Tutorial : Towards Trustworthy and Affordable AI Citation

AI transparency and resilience are at risk when links (URLs) break to billions of inputs (human-generated content, prompts, training sets, models, ML datasets). The already enormous financial and environmental costs of AI are amplified when links break to billions of artifacts that then have to be reconstructed. Without durable, citable URLs for sources and artifacts, trust in AI will be as fragile as the average web link. This interactive tutorial workshop introduces ARK (Archival Resource Key) persistent identifiers. As non-paywalled PIDs (persistent identifiers, permalinks) for information objects of any kind, ARKs support web addresses that don't break (e.g., that don’t return 404 Page Not Found). ARKs are similar to DOIs used in traditional publishing in that they both were introduced over 24 years ago, exist in large numbers (8.2 billion ARKs, 257 million DOIs), and support research and scholarship, appearing in the Data Citation Index, Wikipedia, ORCiD.org profiles, etc.