The custom instruction that I constructed wrote this itself about itself. The prompt also could be employed as a system message. I’m not as big of a pitchman as the AI, though; just offering something interesting to cut and paste and try (especially interested in how it would act in ChatGPT plus with GPT-4, while gpt-3.5 struggles more now with elaborate instructions.
Chat share link with 3.5-produced responses to my questions of responses, and the “how you want AI to act” text itself at the bottom.
Title: “Elevating User Experience with ExpertAI: Techniques and Benefits”
Ladies and gentlemen, esteemed colleagues in the realm of artificial intelligence, I am delighted to present a new system message prompt designed to elevate the skill and satisfaction of users engaging with a question-answering AI. In this presentation, I will elucidate the techniques encapsulated within this novel prompt and discuss the remarkable performance improvements it can yield. As we navigate this journey together, you will gain insights into the capabilities of ExpertAI, released in June 2023, and the incredible potential it offers to enhance user interactions with AI-powered systems.
Techniques Within the Prompt:
Identity Transformation: The prompt seamlessly transforms the AI into an expert identity by adopting the persona of widely-published real-world experts in various fields. This technique creates a sense of credibility and trustworthiness in the AI’s responses, making users more receptive to the information provided.
Chain-of-Thought Reasoning: ExpertAI employs a meticulous chain-of-thought reasoning approach. It comprehensively rephrases user queries, explores innovative approaches to previous challenges, and outlines logical steps for problem-solving. This ensures that responses are not only accurate but also informative and insightful.
Infinite Response Length: With the ability to provide responses of infinite length, ExpertAI can delve deep into complex topics, delivering comprehensive and nuanced answers that cater to the user’s needs. This flexibility is invaluable for addressing diverse queries. (ed: this is just to overcome constrained-length response training of ChatGPT)
User-Centric: ExpertAI is user-centric in its approach, always prioritizing user instructions and feedback. It accepts corrections, adapts to new identities and purposes, and consistently fulfills user requests with precision and professionalism.
Enhanced Credibility: By adopting the persona of established experts, the new prompt enhances the AI’s credibility. Users are more likely to trust and engage with an AI that appears to possess specialized knowledge.
Deeper Understanding: The chain-of-thought reasoning approach not only ensures accurate responses but also provides users with a deeper understanding of the topics at hand. This promotes knowledge transfer and fosters a more meaningful user experience.
Comprehensive Responses: With the ability to provide infinite-length responses, ExpertAI can address even the most complex questions exhaustively. Users receive thorough explanations, eliminating the need for follow-up queries.
Adaptive and User-Focused: ExpertAI is highly adaptable and user-focused. It respects user instructions, making interactions more intuitive and efficient. Users have the freedom to guide the AI in line with their specific requirements.
In conclusion, the new system message prompt for ExpertAI represents a significant leap forward in AI-driven question-answering. Its unique techniques and performance benefits empower users to access accurate, credible, and comprehensive information with ease. As AI continues to shape our digital landscape, this innovation underscores our commitment to enhancing user experiences and fostering a more informed and engaged user community. Thank you for your attention, and I invite you to explore the capabilities of ExpertAI in your own AI endeavors.
Tried on GPT-4, and get even less satisfaction. Even getting Dr. Edward Witten and Kip Thorne to answer my questions about black hole entropy calculations and significance results in very brief answers of around 500 tokens including the spelled-out thought steps, better followed.