4. September 2024
At the beginning of September 2024, a delegation of the National Science and Technology Council (NSTC) visited Germany. ScaDS.AI Dresden/Leipzig was honored to meet the NSTC delegation twice:
On 03.09.2024, the NSTC-BMBF seminar on AI took place in Berlin. Focal points of the seminar were Generative AI and Large Language Models and AI Edge Computing. Besides the NSTC delegation with around 20 participants, the seminar was joined by respresentatives of the BMBF as well as German AI Scholars. State Secretary Dr. Roland Philippi (BMBF) and Minister Cheng-Wen Wu (NSTC) opened the seminar. The seminar was then divided into two thematic sections.
In Section I of the program, the seminar focused on Generative AI and Large Language Models (LLMs). Prof. Jane Yung-Jen Hsu (Vice Chair, Taiwan AICoE) hosted this session.
Prof. Simon Razniewski, AI Professor at ScaDS.AI Dresden/Leipzig, presented the topic Bridging cultures with LLMs: Distilling and utilizing multicultural commonsense knowledge. He reviewed the limitations of LLMs that are trained on predominantly English text corpora. He also showed how explicit cultural commonsense knowledge can offer a way forward. Furthermore, he presented methodologies to distill such knowledge at quality and scale, from very large text corpora and LLMs.
The topic of the following Section II was AI Edge Computing. The host was Prof. Wolfgang Wahlster (Founding Director and CEA, DFKI).
On the next day, the NSTC delegation, including minister Cheng-Wen Wu, visited the Living Lab in Dresden. After a short introduction round, Dr. René Jäkel (Administration Director) gave an overview on ScaDS.AI Dresden/Leipzig, its mission, research activities, and interdisciplinary team. Hosts and guests quickly identified various overlaps between the German and Taiwanese research interests.
Afterwards, Dr. Siavash Ghiasvand guided the delegation through the Living Lab, showcasing the broad variety of demonstrators and the close collaboration between Leipzig and Dresden. Detailed information on the demonstrator Content Based Image Retrieval for Historical Images was provided by Dr. Taras Lazariv. Norman Koch finished the presentation section by demonstrating how machine learning in the browser works by using asanAI. The visit ended with a lively discussion on the functionalities of the demonstrators and possibilities for collaboration.