12.30 - 12.35 h | Question time
12.35 – 13.05 h | Impossible ideas, feasible projects
Pep Casals, co-founder of Nubilum & Coeli Platform
How can we turn a priori impossible ideas into real projects? This presentation uses a set of practical cases to explore the strategies to meet the needs and increase the efficiency of teams in the heritage sector, highlighting how the implementation of artificial intelligence (AI) tools facilitates these successes and emphasising the application of professional judgement.
13.05 – 13.10 h | Question time
13.10 – 15.00 h | LUNCH BREAK (free time)
15.00 – 15.20 h | Through the mirror: an analysis of the heritage content created by GenAI
Xavier Rubio Campillo, Ramón y Cajal researcher at the Universitat de Barcelona and Director of the DIDPATRI (Didactics and Heritage) research group
15.20 – 15.40 h | Use of generative AI in public administrations: audio guides
J. Ignasi Bonet, innovation facilitator at CTTI
Generative AI has recently reached the whole of society. Initial tests have raised great expectations. Given that the results are apparently spectacular, an inevitable question arises: how can I use these tools? But they need to be tested before they can be used in public services.
One notable test that has been carried out is that of the automatic generation of audio guides. The aim is to evaluate the feasibility of an automatic solution that generates versions by language or audience, as well as generating the narration, in order to streamline the creation of audio guides at an affordable cost.
15.40 – 16.00 h | Small data, big models and the rewards of reasoning
Maria Cristina Marinescu, Professor at IQS and researcher at the National Supercomputing Center.
Areas that do not have a lot of data by AI standards, as is the case with cultural heritage, have generally been left out of the main guidelines for applied research. With the advent of large language models, a powerful tool for text generation has become available to everyone, such as for image descriptions. Do these models make other efforts to provide high quality metadata for cultural heritage images obsolete, or are there tasks where they can still be improved? What alternatives can complement the statistical approach on which these models are based for small data domains? We will illustrate our view in this regard through the lessons learned during our European project: Saint George on a Bike.
16.00 – 16.20 h | Didactic iconography in history and generative AI
Francesc Xavier Hernàndez, Professor of Social Science Teaching at the Universitat de Barcelona
Iconography is a key resource for approaching past in a comprehensive manner. The use of Computed-Generated Imagery (CGI) has made it possible to obtain images of high quality and at reasonable cost (3D, matte painting...). Generative AI now offers interesting additions to provide views of the past.
16.20 – 16.30 h | Question time
16.30 | Conclusion
Jordi Medina, Head of the Mediation Area of the Catalan Cultural Heritage Agency.