Abstract

This presentation aims to highlight the critical role of open-source and locally run Large Language Models (LLMs) for European universities. The target audience includes a broad spectrum of European research supporters with limited knowledge of AI. The main objectives are to emphasize the importance of independence from commercial offers, the benefits of data safety, and the ability to tailor tools to specific needs.

Open-source LLMs are crucial for increased security and allow universities to leverage their in-house expertise to create tools that surpass commercial offerings. The main benefit of using open-source LLMs locally is data safety. Case studies include the development of drafting and writing support tools to streamline grant applications.

Challenges such as computational limitations are being addressed by the industry with specialized processors and model size reduction techniques. Future directions recommend the creation of a common open-source platform for European universities to improve tool development, reduce costs, and enhance research support workflows. This approach will also free up time for scientists to focus on their research.