Feature Request: MindMap Integration for Enhanced ChatGPT Experience

Dear OpenAI ChatGPT Team,

I hope this message finds you well. As a machine learning engineer who’s loves ChatGPT, I’ve been inspired to share an idea that I believe would significantly improve the user experience.

Problem Encountered:
While using ChatGPT, the linear conversation format can hinder the potential for more complex interactions. In real life, when discussing issues with others, I often turn to mind map editors (such as XMind) to break down core problems and track how we arrive at branching ideas. Tree-like narratives and a “divide and conquer” approach are more representative of our problem-solving processes. By creating a visual mind map, we can grasp the bigger picture and avoid getting lost in minor details. Additionally, as we switch between branches, it helps us quickly recall the context of each idea and how we arrived at that point. In summary, mind maps facilitate better thinking in a nonlinear way, and we can explore and expand upon ideas by asking questions.

It would be fantastic if ChatGPT conversations could take place within a mind map! I often face two dilemmas while using ChatGPT:

  1. After receiving an answer, I have multiple follow-up questions (Q1, Q2, etc.). When I choose to ask one (Q1) and get a response, the new answer may spark more questions. After several rounds, it becomes challenging to recall the intriguing ideas that came to mind after the initial answer (Q2, etc.). As Bret Victor put it, ideas are precious and fleeting.
  2. Even if I remember my other questions about the first answer, I face a difficult choice: should I edit Q1 to Q2 or should I add Q2 at the bottom of the conversation? In the former case, resubmitting Q1 as Q2 would delete any sparking questions that originated from Q1, while in the latter, the context of the conversation becomes unreasonable, as Q2 would appear disconnected from its proper context.

These problems wouldn’t exist in a mind map, where Q2 could be introduced as a new branch, allowing the conversation to continue seamlessly. Given the nature of GPT, this should be entirely feasible! Just as we can edit Q1 to Q2 now, OpenAI would only need to record the embeddings for each branch—although this might pose a significant storage challenge.

I sincerely hope this feature can become a reality, even if only for a select group of Plus users. Drawing upon Doug Engelbart’s vision of computers “augmenting human intellect”, I’m confident that even in its simplest form, this feature would revolutionize human-AI interactions.

This letter has been edited by GPT-4