#podcast #T5E10 15 julio, 2025
In this episode, we talk with Arantxa Urrea, a trained statistician and PhD student at the DaSCI Institute, about a crucial yet lesser-known topic in the development of artificial intelligence systems: out-of-distribution data (OOD).
During the conversation, Arantxa explained:
🧠 What OOD data is:
These are data that differ from what the model has seen during training. They can appear in real-world situations where the environment changes or unexpected cases arise.
🚨 Why it’s a problem:
AI models often assume that the data they receive in production are similar to the training data. When this isn’t the case, they may make incorrect decisions with high confidence.
🛠️ How to detect and manage them:
Arantxa shared current approaches to identifying these data and strategies to make models more robust and aware of their own uncertainty.
With a clear and outreach-oriented approach, the episode highlights the importance of designing AI systems that are not only accurate but also aware of their limitations.
Arantxa Urrea is employed thanks to the Strategic Project “Ethical, Responsible, and General-Purpose Artificial Intelligence for Cybersecurity” (IAFER-Cib, Ref.: C074/23), a result of the collaboration agreement between the National Cybersecurity Institute (INCIBE) and the University of Granada. This initiative is part of the Recovery, Transformation and Resilience Plan, funded by the European Union – Next Generation EU.
a se lleva a cabo en el marco de los fondos del Plan de Recuperación, Transformación y Resiliencia, financiado por la Unión Europea-Next Generation EU.
Don’t miss a thing and share your thoughts and reflections with us on social media (@sintonia_dasci).
https://www.spreaker.com/episode/episodio-48-conociendo-a-l-s-doctorand-s-la-ia-que-la-lia–66983408
