OpenAI co-founder Sam Altman revealed that ChatGPT has 100 million weekly users who are using the chatbot to boost their personal brand, adopt a winner’s mindset, and outperform competitors. However, many users may not be utilizing the tool to its full potential. Researchers at Mohamed bin Zayed University of AI developed a set of principles for effective prompting in a paper titled Principled Instructions Are All You Need for Questioning LLaMA-1/2, GPT-3.5/4, which entrepreneurs and business leaders can use to enhance their ChatGPT usage.
The researchers suggest several unconventional pointers to optimize ChatGPT prompts. They advise users to forget politeness when addressing the chatbot and focus on providing clear instructions instead. It is recommended to include references to the intended audience in the prompt, use sequences for complex tasks, and give positive commands for better clarity. Asking for further clarification, including examples, and even threatening punishment can also improve the effectiveness of the prompts.
Additional tips to improve ChatGPT prompts include requesting the chatbot to generate more human-like responses, instructing it on how to process information, avoiding stereotypes, and asking it to identify gaps in the provided information. Assigning a specific role to ChatGPT, using delimiters in prompts, repeating important words or phrases, combining different techniques, and prompting the desired response in advance are also suggested to enhance the quality of the generated output.
Users are encouraged to provide sample copy for emulation, be clear and concise in their prompts, avoid unnecessary politeness, and standardize their instructions for better results from ChatGPT. The art of prompting the chatbot is considered a skill that can be honed through practice, experimentation, and learning from successful outcomes. By mastering the technique of writing effective prompts, users can access the benefits of higher productivity and better-quality outputs from ChatGPT with minimal effort.
