New episode of ADAPT Radio’s ‘HumanAIse’ focuses on Recovering History with AI

03 August 2023

ADAPT Radio’s HumanAIse series continues this month with a discussion on utilising Artificial Intelligence to recover historical documents with Senior Researcher at the Virtual Treasury of Ireland, Trinity College Dublin, Dr. David Brown, and Managing Director of Transkribus developer Read-Coop, Andy Stauder  

Conversations about AI are often future focused, but these emerging technologies can also help us bring the past to life. In this month’s ADAPT Radio Podcast, discover how AI has been used in Beyond 2022, a flagship research project led by Trinity College Dublin which will digitally recreate seven centuries of historical records of the Public Record Office of Ireland destroyed by fire at Dublin’s Four Courts at the beginning of the Irish Civil War. Joining the podcast to discuss are Senior Researcher at the Virtual Treasury of Ireland of Trinity College Dublin, Dr. David Brown, and Managing Director of Transkribus developer Read-Coop, Andy Stauder

Kicking off the podcast, Dr. David Brown speaks on the drive of the Virtual Record Treasury to restore historical records that were destroyed in the fire at the Public Records Office in Dublin during the Civil War. The Beyond 2022: Ireland’s Virtual Record Treasury research project is an exciting all-island and international collaborative project. It seeks to re-imagine and re-create, through virtual reality, this national treasure and the archival collections that were lost, comprising records of seven centuries of Irish history, genealogy and administration. Dr. Brown described the inclusion of Transkribus as serendipitous as they were developing exactly the software the Virtual Record Treasury needed to advance in their work. 

Stauder also provides an overview of how the transcribing technology works by utilising AI models to create searchable text from images of documents much faster than you would without this technology. For instance, according to Dr. Stauber, the type of transcribing they undertake would take 6 months without the technology, with it the same task can be done in a few days. As well as this, AI models can read handwriting we may not recognise with the human eye. Stauder explains how Transkribus processes and understands different styles of handwriting from different people and eras. According to Dr. Brown, at least 200,000 pages were processed throughout the project, most of it handwritten, that ranged in date from the 1280s up to the 1920s.

Other potential applications of this kind of historical research focused technology can be seen in more modern scientific applications on a much wider scale. For example many medical documents were produced on paper which impacts how easy they are to distribute. However, utilising this type of technology in these cases would allow easier and much more far reaching sharing of scientific or medical documents which could aid in research. One specific example Stabur relates to is in the case of the Spanish flu pandemic. Finding and researching records from this time period may inform future pandemic research. Finally, Stauber highlights the importance of future proofing data accessibility and addressing challenges of digital obsolescence and fragility as well as offering possible solutions.

For further insight into the interactions of humans and AI, catch HumanAIse on SoundCloud, iTunes, Spotify, and Google Podcasts.

ADAPT Radio: HumanAIse is ADAPT’s newest podcast series providing an in-depth look at the future of AI, automation and the implications of entrusting machines with our most sensitive information and decisions.