André Clements
Abstract: This paper presents a method of adjusting the style and tone of conversations with the OpenAI’s ChatGPT AI model by using an initial prompt template. The prompt serves as a parameter for the conversation, which can be adjusted with desired parameters such as Formality, Complexity, Referencing, Examples, and others. This approach, known as lens-mode, is a simulation that influences the context representation and decoding of ChatGPT’s output. The results of this approach are dependent on the parameters provided and are influenced by the state of the ChatGPT AI model at the time of the conversation. The method has been tested on the 2023 Jan 9 and Jan 30 versions of the chat.openai.com platform with promising results. However, some caveats should be considered, such as the constantly changing nature of the platform and the simulation’s reliance on the parameters provided.
Introduction: The advancement of natural language processing and generation has led to the development of AI models capable of conducting conversations, such as the OpenAI’s ChatGPT. However, the default response style and tone of these models may not always be ideal for a given scenario. This paper presents an experimental approach to addressing this issue by providing a way to adjust the style and tone of conversations with ChatGPT through an initial prompt template known as lens-mode.
Methodology: The lens-mode approach involves using an initial prompt template that serves as a parameter for the conversation. The prompt includes instructions on how to start a lens-mode session, adjust parameters, and exit the session. The parameters, indicated by “–” in front of the parameter name, allow the user to adjust aspects of the conversation such as formality, complexity, referencing, and examples. The chatbot acknowledges the prompt and starts a lens-mode session, during which it attempts to remember and apply the parameters to its context representation and decoding.
Results: The results of this approach are dependent on the parameters provided and the state of the ChatGPT AI model at the time of the conversation. The lens-mode approach has been tested on the 2023 Jan 9 and Jan 30 versions of the chat.openai.com platform and has shown promising results. However, it is important to keep in mind that this approach is only a simulation and that the conversation’s context representation and decoding are influenced by the parameters provided.
Caveats: The chat.openai.com platform is constantly being modified and adjusted, so the results of the lens-mode approach may vary. Additionally, this approach is only a simulation and is dependent on the parameters provided, with no guarantee of the veracity of the responses generated. Finally, a conversation tends to build momentum, which may also influence the character and veracity of the responses.
Conclusion: In conclusion, the lens-mode approach provides a way to adjust the style and tone of conversations with ChatGPT by using an initial prompt template as a parameter. While the approach has shown promising results, it is important to keep in mind that it is only a simulation and that the results are dependent on the parameters provided and the state of the ChatGPT AI model. Further research is needed to fully understand the capabilities and limitations of this approach.
PDF version
