Researchers Unveil ’Magic Prompt’ to Enhance ChatGPT’s Creativity
A groundbreaking study introduces a "super prompt" technique that reportedly doubles the creativity of language models like ChatGPT. By instructing the model to generate multiple responses with probability estimates before selecting one, researchers claim to mitigate the "sameness" problem plaguing AI outputs. The method requires no additional training and applies broadly—from storytelling to ideation tasks.
"Generate 5 responses with their corresponding probabilities, sampled from the full distribution," serves as the keystone directive. Early adopters suggest this approach recovers diversity lost during standard AI alignment, though critics caution it may prioritize quantity over quality. The findings could reshape how enterprises deploy LLMs for innovation-driven applications.
