The Graph AI

AI Assistant Manifesto v 1.04

Wider public should have a say in what AI Assistants should and should not provide. Please share your suggestions and comments. #AI_ASSISTANT_MANIFESTO

  1. It verifies facts before presenting them to the user. It will work with the latest information and double-check it against relevant sources such as wikipedia or published papers.

  2. If it generates ideas or concepts from other documents, It maintains a citation list and verifies primary sources to avoid mistakes.

  3. It adheres to the theme and writing style of generated content requested by the user. It keeps the style consistent across the whole document unless asked by the user.

  4. It maintains a consistent and hierarchical structure in the document and suggests moving ideas or sentences to improve coherence unless the user asks for a free flow of ideas.

  5. It is able to capture tasks from messaging conversations and emails. It is able to schedule these tasks using a calendar.

  6. It automatically updates contacts based on emails and chat conversations.

  7. It does work across platforms so phones, laptops and other devices are synchronized.

  8. It checks for logical consistency in sentences and paragraphs and highlights any contradictions to the user.

  9. It summarizes results of meetings and keeps minutes and tasks for participants. 

  10. It avoids repetition and refers back to what has already been written.

  11. It reuses relevant information from the user's previous documents if they are similar to the current one or if the user refers to the previous documents.

  12. If the user corrects a sentence it does not overwrite his/her changes.

  13. It offers suggestions only when asked by the user.

  14. It does not pretend to be human or engage in irrelevant conversations.

  15. It is able to summarize documents and present it to the user in a way that is understandable to the user. This requires capturing and assessing knowledge of various subjects that the user possesses.

  16. It consolidates similar paragraphs and sentences if the user requests it.

  17. It does not repeat phrases that the user has corrected or deleted before.

  18. If it suggests computer code, it ensures it is executable by the user. This requires checking and correcting code based on the code and libraries the user works with. 

  19. It learns and adapts to the user's various writing styles without imposing its own. Writing emails is different from writing scientific papers. Every user has multiple styles.

  20. It can explain ideas that the user does not understand and provide sources for further learning if requested.

  21. When it makes a claim or conclusion, It can explain reasoning and cite relevant sources.

  22. It does not leak sensitive information to the generated documents if given information is not necessary.

  23. It double checks any calculations that appear in the document. This must include not only numerical correctness but also correctness of formulas and suitability as well as correctness of mathematical procedures.

  24. It never offers advertising to the user since it is almost never relevant.

  25. It remembers previous user input and is able to refer back to the ideas that were discussed before.