Izvajalec / Course provider: Jožef Stefan Institute (JSI)
Predavatelji / Instructors: Marko Debeljak (JSI)
Content: The event aims to provide agricultural stakeholders with a clear, realistic, and application-oriented understanding of large language models (LLMs) and their potential role in Slovenian agriculture. Participants will be introduced to the rationale for focusing on high-quality, domain-specific data and benchmarks rather than individual model versions, and to the role of the national AI infrastructure (SLAIF) in supporting domain- and language-specific AI solutions. A key objective is to foster informed expectations, trust, and early engagement of future users by openly addressing both the capabilities and limitations of current AI technologies in regulated and practice-oriented agricultural contexts.
The event presents current activities related to the development and application of large language models within the Slovenian agricultural domain in the context of SLAIF. It introduces the fundamentals of LLMs in an accessible manner, with a focus on their applicability, risks, and constraints in agriculture. Participants will be presented with a proposed taxonomy of Slovenian agriculture, including key sectors and subdomains, along with example benchmark tasks designed to evaluate model performance in Slovenian. Preliminary benchmark results will be demonstrated to illustrate how existing models perform on selected agricultural use cases. The event further provides a structured forum for validating assumptions about users, use cases, and sectoral coverage, and introduces a dedicated web application for the collection of high-quality question–answer pairs and other expert inputs to support model adaptation and evaluation.
After the event, participants will understand the importance of domain-specific data, benchmarks, and expert validation for the sustainable development of AI solutions in agriculture. They will have a realistic understanding of what current large language models can and cannot do in the Slovenian agricultural context, and how national AI infrastructure supports their adaptation. Participants will be able to critically assess proposed use cases and benchmarks from a practitioner’s perspective and will be equipped to contribute expert knowledge, feedback, and structured data inputs for the continued development and evaluation of AI tools for Slovenian agriculture.