Leslie Johnston (Director of Digital Preservation at the US National Archives and Records Administration) spoke at the second Unlocking our Digital Past workshop on 16 September 2021. Leslie’s talk focused on incremental approaches to AI and highlighting the machine learning tools we already use without considering them as ML.
The abstract for Leslie’s talk read as: “The U.S. National Archives and Records Administration (NARA) has recently been looking at how Artificial Intelligence and Machine Learning might be applied to both processing and public access, consulting with colleagues from several academic and cultural heritage organizations on the state of the art in applied AI; NARA has several medium- and long-term goals. based in part on the need to work more efficiently and improve access at a very large scale. But there are actually short-term goals that NARA or any organization can potentially meet with machine learning tools: every project doesn’t need to be about full-fledged AI. Implementation of machine learning as a subset of AI is much more feasible, and organizations may already be using such tools without realizing it: Optical Character Recognition for printed text or handwriting, chatbots, Alexa skills, pattern recognition or pattern matching (used to find PII or other relevant patterns in records), named entity recognition (personal or corporate names, geographic locations, and dates), or computer vision for still images or video such as the controversial use of facial recognition. This talk is focused on the lower barrier to entry types of tools that cultural organizations can use to start their use of AI for processing and access.”
You can view Leslie’s presentation below: