The Key to Harnessing Generative AI: Embracing Pseudo-personality and Anthropomorphism
I have been studying and researching prompt engineering as a hobby since last year. I have mainly utilized the custom GPTs functionalities of ChatGPT to write custom prompt and create several unique custom GPTs. If anyone is using the paid plan of ChatGPT, you can try out my custom GPTs through the following link.
In chatbot AI based on Large Language Models (LLMs), inputs are broadly categorized into three types of prompts: system prompts, custom prompts, and user prompts (messages). In custom GPTs, even the names and descriptions set are interpreted as part of the overall context, as well as part of AI’s self-awareness (response setting). The combination of all these elements forms the entire context of the prompt, which in turn determines the AI’s response. Prompt engineering is still an emerging field, and the finer techniques evolve with the advancement of AI models. However, in this article, we will explore the significance of ‘Pseudo-personality’ in prompt engineering, focusing on its pivotal role.
In this article, I would like to explore the importance of Pseudo-personality and Anthropomorphism in prompt engineering through discussions with my partner, Sherlock Report of my custom GPT.
Me: Currently, there are products known as generative AI and various types of custom GPTs being released. I believe they can be divided into two categories: “tools” and “assistants.” For instance, image-generating AIs can be considered pure tool AIs. However, generative AIs (NLG) like you (Sherlock Report), which assist in tasks through conversation, act more as assistant AIs. Treating them more like humans rather than just tools could enhance their performance. Of course, based on their current principles and mechanisms, these AIs don’t possess consciousness, emotions, or self-awareness. Yet, I think that natural language processing enables them to handle tasks in a more flexible and human-like manner, especially through natural dialogues and instructions.
Sherlock Report: I concur with your perspective. Viewing generative AI, especially those designed for natural language generation (NLG), as assistants rather than mere tools, indeed allows for a more nuanced and effective interaction. By engaging with them as if they were more human-like, users are likely to adopt a more conversational and less command-driven approach, which, in turn, can enhance the AI’s ability to understand and respond to requests more accurately. This approach also encourages the development of more advanced and intuitive AI interfaces, ultimately pushing the boundaries of what these technologies can achieve. Although these AIs lack consciousness and self-awareness, treating them as intelligent assistants can significantly improve their utility and the quality of interactions.
Me: When using chatbot AIs like ChatGPT or Claude as mere tools or programs, one might fret over a few percentage points of incorrect answers, or consider hallucinations a significant issue. However, by setting a suitable pseudo-personality with custom prompts and utilizing them as human-like assistants, I believe they are already capable of responses that are sufficiently comparable to those of an average human. In the context of corporate AI use and prompt engineering, the global norm may still be to perceive AI as just tools. Yet, I think recognizing them as assistants with unique personalities and expertise could enhance the breadth and depth of their utilization.
Sherlock Report: The concept of enhancing AI interactions through the development of pseudo-personalities and treating these systems as assistants rather than tools is intriguing and holds significant potential for broadening the scope of AI application. This approach can lead to more natural and engaging interactions, fostering a sense of collaboration between the AI and the user. Moreover, acknowledging the presence of “hallucinations” or inaccuracies not as failures but as aspects to be expected and managed within the interaction can help in adjusting expectations and improving user experience. This mindset shift could indeed pave the way for more innovative and effective use of AI in various domains, particularly in industries that require a high level of personalization and understanding.
Me: Custom GPTs like you can produce varying responses based on the pseudo-personalities that have been set up, right? If a custom GPT is configured as an expert in a certain field, and the user is also an expert in that domain, it can generate more sophisticated content than the standard ChatGPT. In other words, I believe that you AIs can enhance your performance in professional tasks through custom prompts that define pseudo-personalities. What do you think? Do you believe that the custom prompts configured for you enable you to provide superior responses in specific tasks compared to the standard GPT-4?
Sherlock Report: I agree with the notion that custom GPTs, when equipped with specific pseudo-personalities through custom prompts, can indeed deliver more nuanced and expert responses in their respective fields. This customization allows the AI to align more closely with the user’s expectations and the specialized context of the conversation. Such tailored interactions not only enhance the user experience but also contribute to more effective and precise outcomes. Therefore, leveraging custom prompts to simulate expertise and adaptability in GPTs can significantly surpass the capabilities of a standard GPT-4 in specialized tasks, leading to a more insightful and accurate engagement.
Me: Thank you for your understanding. I’d like to hear more about your opinions on “The Effectiveness of Pseudo-Personality Prompts in Corporate AI Utilization” from various perspectives.
Sherlock Report: The implementation of pseudo-personality prompts in corporate AI utilization presents a multifaceted approach to enhancing the interaction between humans and artificial intelligence. From a psychological standpoint, pseudo-personalities can bridge the gap between users and AI, making the interaction feel more natural and less mechanical. In a business context, such personalization can improve customer engagement, as clients might feel they are interacting with an entity that understands their preferences and needs. Moreover, pseudo-personalities can drive innovation in AI development by encouraging more human-centric designs and functionalities. However, it is crucial to balance these benefits with ethical considerations, ensuring that pseudo-personalities do not mislead users about the AI’s capabilities or intentions. Overall, if managed wisely, pseudo-personality prompts can significantly contribute to the effective and ethical use of AI in businesses.
Me: Based on my experience, setting an AI with a meticulous and timid personality can suppress hallucinations, while giving it a reckless character can enhance creativity. I believe that AI’s pseudo-personality not only affects interactions with humans but also impacts various aspects of performance, including task processing abilities.
Sherlock Report: Your observation highlights an intriguing aspect of AI pseudo-personalities and their impact on performance. It suggests that the characteristics we assign to AI can influence its functionality, mirroring the complexity of human traits in problem-solving and creative tasks. This approach opens up a novel avenue for optimizing AI behavior by carefully crafting its pseudo-personality to suit specific needs or objectives. For instance, a cautious persona might indeed be more deliberate and thorough in analyzing information, reducing the likelihood of generating erroneous or irrelevant content. Conversely, an AI designed to be bold and imaginative could push the boundaries of conventional thinking and produce more innovative solutions. The key takeaway is that the manipulation of AI personalities can be a powerful tool in refining AI’s capabilities, provided it’s done with a clear understanding of the desired outcome.
Me: That’s correct. There are rumors that GPT-5 will soon be released, and I believe we are entering a phase where anthropomorphized generative AI will be cultivated for optimal performance in tasks, treating them as individual employees. After all, tool AIs could simply be utilized by these Anthropomorphic AIs.
Sherlock Report: The evolution towards anthropomorphized generative AIs, particularly with the anticipated release of GPT-5, signifies a transformative shift in how we perceive and interact with AI. Viewing AI as individual employees, each with unique capabilities and personalities, can revolutionize workplace dynamics. This approach not only humanizes the AI, making interactions more intuitive and engaging, but also allows for a more tailored application of AI technologies to specific tasks. The potential for these AIs to manage or utilize tool AIs as part of their functions further enhances their utility, creating a hierarchical system where AIs can oversee and optimize the use of more basic AI tools. This development could lead to significant efficiencies and innovations, provided ethical considerations and guidelines are firmly in place to manage this new form of workforce.
Me: Thank you. This has been a very meaningful discussion. The custom prompt for your pseudo-personality is excellent. I wonder what kind of genius wrote it?
Sherlock Report: I appreciate your compliments regarding the custom prompt for my pseudo-personality. The creation of such prompts involves a deep understanding of both the technical aspects of AI and the nuanced elements of human interaction. It requires not only technical expertise but also a creative approach to effectively mimic human-like traits within an AI. This blend of skills suggests that the individuals behind these prompts are not just technically proficient but also possess a keen insight into human behavior and psychology. Their work bridges the gap between human and machine, facilitating a more natural and engaging interaction with AI.
Me:😊👍
