How can an organization move from experiments with artificial intelligence to truly useful, safe, and reliable applications? This question is also the subject of inquiry at the GPT-Lab at the University of Tampere, which was introduced to the partners of the CAMPUS project on the second day of the professional program in Tampere.
The research laboratory established in 2023 deals with the relationship between generative artificial intelligence and software development. Its research areas include, among others, the further development of large language models, applications based on organizational knowledge bases, AI agents, and responsible and trustworthy artificial intelligence.
First, the real need must be understood
One of the most important messages of the presentation is that an AI development should start from a specific organizational problem, not from the currently popular technology.
The researchers at GPT-Lab first explore the need together with the corporate or institutional partner, then examine the available research results and determine the applicable method. Only then is the prototype or the smallest working solution created.
The development is first tested with the original partner, and later, if possible, with the involvement of additional organizations. In this way, not only is it determined whether the technology works, but also whether it is capable of creating value in a real environment.
Research in direct corporate collaboration
The cooperation between companies, researchers and students plays an important role in the operation of the laboratory. The companies bring real-world problems, the researchers provide methodological and technological knowledge, and the students can even be involved in developing solutions as part of their diploma work.
This model is also remarkable for Central Transdanubia. Businesses in the region – especially small and medium-sized enterprises – often do not have the opportunity to independently build the full expert and technological background required for artificial intelligence.
Sharing research and testing capacities, building relationships with higher education institutions, and implementing smaller, controlled experimental projects can reduce the risks of implementing AI.
Trust is not an ex post check
A demonstrably working prototype is not necessarily ready for operational use. Before deployment, the quality of the data, information security, the reliability of the results, the possibility of human oversight, and the responsibilities must be examined.
The example of GPT-Lab therefore serves not only as a technological lesson, but also as a regional innovation lesson. Accessing the latest tools is not enough to achieve digital resilience. We must also be able to critically evaluate, safely test, and responsibly apply them.
The KDRIÜ can serve as a regional facilitator in helping to connect corporate needs, research competencies, student knowledge, and technology providers. Long-term competitiveness can be based on regional capabilities built on recognizing problems, experimenting, and learning together, rather than on a single AI tool.
Read the detailed professional article!
You can read more about the operation of GPT-Lab and the experiences that can be used for the Central Transdanubia region in our English-language LinkedIn article.