Angelie Kraft (M.Sc.)

Angelie Kraft (M.Sc.)

Doctoral candidate at University of Hamburg

Talks and presentations

Introduction to LLMs

April 13, 2024

Talk, Yearly meeting of the Women in IT group of the German Informatics Society, Heidelberg, Germany

I was kindly invited by the Women in IT group of the German Informatics Society to give an introductory talk on the topic of LLMs at their yearly meeting. I explained the basic principles of what LLMs are and how they are created and briefly explained how RLHF, tool use, and RAG are used to tune LLMs towards more helpful assistants.
More importantly, I explained how factual inaccuracies happen (why they are not actually “hallucinations”) and why training on the whole internet does not equal training on the whole truth.

Unpacking Chat-GPT

July 05, 2023

Talk, Airbus Buxtehude, Buxtehude, Germany

I was invited to give a talk about the inner workings of ChatGPT, as well as its strengths, limitations, and ethical concerns. More details about our lab visit to Airbus can be found here.

The Ethical Risks of Analyzing Crisis Events on Social Media with Machine Learning

July 06, 2022

Talk, International Workshop on Data-driven Resilience Research 2022 (D2R2'22), Leipzig, Germany

I was invited to present our work “The Ethical Risks of Analyzing Crisis Events on Social Media with Machine Learning” at the D2R2’22 workshop in Leipzig. Although I sadly could not give the talk due to sickness, my slides can be found here.

Virtueller Forschungsabend: Neue Trends aus der KI-Forschung

November 04, 2021

Talk, Technologie- und Innovationszentrum Gießen, Online

At this meetup, different AI researchers gave a non-expert audience insights into current trends in our field. I gave an overview of important problems and approaches regarding social bias in large multimodal and language models. Watch a recording of the event here.

Expo 2021 - Student project exhibition

October 13, 2021

Poster presentation, Informatics Department, University of Hamburg, Hamburg, Germany

I presented a poster on my Master’s thesis “Triggering Models: Measuring and Mitigating Bias in German Language Generation” at the yearly EXPO of the Informatics Department at the University of Hamburg. I won “Best student project 2021”.

Bias in Künstlicher Intelligenz und Machine Learning

September 02, 2021

Interview, InstaLive with inovex GmbH, Online

In this casual InstaLive, Lukas (inovex), Eva, and I talk about algorithmic bias, its harms and what can be done to counter it. While I address the topic of bias in large language models, Eva gives some insights into biased recommender systems.

Webinar: Einsatzmöglichkeiten und Potenziale großer KI-Sprachmodelle

July 01, 2021

Talk, eco - Verband der Internetwirtschaft, Online

In this webinar organized and hosted by eco - Association of the Internet Industry, I was invited alongside two other speakers (including Jörg Bienert, chair of the German AI Association) to talk about the potentials and possible applications of large-scale language models like GPT-3. In my talk, I addressed the issue of reproduced social biases and stereotypes and practical implications.

Über Bias - Projektvorstellung justAI

September 11, 2020

Interview, Social Developers Club, Hamburg, Germany

The Social Developers Club interviewed me about algorithmic bias and its relevance to society. This interview was part of their project proposal for their workshop series justAI which later won the “IDEA Preis 2020”. Their goal is to provide introductory AI courses for kids and teenagers with an emphasis on responsible AI.