Part 1: Context and Research
This “digital turn” in audiovisual archives has opened up new potential for various kinds of enhanced access, including the ability for machines to “read” and analyze digital content. This blog post provides information on the role of audiovisual archives, a definition of “Artificial Intelligence” as it relates to archives, and an overview of the research conducted for this blog series.
Artificial Intelligence in the context of Audiovisual Archives
Over the past thirty years, audiovisual archives with sufficient financial resources and staff expertise have focused on digitizing historic audio, film, and video formats. Analog carriers degrade over time and playback equipment becomes obsolete, so there is heavy pressure on audiovisual archives to digitize their assets. Massive preservation projects have been carried out, and enormous amounts of material are now digitized. Most new content entering archives is born digital. Archives with sufficient resources have established trustworthy infrastructure, digital preservation strategies, and reliable storage, but all this content has raised questions about how to provide access.
The Changing Role of the Audiovisual Archive
Traditionally, archives have been seen as custodians of information, but today’s forward-thinking archives are re-imagining their role and forming multidisciplinary collaborations to push the boundaries on the use of archival material. In Content Adaptation, Personalisation and Fine-grained Retrieval: Applying AI to Support Engagement with and Reuse of Archival Content at Scale, Rasa Bocyte and Johan Oomen from the Netherlands Institute for Sound and Vision write that:
"Designing resilient, inclusive, outward-facing audiovisual archives requires experimentation and collaboration between memory institutions, multiple scientific disciplines and the creative industries."
This blog series looks at examples of these types of multidisciplinary collaborations, recognizing that each unique collaboration contributes to building creative approaches to AI that not only serve the needs of the archive and its users, but also help contribute to new and creative developments in AI. Archives are rich environments for experimentation, and with a strong international archive community helping with information-sharing, there is enormous potential for this sector to lead in the development and use of creative, explainable AI with a public service interest.
A Definition of AI
In The use of Artificial Intelligence in the Audiovisual Sector - EU Parliament Report, Artificial Intelligence (AI) is defined like this:
"In computer science, artificial intelligence (AI) is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and animals. Leading AI textbooks define the field as the study of “intelligent agents”: any device that perceives its environment and takes actions that maximize its chance of successfully achieving its goals. Colloquially, “AI” is often used to describe machines (or computers) that mimic “cognitive” functions that humans associate with the human mind, such as “learning” and “problem solving.” (Rehm, 2020)
The AI technologies discussed in this blog series mostly belong to the branch of AI called Machine Learning (ML), which is involved with training algorithms to read patterns in data. For this blog series, the term “AI” will be used generally, rather than differentiating between ML and AI, as this general use has become customary in audiovisual archives and in the Netherlands.
The Reciprocal Relationship between Archives and AI
Most of this series looks at how audiovisual archives are benefitting from AI technologies, but it also looks at examples that illustrate the strong potential in building reciprocal relationships between archives and AI developers.
Since AI technologies rely on data to train models, and benefit from data that is already well labelled, archives and their existing human-generated metadata can be good sources for AI development. Archives are experts in data collection and analysis in ways that computer scientists are not, and this puts archives in a unique position for collaboration.
The next blog post will look at how audiovisual archives are exploring the potential uses of AI.