Research project P4S/251/PASTFORWARD-AI (Research action P4S)
PASTFORWARD.AI is a KIK-IRPA research initiative that strategically integrates open-source artificial intelligence into the cultural heritage sector. Rather than developing new AI systems from scratch or relying on large commercial platforms, both of which come with high costs, significant skill requirements, and data privacy concerns, the project focuses on identifying, evaluating, and applying readily available, adaptable open-source models to enhance KIK-IRPA's capabilities and services.
The initiative targets five concrete application areas across the institution's three core departments: advanced reverse image search for visual similarity, Retrieval Augmented Generation (RAG) for intuitive document querying, metadata enrichment through image recognition, automatic translation for multilingual access, and advanced optical character recognition for digitised texts. Together, these tools aim to significantly improve the findability, accessibility, interoperability, and reusability of KIK-IRPA's collections, particularly through its BALaT platform, while automating routine tasks and strengthening research data management.
The project unfolds in three phases. It begins by establishing governance structures, data and ethical frameworks, and a global survey of relevant open-source AI models. This foundation leads into an experimentation phase, where suitable models are selected through departmental consultation and tested via iterative proofs-of-concept built on KIK-IRPA's own data. The project concludes by evaluating these pilots, consolidating lessons learned, and translating findings into a strategic AI integration roadmap for the institution.
Responsible deployment is central to the project's identity. An internal AI Ethics Assessment Framework and clear usage policies are developed as core deliverables rather than as an afterthought, ensuring that every application of AI is transparent and institution-wide. Staff workshops on AI literacy and change readiness support this commitment from the inside.
Externally, PASTFORWARD.AI actively engages communities such as AI4LAM, sharing progress, gathering feedback, and aligning with emerging best practices. All code and findings will be published openly via a project website and public repositories. The expected outcomes, including a long-term AI roadmap, enhanced institutional knowledge, and a suite of tested and scalable proofs-of-concept, are designed to demonstrate how responsibly applied open-source AI can preserve, analyse, enrich, and disseminate cultural heritage, in direct support of KIK-IRPA's ambition as a digital hub.
