Skip to main content

Full AI profile: https://spdx.github.io/spdx-spec/v3.0.1/model/AI/AI/

https://spdx.dev/learn/areas-of-interest/ai/

Hugging face paper@ https://huggingface.co/blog/model-cards

IBM paper: https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/factsheets-model-inventory.html?context=cpdaas

Description (ChatGPT summary)

The AI Bill of Materials (AI BOM) initiative, as defined in the SPDX 3.0.1 specification, aims to provide a standardized framework for documenting and sharing comprehensive details about AI systems and models. This initiative introduces the AIPackage class, which serves as a container for metadata related to AI artifacts. Key properties of the AIPackage include:

  • Autonomy Type: Specifies the level of autonomy exhibited by the AI system.
  • Domain: Indicates the specific field or industry where the AI model is applied.
  • Energy Consumption: Details the energy usage during various stages of the AI model's lifecycle, such as training and inference.
  • Hyperparameters: Lists the parameters that were set prior to the training process and are crucial for model performance.
  • Model Explainability: Provides information on how the decisions made by the AI model can be interpreted and understood.
  • Safety Risk Assessment: Documents any potential safety risks associated with the AI system and the measures taken to mitigate them.
  • Standard Compliance: Indicates adherence to relevant standards and regulations applicable to the AI system.
  • Type of Model: Describes the architecture or category of the AI model, such as neural networks or decision trees.

By incorporating these properties, the AI BOM facilitates transparency, accountability, and reproducibility in AI deployments, ensuring that all stakeholders have access to essential information about AI systems.