DE LA CREATIVITATEA UMANĂ LA CREATIVITATEA COMPUTAȚIONALĂ
FROM HUMAN TO COMPUTATIONAL CREATIVITY

Summary/Abstract: This paper is an overview of the evolution of computational creativity, focusing on the recent developments of Large Language Models’ (LLMs) creative potential. It covers topics from the theoretical background of creativity, mainly from a traditional psychological perspective to the newly designed creativity tests applied to both humans and machines. The paper summarizes the results of the studies that compare human and LLMs creative capabilities, from various perspectives, such as verbal ideational creativity, the influence of thematic roles, fanfiction generation, or stylistic and language creativity. It also points out that LLMs do not merely regurgitate information from the training data, but exhibit genuine creative behaviour, albeit merely of a combinatorial nature, proved by the results of formal creativity tests. For the first time in history, the machines, represented by LLMs, show genuine creative potential, expressed directly in natural language, and not just as a result of algorithms, like, for example, finding completely new moves or schemes in games like go or chess. The way this innovative potential will be used in future years depends only on mankind. If humans manage to overcome the initial shock and stop perceiving the machine as a threat to certain sectors like IT or the creative industry, and if human discernment remains central to the human-machine interactions, then the creativity of the LLMs can be used to boost human creativity in its own benefit. Moreover, machine creativity represents a great opportunity for creative professionals, as it can make their work easier, shorter and more efficient, as long as it is used correctly, as an assistant. On the contrary, used by creators as a subcontractor or by the common user who delegates the entire creative process to the machine, represents a threat and a degradation of the creative field, losing quality and diversity of the outputs.

• Details
• Contents