Model Cards
Standardized documentation for ML models describing training, capabilities, limitations, bias analyses, and recommended use cases.
Model Cards becoming compliance requirement (EU AI Act). Marketing should review them for used models: Does the model fit the use case? What risks exist?
Explanation
Model Cards contain: Model details (architecture, training data), intended use, out-of-scope use, bias & fairness analyses, performance metrics, limitations, ethical considerations. Introduced by Google 2019.
Marketing Relevance
Model Cards becoming compliance requirement (EU AI Act). Marketing should review them for used models: Does the model fit the use case? What risks exist?
Example
Hugging Face shows model cards for all hosted models: Llama-2-70B card explains RedPajama training, benchmark scores, known bias problems.
Common Pitfalls
Model cards often incomplete or outdated. No standardization of content. Bias tests not always relevant for specific use case.
Origin & History
Model Cards is an established concept in the field of Artificial Intelligence. The concept has evolved alongside the growing importance of AI and data-driven methods.