A Mathematical Model to Enhance Creativity in Generative AI Systems

Abstract

People often think it will be hard for artificial intelligence (AI) systems to copy creativity because it is a complicated and important part of human intelligence. Generative AI, which tries to make new and useful things, has shown promise in being able to copy creative processes. However, generative AI systems that are already in use often have trouble making results that are truly original. This is mostly because there isn't a clear scientific framework to help people be creative. We present a new mathematical model to improve creativity in creative AI systems in this work. Our model is based on the ideas of lateral and convergent thinking, which are important parts of how people think creatively. Divergent thinking means coming up with many ideas or solutions, while convergent thinking means picking the best ones and making them even better. To put our model into action, we come up with the idea of a "creativity score," which measures how new and useful the results are. The creativity score is determined by looking at a number of things, such as the variety of outputs, how different they are from current solutions, and how useful they are for fixing the problem at hand. Finding the right balance between divergent and convergent thinking is one of the hardest parts of making generative AI systems more creative. Divergent thinking is important for coming up with many ideas, but convergent thinking is needed to pick out the best ones and make them even better. This problem is solved by our model, which uses both divergent and convergent thought. To make sure our model works, we test it on different creative AI tasks, such as making images and writing text. Our results show that our model can greatly improve the creativity of generative AI systems, resulting in more varied, unique, and useful outputs.