Gutenberg Test Part 2: Testing a Creative AI Agent

This paper is the second part of the "Gutenberg Test", which examines the ability of an artist-created AI agent to reproduce creative processes. In the first part of the experiment, the author-artist created a "creative AI agent" consisting of three "artificial personalities", each with its own unique specializations, knowledge, and biographical details. This paper examines the practical part of the "Gutenberg Test", during which a creative AI agent is tested for its ability to reproduce the creative processes laid down by the artist. The "AI agent" is given to a user who does not have creative skills, who uses it to create generative images. In addition to testing in various scenarios, the agent is tested for its ability to generate artifacts in the "original" artistic image. In the final part of the experiment, the generated images are given to the artist, the creator of the AI agent, for analysis and evaluation. The final part of the article is devoted to the analysis and interpretation of the mechanisms involved in the process of creating and evaluating creative works, as well as the possibilities of reproducing the generated artifacts by non-digital means. The experiment demonstrates both the possibilities and limitations of large language models in the field of creativity, providing a cumulative assessment of the “reproducibility” of creative processes within the framework of neural network models.