Combined rule-based and generative artificial intelligence in the design of smartphone optics

This paper reports on a study of design methodology for smartphone lenses utilizing generative and rule-based artificial intelligence (AI) algorithms. The proposed innovative design method utilizes the GPT-4, an OpenAI model, to generate macros for global optimization algorithms used in designing smartphone lenses. A comprehensive global search for optimal starting points of smartphone lenses has been conducted to obtain training sets. The training of a generative AI model for lens design was carried out through the application of prompt engineering techniques. We trained a GPT-4 model developing a framework for coding macros, creating a merit function, and evaluating the obtained starting designs. The results demonstrate the practical value of the proposed design methodology based on AI algorithms in the design of a 21.4 megapixel smartphone lens.