics.

Keywords

• art discourse
• art description
• artificial intelligence
• neural networks
• image generation
• stylistic devises
• linguistic pragmatics