Carrie Bickner

Editorial Note: This document has been updated and incorporated into the MARTI Framework master document, which provides a comprehensive overview of the latest developments. For the most current and complete information about MARTI, please visit: https://carriebickner.com/marti/

I used a chat bot to develop MARTI, a metadata framework for managing human/generative AI output. 

MARTI is a metadata framework for generative AI/human output.

It is not complete, but it is at a point where it would benefit from the input of authors, artists, programmers, researchers, scientists and, above all, people who create LLMs and the tools that exploit them.

To begin, I am going to share two documents that introduce the core principles, purposes and goals, and define the criteria and functional requirements for the framework.

• MARTI Framework Overview
  Provides an introduction to MARTI, outlining its core values, purpose, and intended applications.
• Metadata Requirements Outline
  Defines the criteria for metadata within MARTI, covering task-based and suite/environment metadata and emphasizing preservation and provenance.

Over the next several weeks I will publish technical documentation, including practical guides and technical standards, and implementation recommendations outlining the following:

• Contribution Table and Schema Document
  Defines the structure of the contribution table, illustrating how MARTI organizes roles and names for consistent documentation of contributions.
• Output-Focused Metadata Rubric
  Provides a rubric with quality standards for metadata specific to generative AI outputs.
• Metadata Generation and Voluntariness
  Addresses the distinction between voluntarily created and automatically generated metadata, outlining intellectual responsibility.
• Instance and Variation Delineation Document

Elaborates on metadata requirements for differentiating instances and variations.

• Quality Control and Compliance Standards
  Establishes standards for metadata integrity, consistency, and compliance, ensuring MARTI’s quality assurance and ethical guidelines.

This work is not complete; it is intended to serve as a conversation starter and call for collaboration.

MARTI is an open framework. It is meant to be used and adapted across the whole range of human and generative AI collaboration from creating a single, static image to dynamic and evolving LLMs.

The MARTI framework is designed to integrate with existing metadata standards, such as Dublin Core, the C2PA Content Credentials standard or any standards that come from the W3C, enhancing interoperability and ensuring consistency across diverse metadata ecosystems. By aligning MARTI with these widely adopted frameworks, we aim to facilitate broader adoption and compatibility within digital asset management, preservation, and ethical Al documentation. (Bickner et al. MARTI Framework Overview (Version 1.1)),

I look forward to any and all comments.