“Perfect is the enemy of good. We want to be as good as possible but don't expect to be perfect”, emphasised Fredrik Heintz.
Danila Petrelli, Senior Data Manager at AI Sweden, presented how TrustLLM’s data strategy has evolved:
“The TrustLLM data strategy shouldn't be seen as a finished solution, but as a structured response to real constraints. We have built it as a path or a process, not an endpoint.”
Ethics-by-design approach
Laurynas Adomaitis from Department of Computer Science at RISE Research Institutes of Sweden introduced an ethics-by-design approach, positioning ethics as a process of ongoing dialogue. With tools for the method, developers are encouraged to reflect on and address potential ethical issues while developing the systems. The tool facilitates a dialogue which leads to internal insights and changes in the design. He stressed the complexity of ethical questions in AI, which often invoke deeper values such as justice, freedom, and autonomy.
A panel discussion featuring Fredrik Heintz, Laurynas Adomaitis, Peter Strömbäck (External Ethics Advisory Board), and Billy Jörgensen (AI Sweden), moderated by Tohid Ardeshiri (TrustLLM project manager), concluded the seminar. Together, they explored the roles of regulation, responsibility, and cross-sector collaboration in building trustworthy AI.
Successful examples and courage
"All successful forward leaps in LLM development involved an innovation. We need to learn from these successful examples worldwide and most importantly have the courage to innovate ourselves in our data strategy and data usage within the European ethical and legal framework", Tohid Ardeshiri urges.
