Machine learning for business users

We know it from ChatGPT: Even users without technical expertise can elicit passable answers from a language model on the spot; for optimal results, however, more experience is needed. There is already talk of a new core competence called "prompt engineering". This is precisely where the INPROMPT project comes in, which the DFG now intends to support with a grant in kind. INPROMPT stands for "Interactive Prompt Optimization with Humans in the Loop for the Development and Intervention of Natural Language Understanding Models."

The goal of the project: an assistance system for users without technical training in machine learning. They are to create well-performing prompts using automatic optimization techniques. The prompts are modified iteratively; an objective function estimates the quality of the adjusted inputs. Factors included are: automatic predictions on text instances, the readability of the prompt, and the conclusiveness of a decision-making explanation.

Applicant Prof. Roman Klinger explains, "We expect our project to significantly improve the transparency of prompt-based models and contribute to the democratization of the use of machine learning."

The INPROMPT project is expected to start in April 2024.