Artificial Intelligence: Creating new opportunities for archives
Denis Leconte
Issue: March/April 2024

Artificial Intelligence: Creating new opportunities for archives

As industries of all kinds grapple with how to put artificial intelligence to work, you’d be hard pressed to find a better immediate fit for the emerging technology than the archiving business. 

At Iron Mountain Media and Archive Services, where millions of assets in every imaginable form are securely stored for the world’s most famous studios, artists and brands, AI hasn’t just accelerated work previously done by humans — it’s opened new realms of possibility for these vast historical and cultural treasure troves. 

Since the division was launched in 1988, Iron Mountain Media and Archive Services has become the go-to physical and digital media archiving solution for the biggest names in entertainment and media. But the vast scale of those archives has historically made it challenging to know exactly what was in each collection — and you can’t retrieve and use what you can’t find. 

With innovative new digital tools for intaking and indexing assets with much greater efficiency, that is changing. The recent development of a mobile scanning device (the Automated Media Image Capture System, or AMICS) and AI-powered Smart Vault gives Iron Mountain the power to greatly accelerate the process of cataloging and retrieval. With these tools, assets are infinitely more discoverable, accessible, searchable — and as a result, more monetizable. 

As director of technology and later VP of technology, my team began experimenting in the mid 2010s with machine learning. In those early days, we were putting early AI applications, like facial recognition, to some use. These new tools were useful, but required a fair amount of tuning and were fairly error-prone. It wasn’t always easy to get early AI to do what you wanted. 

The technology began to get better around 2017, and progress and the quality of the results greatly accelerated after that. Classification tools became far more accurate and dependable — but asset classification results were still lists of keywords and phrases, much easier for a search engine to zero in on a given asset, but not particularly readable. But now, with the emergence of generative AI, it becomes possible to use a layer of AI with some understanding of language to summarize the keyword-based knowledge coming from classification into plain language descriptions — and potentially summaries of descriptions in the case of combined assets (a video with many cuts for example, where each cut gets summarized and then the overall video is also summarized). This allows an unprecedented level of interaction with the asset search space, through a very approachable experience, leading to much improved asset discoverability. One of the important benefits of recent AI development is the ability to produce the semantic environment for this sort of experience automatically, something that was only possible before through painstaking, time-consuming and specialized human labor. 

But to get to the point where that was possible, the intake process also needed an overhaul. Until very recently, every asset had to be photographed, categorized, organized and tagged by hand — just to be searchable in the first place. Now, can you hire enough specialists to complete that task on a very large archive? It’s certainly possible, but the cost is often prohibitive — which is why it has not been done for many archives.
That’s where AMICS comes in. The Automated Media Image Capture System is our “R2-D2-like” scanning device that’s mobile and nimble enough to be easily moved around Iron Mountain’s vast physical archives. It’s designed to simultaneously photograph all sides of any given asset and process the images to extract metadata. AMICS can recognize and interpret everything from barcodes and handwritten notes to the manufacturer of cardboard storage boxes themselves. 

All of this metadata is then fed into Smart Vault, where this representation of the physical asset can be associated with the digital instance of the asset — a video file, for example, in the case of a videotape — and the metadata sets coming from both instances can be married to provide a more holistic view of the asset, thereby further facilitating its accessibility. 

On the user end, imagination becomes the only limit. AMICS and Smart Vault have brought us to a place where no level of detail or hint of knowledge about an asset is impossible to search for. 

The ability to search through your whole archive, in as natural and detailed a way as possible, has always been the Holy Grail for the archiving industry — and now, thanks to AI, it’s becoming possible. At Iron Mountain Media and Archive Services, where we still say,  “It’s only Rock ‘n’ Roll if you can find it,” we’re making quick progress in that direction.   

Denis Leconte is the Vice President of Technology at Iron Mountain Media and Archive Services (