A Model of Computational Creativity based on Engram Cell Theory

Abstract

Artificial Intelligence technology has made remarkable progress in machine learning, but it is still in its infancy in creative thinking or computational creativity. In 2018, Yang and Li proposed that the physiological basis for the formation of memories and concepts in the human brain is engram cells (interneuron), and creative thinking is the process of forming new engram cells to connect previously seemingly unrelated concepts. During this process, association and prediction play a key role. In this study, a computational model based on engram cell theory was coded in Python to mimic the process of creative thinking. The validity of the model was tested by simulating the phenomenon of language generation and summarizing the artificial food-set regularity in the plus maze. The results show that, given 29 initial words and certain grammatical rules, the language generation program generates 25,405 sentences after 130,000 calculations, and these generated sentences can be combined into various short paragraphs. After 50 times of training in the cross maze puzzle solving program, the model can master 100% of the rules of artificial food settings. In conclusion, a computational model of creative thinking based on engram cell theory can creatively and automatically generate sentences and paragraphs, and can learn and summarize laws to solve simple puzzles. We plan to further use this model to address complex real-world problems, such as the study of cancer therapeutic targets