#25 Meetup Women in Machine Learning & Data Science at WMI UAM

For this special event, our invited speakers will come from Warsaw to share their experience with WiMLDS Poznań community!

Zuza Kwiatkowska - On the technical side of the AI force for nearly 10 years. For the past 2 years, she has been translating from scientific to human through social media, public speaking, and trainings. Creatively fueled by ADHD. [LinkedIn]

Inez Okulska - Editor-in-Chief of "hAI Magazine" - the first printed magazine on artificial intelligence in Poland, and VP of Research at CampusAI. Researcher and creator of algorithms, scientist, currently senior research engineer in the area of ​​natural language processing (NLP) at the Wrocław University of Science and Technology. By education, she is both a humanist (Polish philology, comparative studies) and an engineer (automation and robotics). She used to visit Harvard. Author of numerous speeches and lectures (including on TedX and I love Marketing), where she talks about non-human intelligence in human language. Many times distinguished in prestigious rankings, including Top100 Women in AI and Top100 Women in Data Science in PL. [LinkedIn]

Agenda (start at 6:00 pm)

Event opening
Talks

1. "The Solution to the Deepfake Problem Might Already Exist" - Zuza Kwiatkowska

Deepfake technology, which can be used to create fake digital media, has become a major issue that is deeply concerning us all. The tools created by scientists give people the power to undermine someone's reputation or make false accusations with just a few clicks.

But is the problem really with the technology itself and the scientists who developed it? Or should we be looking more at ourselves, the users of these digital illusions?

Interestingly, the solution to the deepfake problem may already be out there. However, accepting and implementing this solution could be an uncomfortable and difficult challenge for us as a society.

2. “When RAG Evaluation Goes Rouge: Metrics, Dataset Preparation and Other Headaches” - Inez Okulska

Evaluating Retrieval-Augmented Generation (RAG) systems often feels like navigating a labyrinth of unmet expectations, absent benchmarks, and overlooked nuances. While much of the discourse focuses on building RAG-based solutions, the real challenges emerge when trying to measure their effectiveness.

Constructing custom evaluation datasets is no small feat—especially in languages like Polish, where source material is often poorly digitized. Drawing from my own experience, I will offer some food for thought on the questions of what and how to evaluate in RAG-based systems, or their individual components, such as information retrieval and text generation.

Networking
Pizza time by GFT

WiMLDS Code of Conduct
https://github.com/WiMLDS/starter-kit/wiki/Code-of-conduct

More details soon. Stay tuned!