Digdeep Podcast: Why Does AI Have an Incomplete Picture of Our World, Wiebke Hutiri?

In this episode of #digdeep speeks Wiebke Hutiri, Senior Research Scientist at Sony.

The podcast describes how AI systems make decisions about people every day, but often rely on unbalanced training data collected without consent.

Many population groups are underrepresented in these datasets, which means they are less accurately recognized or considered by the models. This problem is fundamental and has real-world consequences, for example, in facial recognition, self-driving cars, and voice assistants. Wiebke Hutiri works on responsible AI and, with her team, has published a diverse, ethically collected image dataset from 81 countries. She emphasizes that AI development should be more strongly driven by societal goals and explains how human diversity (“Long Tail”) leads to errors in AI systems.

Wiebke Hutiri is a Senior Research Scientist at Sony AI and leads an international team for Responsible AI and Ethics. Her team develops and standardizes methods for ethical data collection, fair model evaluation, and trustworthy AI systems.

Just listen in (podcast in German).