Introduction
Ingrid Falk once watched a student celebrate a neural network curve that was almost too smooth. She asked where the measurement gap had gone. The room fell quiet, and then the student noticed that the model had politely hidden the missing data.
Ingrid lives in Heidelberg and mentors young researchers. She began her scientific career when computation meant punched cards and patience, long before deep learning acquired its current shine.
Story of the Path into AI
After retirement Ingrid grew irritated by the phrase “completely new.” Much of AI looked new, but some questions were old: measurement error, model assumptions, overfitting, the temptation to believe a beautiful curve. She did not want to become a nostalgic critic, so she learned the tools.
She read lecture notes, rebuilt small neural networks and compared them with classical physical models. At first she was slower than the students and had to accept that they could implement certain methods faster. Her strength lay elsewhere: asking what a model was allowed to ignore. Her first seminar format asked students to compare AI results with simple physical baselines and to mark uncertainty before presenting performance.
The first cohort found the exercise irritating. They wanted to show progress. Ingrid insisted that a baseline is not an insult; it is a conversation partner.
Current Work
Today Ingrid volunteers as a mentor and writes short essays on scientific humility in AI. In one climate-data exercise she showed how a complex model covered an obvious measurement gap with smooth predictions. The group learned to visualize uncertainty instead of hiding it in elegance.
Her mentoring helps young researchers connect modern tools with methodological discipline. Ingrid is not against complexity. She is against forgetting that every model is a calculation with assumptions, and that assumptions become dangerous when they stop being visible.
Personal Advice
“A model is not an oracle; it is a calculation with prerequisites,” Ingrid says. Her advice to younger researchers is to keep simple comparisons close. They do not make the work smaller. They keep it honest.
Key Facts
Age and place: 80, Heidelberg.
Background: physics, retirement, bridge between generations.
Entry into AI: seminar format comparing AI results with physical baselines.
Focus today: research methodology and mentoring.
Typical tools: statistics, model comparison, scientific mentoring.
Werkstattnotiz
Ingrid has one slide that contains only a blank axis and the question: “What did we not measure?” Students remember it more often than the equations. She is now collecting examples in which uncertainty was present in the lab notebook but disappeared before the final chart.