28.05.2026
Zeynep Akata: To Trust AI, We Need to Understand What Goes on Behind the Scenes
Research Film
How can we build AI systems that are transparent enough to be trusted with decisions that affect our health and daily lives?
«You can think of a foundation model like a mirror to society. If the mirror is distorted, then the image that comes out is going to be distorted too.»
Zeynep Akata
MCML PI
MCML PI Zeynep Akata explains why bias in foundation models is a critical challenge and why explainability should guide model development rather than just auditing it. By identifying which model components are responsible for flawed decisions and adjusting them, researchers can steer AI toward more trustworthy behaviour.
In this video series, discover how MCML researchers are bridging theory and practice to build AI systems that prioritize accessibility, sovereignty, and impact.
©MCML
The film was produced and edited by Nicole Huminski and Nikolai Huber.
Related
29.10.2026
Teaching Language Models to Express Uncertainty
An ICLR 2026 paper shows how reinforcement learning improves confidence estimates from language models.
27.10.2026
Runner-Up Prize at AutoML 2026 for Tabular-DC-Bench
Tabular-DC-Bench, by MCML Junior Member Mamdouh Aljoud and MCML Director Thomas Seidl, among others, won a runner-up prize at AutoML 2026.
