Keynote 1:
Nils Reimers, VP AI & Search, Cohere

Title:
What happens when your users are not longer humans?

Abstract:
For the past 80 years, humans where the users of search systems. The systems had to be fast, precise in the returned information, and understandable for humans. With the rise of LLMs, we see a drastic increase of AI-powered search queries for applications like RAG, Deep Research and AI Agents. What will change, when no longer humans are the primary users of a search system, but AI systems? Join me for this talk to get an overview of state-of-the-art agentic AI search, and a look into the future how search systems need to evolve.

Bio:
Nils Reimers has been in the “AI-powered search” domain since 2019 with the publication of the “Sentence-BERT” paper, showing that transformer networks are capable to learn suitable representation for search. Since then, he led many research projects in that area, including the creation of the BEIR- and the MTEB-benchmarks. Since 2022 he works as VP AI & Search for Cohere.com, where his teams develop advanced retrieval models that power search applications for tens of millions of users.

Keynote 2:
Martin Potthast, hessian.AI Chair for Deep Semantic Learning at the University of Kassel

Title:
Who writes the web? Who reads it? Who judges it? And who reports back to us?
About authenticity in information retrieval.

Abstract:
Retrieval-augmented generation (RAG) leverages LLMs to automatically find relevant documents and to provide users with direct answers to their queries. Supplying retrieved documents deemed to be relevant during answer generation reduces LLM “hallucinations.” However, as generative AI proliferates, a RAG system will increasingly find documents that have been generated as well. Although they can be useful, we observe that LLMs favor generated over authentic content during retrieval, whereas we show that detecting LLM-generated content proves to be inherently more difficult than expected. Likewise, citing a source for a generated statement makes it verifiable. But it often remains unclear which sources have been discarded in favor of the cited one. We demonstrate this kind of framing in the context of product search. Direct answers often narrow the user’s perspective of the available diversity of choices. The sincerity of a RAG system must therefore be judged by the authenticity of its answers.

Bio:
Martin Potthast holds the hessian.AI Chair for Deep Semantic Learning at the University of Kassel. His research focuses on language technologies, search engines and the analysis and synthesis of information. Martin contributes to the research areas of information retrieval, natural language processing, and artificial intelligence. Martin Potthast studied computer science at the University of Paderborn (2001-2006) and successfully completed his doctorate at the Bauhaus-Universität Weimar in December 2011, where he then worked as a postdoctoral researcher at the Digital Bauhaus Lab. In October 2017 he was appointed junior professor at Leipzig University and in October 2023 full professor, thus successfully completing his tenure track. Martin is a founding member and board member of the ScaDS.AI Center for Scalable Data Analytics and Artificial Intelligence. In April 2024, he was appointed Professor of Deep Semantic Learning at the University of Kassel and joined the Hessian Center for Artificial Intelligence hessian.AI.