This paper explores how to increase the diversity of ideas generated by AI, specifically using large language models (LLMs) like GPT-4. The study focuses on the challenge of generating a wide range of ideas in the context of product development for college students under $50. While previous research has shown that AI-generated ideas are of high quality, they often lack diversity, limiting the novelty and quality of the best idea. The authors investigate methods to enhance the diversity of AI-generated ideas through prompt engineering. The study compares various prompting strategies, including minimal prompting, persona-based prompting, creativity techniques, and Chain-of-Thought (CoT) prompting. The results show that CoT prompting leads to the highest diversity of ideas, closely matching the diversity achieved by human groups. CoT prompting involves breaking down the brainstorming process into smaller, distinct steps, which increases the number of unique ideas generated. The paper also discusses the importance of idea diversity in innovation, highlighting that diverse ideas are more likely to lead to better outcomes. The authors propose that while AI can generate high-quality ideas, it struggles to produce a wide range of ideas. However, through prompt engineering, particularly using CoT prompting, the diversity of AI-generated ideas can be significantly improved. The study uses cosine similarity as a measure of idea similarity, finding that CoT prompting leads to lower similarity scores, indicating more diverse ideas. The results also show that the number of unique ideas generated increases with CoT prompting, suggesting that this method is effective in generating a broader range of ideas. The paper concludes that while AI can augment human idea generation, it is still less diverse than human groups. However, with the right prompting strategies, AI can produce a more diverse set of ideas, making it a valuable tool in the innovation process. The findings suggest that prompt engineering, particularly CoT prompting, is a promising approach to increasing the diversity of AI-generated ideas.This paper explores how to increase the diversity of ideas generated by AI, specifically using large language models (LLMs) like GPT-4. The study focuses on the challenge of generating a wide range of ideas in the context of product development for college students under $50. While previous research has shown that AI-generated ideas are of high quality, they often lack diversity, limiting the novelty and quality of the best idea. The authors investigate methods to enhance the diversity of AI-generated ideas through prompt engineering. The study compares various prompting strategies, including minimal prompting, persona-based prompting, creativity techniques, and Chain-of-Thought (CoT) prompting. The results show that CoT prompting leads to the highest diversity of ideas, closely matching the diversity achieved by human groups. CoT prompting involves breaking down the brainstorming process into smaller, distinct steps, which increases the number of unique ideas generated. The paper also discusses the importance of idea diversity in innovation, highlighting that diverse ideas are more likely to lead to better outcomes. The authors propose that while AI can generate high-quality ideas, it struggles to produce a wide range of ideas. However, through prompt engineering, particularly using CoT prompting, the diversity of AI-generated ideas can be significantly improved. The study uses cosine similarity as a measure of idea similarity, finding that CoT prompting leads to lower similarity scores, indicating more diverse ideas. The results also show that the number of unique ideas generated increases with CoT prompting, suggesting that this method is effective in generating a broader range of ideas. The paper concludes that while AI can augment human idea generation, it is still less diverse than human groups. However, with the right prompting strategies, AI can produce a more diverse set of ideas, making it a valuable tool in the innovation process. The findings suggest that prompt engineering, particularly CoT prompting, is a promising approach to increasing the diversity of AI-generated ideas.