Core Viewpoint - The article discusses a creative competition among various AI models to generate couplets for the upcoming Year of the Horse, targeting different demographics: elderly, young adults, and elementary school students. The performance of each model is evaluated based on their ability to meet the specific needs and preferences of these groups. Group 1: Elderly Audience - The elderly prefer couplets that convey stability and traditional values, focusing on themes like health and family happiness [5][12]. - The AI model "Doubao" produced a well-rounded couplet that met traditional expectations, while "Deepseek" impressed with its cultural depth [6][8]. - "Gemini 3" provided a straightforward yet relatable couplet, while "ChatGPT" was criticized for being overly conventional and lacking creativity [11][12]. Group 2: Young Adult Audience - Young adults seek couplets that are trendy, relatable, and motivational, incorporating modern language and internet culture [13][14]. - "Doubao" effectively captured the aspirations of young workers with its use of popular phrases, while "Gemini 3" was praised for its clever integration of relatable themes [17][21]. - "Deepseek" was noted for its visual appeal but did not resonate as strongly, and "ChatGPT" was seen as too rigid and lacking in contemporary relevance [24][25]. Group 3: Elementary School Audience - For elementary students, the couplets should be simple, fun, and easy to understand, focusing on themes of learning and growth [25][26]. - "Doubao" created a clear and engaging couplet suitable for both children and parents, while "Deepseek" produced a playful couplet that resembled a children's song [27][28]. - "Gemini" also performed well, creating a lively image of a smart and eager child, whereas "ChatGPT" was deemed too formal and lacking in charm [30][31]. Group 4: AI Models' Self-Evaluation - The AI models engaged in a self-assessment round, with "Deepseek" critiquing others for their strengths and weaknesses, particularly highlighting "ChatGPT" for its lack of modern appeal [32][34]. - "ChatGPT" countered by pointing out significant errors in "Deepseek's" output, emphasizing the importance of accuracy in couplet creation [36][39]. - The final rankings indicated that "Doubao" was the most balanced, "ChatGPT" was traditional but safe, "Gemini 3" was creative but less formal, and "Deepseek" suffered from critical errors [40][44].