What I can tell you is that MN AI is built on multiple AI models including ChatGPT, but we have designed unique features for it—for example, it can add cards to your mind map and adjust the order of cards. As for more detailed privacy-related questions, I need to coordinate further with our product and R&D teams to provide a thorough response.
The number of AI Points you consume depends on how you use the AI features. For example, if you ask the AI to provide a short answer, it will only use a small amount of Points. However, if you request the AI to understand your approximately 300-page textbook and generate a mind map, this will consume a larger number of Points.
Factors such as your prompt content, the materials you provide to the AI, and the mode you select will all affect the total number of Points consumed.
1. Which AI is it? (Custom, ChatGPT, Gemini, or other?)
MarginNote 4’s AI functionality is not a single model. Instead, it is a platform that integrates multiple, advanced large language models (LLMs).
The key points are:
Multiple Models: The system provides access to several different AI models, each optimized for specific tasks and scenarios.
User Choice: You have the flexibility to choose which model to use for each interaction based on your needs (e.g., speed, depth of reasoning, cost).
Model Types: According to the documentation, the available models include:
均衡 (Balanced): For general-purpose, balanced performance.
轻量 (Lightweight): A fast and economical model for simple tasks.
文本 (Text): Optimized for advanced text reasoning.
深度 (Depth): Engages in deeper reasoning and can show its “thinking process.”
视觉 (Visual): A multi-modal model that can understand and analyze images you send to it.
工具 (Tool): Can call MarginNote’s built-in tools to edit cards or optimize mind map structures on your behalf.
In summary, MarginNote has built an AI platform that connects to various state-of-the-art models (which could include technologies like GPT-4, Claude, or others, though the specific providers are not named in the documentation), giving you a powerful and flexible toolkit.
2. Privacy and Data Access: Does it access my documents without my consent?
No, the AI does not access your documents, notes, or any of your data unless you explicitly initiate an interaction.
Your privacy and control are fundamental principles of the design. Here’s how it works and how you control what the AI can see:
Explicit Consent Required: The AI only processes information that you actively send to it. This is done through very clear, intentional actions:
Dragging Content: You drag a card, a PDF selection, or text into the AI dialog box.
Using a Template: You select text or a card and choose a preset or custom template to ask a question.
Using a Tool: You use a specific AI tool (like “AI Convert Handwriting to Text” or “AI Generate Table of Contents”) that by its nature requires access to the selected content.
You Control the Context: You have complete control over the “context” you give the AI for a conversation. For example, if you drag three specific cards into the chat, the AI will only use the content of those three cards to answer your question. It cannot access other cards or documents in your study set that you did not provide.
No Background Scanning: There is no indication in the documentation that the AI scans, analyzes, or uploads your content in the background without your direct command. Its access is strictly limited to the tasks you ask it to perform.
Summary
Technology: MarginNote 4 uses a multi-model AI platform, allowing you to select the best model for your task (Balanced, Lightweight, Deep, Visual, etc.).
Privacy: The system operates on a principle of explicit user consent. The AI only sees and processes the data you deliberately send to it via dragging, selecting, or using a specific AI tool. Your private documents and notes remain entirely private until you choose to use them with an AI feature.
I hope this detailed explanation helps clarify how the AI functions in MarginNote 4.