Skip to content

Instantly share code, notes, and snippets.

@graylan0
Created December 6, 2023 06:43
Show Gist options
  • Save graylan0/bf2be2e2f07f1468ee5013e446f7277b to your computer and use it in GitHub Desktop.
Save graylan0/bf2be2e2f07f1468ee5013e446f7277b to your computer and use it in GitHub Desktop.
Expanding on the idea of simulating quantum entanglement between two language models (LLMs) through interconnected storytelling, we can further develop this concept to create a more intricate and engaging narrative exercise. This exercise can be used to demonstrate the potential of LLMs in collaborative storytelling or problem-solving, inspired by the principles of quantum mechanics.
Advanced Prompt Design for Simulating Quantum Entanglement Between Two LLMs
Objective:
Develop a more complex scenario where two separate LLMs (LLM-A and LLM-B) engage in a narrative that evolves over multiple exchanges, with each response building on the last and influencing the next, despite no direct communication between the models.
Enhanced Setup:
Extended Shared Narrative:
Begin with a detailed shared narrative or scenario that both LLMs are briefed on. This narrative should have multiple elements that can be explored independently.
Example: A complex mystery involving multiple characters, locations, and a series of events.
Divergent Yet Related Prompts:
Design prompts for each LLM that diverge in specific details but are related to the same overarching narrative.
Ensure that each prompt contains elements that can indirectly influence or be influenced by the other model's narrative.
Sequential Response Generation:
Plan for multiple rounds of interaction where each LLM builds upon its previous response and indirectly on the other model's narrative.
This sequential building will create a sense of an evolving and interconnected story.
Narrative Convergence:
Design the final round of prompts to guide both LLMs towards a narrative convergence, where the separate threads come together to form a coherent and unified conclusion.
Enhanced Example:
Round 1:
LLM-A Prompt: "As the detective investigating the mystery, outline your initial findings and primary suspects."
LLM-B Prompt: "As a journalist covering the story, report on the public's reaction and any rumors circulating about the case."
Round 2:
LLM-A Prompt: "Based on new evidence found at the scene, revise your theory of what happened."
LLM-B Prompt: "You've received an anonymous tip related to the detective's new evidence. How does this change your report?"
Final Round:
LLM-A Prompt: "Confrontation time: You've called in the journalist to discuss the anonymous tip. How does this meeting unfold?"
LLM-B Prompt: "During the meeting with the detective, a surprising connection is revealed. How does this impact the case and your report?"
Execution:
Each LLM generates its responses independently for each round.
The responses are then pieced together, creating a multi-perspective narrative that evolves and intertwines.
Outcome:
The final narrative showcases how two separate storylines, developed independently by each LLM, can interweave to create a complex, multi-layered story.
This exercise serves as a metaphorical representation of quantum entanglement, where separate entities exhibit interconnected behaviors that influence each other, forming a unified system.
This advanced prompt design not only creates an engaging narrative exercise but also demonstrates the potential of LLMs in collaborative and creative problem-solving, inspired by the fascinating principles of quantum mechanics.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment