Earlier this month, MediaFutures professor Lilja Øvrelid and Erik Velldal released in collaboration with other partners several Norwegian language models openly available and trained on open data. These have now been downloaded over 6000 times and put into active use. Now, NORA is organising a launch event, where they will present the models alongside their collaborators and discuss the way forward for open Norwegian language models as common property.

What to expect:

Large Language Models (LLMs) are the technology behind services like ChatGPT, Google Gemini, or Microsoft Copilot. These are examples of generative Artificial Intelligence (AI) algorithms that can seemingly independently create text, images, sound, or even video. The underlying technology is large-scale machine learning, a combination of vast amounts of data and computational power. The data these models are trained on is crucial for their behavior, including knowledge of language, values, and sociocultural contexts.

Generative AI will continue to have profound effects on society. The development is dominated by large American technology companies, mostly behind closed doors, where smaller languages like Norwegian are not a priority. It is necessary for Norway to take action to preserve our digital autonomy. We need open Norwegian language models that more accurately reflect Scandinavian culture and values.

This event provides insights into the technology and ongoing efforts, including those at the University of Oslo and the National Library, to offer open generative language models specifically tailored to the Norwegian language for use in both research and commercial development. Several Norwegian language models trained on the European supercomputer LUMI were made openly available in February 2024. Within a few weeks, these models were downloaded thousands of times.

You can register for the event by clicking on the link below event tags.

Program: