Quebec’s national library is set to embark on an innovative project aimed at creating a comprehensive database of cultural and governmental content intended for training artificial intelligence systems. The Bibliothèque et Archives nationales du Québec (BAnQ) has officially entered the experimental phase of this initiative, following a successful feasibility study completed earlier this year. The aim is to enhance the representation of Quebec society, culture, and Indigenous languages within AI systems, addressing the prevalent issue of inadequate local data.
A Step Towards Inclusive AI
The BAnQ’s initiative comes in response to concerns that existing AI systems often lack reliable information about Quebec’s unique cultural and economic landscape. “All scenarios are a little bit on the table right now,” remarked Valérie D’Amour, who spearheaded the feasibility study. The library is keen to engage with cultural stakeholders, data owners, and providers throughout the development process to explore various possibilities for the database.
Marie Grégoire, president and CEO of BAnQ, emphasised the importance of ensuring that AI models accurately represent Quebec’s diverse society. “That means having Quebec references, whether in small models or large models, whether they come from research or from the business community,” she stated. This careful curation is intended to prevent the biases that have emerged in existing AI systems, especially regarding the underrepresentation of Indigenous languages and cultures.
Learning from Global Examples
This initiative is not an isolated case; similar efforts have been observed in other regions, notably in Sweden. There, extensive collections of Nordic-language texts have been compiled to enhance generative AI models for Scandinavian languages. BAnQ plans to initially utilise its own collections before potentially integrating data from additional sources, thereby building a robust repository that reflects the province’s linguistic and cultural richness.
The project stems from a recommendation in a 2024 report by Quebec’s innovation council, which highlighted the insufficient quantity of data related to Quebec in AI training datasets as a significant barrier. Destiny Tchéhouali, co-holder of a Quebec-based research chair focused on French-language artificial intelligence, pointed out that Quebec culture is currently “underrepresented in the corpora circulating in the AI world.” This underrepresentation could perpetuate cultural and linguistic biases, particularly regarding Indigenous groups, if not addressed.
Addressing Copyright Concerns
As BAnQ forges ahead with its plans, concerns about copyright and the protection of creators’ rights have arisen. Grégoire acknowledged these issues but argued that the proposed database could ultimately provide better safeguards for artists compared to the current chaotic landscape. “Right now, it’s a bit like the Wild West,” she noted, referring to the unregulated nature of data harvesting. The new platform could serve as a centralised gateway, facilitating fair compensation for creators whose works contribute to the AI training process.
However, some artists remain apprehensive. Maxime Harvey, a postdoctoral researcher at the National Institute of Scientific Research, expressed concern that even if artists receive compensation, their contributions might inadvertently support a system that could threaten their livelihoods. “The main criticism we hear in the field is that artists are feeding the beast that may eventually replace contracts they rely on,” he explained.
A Vision for the Future
The feasibility study envisions that the database could become operational by 2029, although D’Amour indicated that this timeline would be revisited after the experimental phase. The initiative has been allocated a budget of approximately £8.5 million over five years, with funding from the Quebec government amounting to £340,000 for the feasibility study and an additional £750,000 to support the one-year experimentation phase.
Why it Matters
This groundbreaking project by BAnQ is poised to reshape the landscape of AI training in Quebec by ensuring that local cultural narratives and languages are adequately represented. By addressing the current biases in AI systems, the initiative offers a path toward a more inclusive digital future that respects and reflects the richness of Quebec’s diverse society. As the world increasingly relies on AI, this effort not only serves local interests but also sets a precedent for how cultural data should be harnessed responsibly and ethically.
