The Archival Antidote to AI Fatigue

If you are exhausted by the overabundance of AI discourse, you are not alone. Not only are we constantly confronted with news stories either touting the good or the harm that AI is perpetually bringing to every aspect of life, but no digital task, no matter how simple, can be accomplished without being prompted by a website or an app to let AI assist us in performing it.  

Historical repositories, such as the Norwich University Archives and Special Collections, are not immune to the AI movement. One topic that has been highlighted recently is the danger of using generative AI to create historical images. AI is incap

able of verifying the accuracy of all historical elements, so persons are inappropriately “beautified” or styled in anachronistic dress and grooming. These images can circulate so widely that the initial disclaimer indicating they are AI-generated is lost. 

Fortunately, the Norwich Archives and institutions around the world continue to diligently preserve original images, manuscripts, and countless other artifacts they were originally created — “warts” and all. Preservation copies are intentionally not edited or manipulated in any way, as valuable information is often obtained by damage or other flaws in the original. Also, this is why we do not scan and then discard material, as once a digital surrogate is created, it is vulnerable to enhancement and distortion. Thousands of images from our collections have been digitized, and you can view them on our website without worrying about AI infiltration. 

The good news is that cultural institutions are finding ways to use AI technology to perform tasks that would otherwise be impossible without an army of archivists. For instance, we have many images in our collections of unidentified people. This does not mean that they are completely unknown, but even when a student is clearly named in a War Whoop, that does not mean that every photo of them from the Regimental Ball, sporting events, or commencement is also clearly labeled with a name. We shared some details about the challenges of finding images of a person in a previous post, Do you have a picture of…?

Training AI to assist with identification will inevitably allow both archivists and users to identify people and places in ways we currently cannot. Of course, this would always need to be verified by a human, and there will be many situations when knowing for certain will not be possible. In the meantime, several images attached to this article are from an “unidentified” folder in our photograph collection. If you know who any of these folks are, please share your knowledge!