Lamarr Researchers Share Expertise on Large AI Models
Large Artificial Intelligence (AI) models have great disruptive potential. Researchers of the Lamarr Institute and its partner organizations have long-standing and broad-ranging experiences with the development and application of the underlying AI technologies. At the LEAM conference initiated by the German AI Association, a study on large AI models for Germany was presented to the public.
Currently, the development and application of Artificial Intelligence is undergoing a dramatic transformation. The recent success of ChatGPT, a chatbot which can create eloquent answers and writes natural-sounding and detail-rich texts, is only the tip of the iceberg of this exciting new development. Behind the success of ChatGPT, which is short for “Chat Generative Pre-trained Transformer”, is GPT-3, a large AI model which deals with natural language processing (NLP).
The LEAM Initiative and Conference
The Lamarr Institute’s scientists have extensive experience in NLP research. Lamarr Principal Investigator Dr. Joachim Köhler, head of the Lamarr Institute’s applied research group on NLP, is also advocating the initiative LEAM, which stands for Large European AI Models and is initiated by the German AI Association (KI Bundesverband). LEAM brings together representatives from businesses and research institutions with the aim of building a competitive European AI ecosystem fostering the development of large AI models.
On January 24, Dr. Joachim Köhler participated in a panel discussion at the LEAM conference which took place in Berlin. The panelists analyzed the requirements for software and hardware to build large AI models. Another panel discussion on the role of language models and how they can be useful in everyday life featured Dr. Nicolas Flores-Herr who is head of Conversational AI at the Fraunhofer IAIS, one of the partner organizations of the Lamarr Institute.
In several other keynote speeches and panel discussions, the LEAM conference aimed at giving an overview of the impressive development of large AI models, shedding light on the potential and benefits for stakeholders from research and industry, and giving an outlook on necessary steps to implement such AI models in Germany.
Study on Large AI Models for Germany
In the context of the LEAM conference, the results of the feasibility study “Große KI-Modelle für Deutschland” (“Large AI Models for Germany”), commissioned by the Federal Ministry for Economic Affairs and Climate Action, was presented. One of the authors of the study is Dr. Gerhard Paaß, senior AI researcher at the Lamarr Institute.
The study analyzes the disruptive potential of large AI models, especially so-called AI foundation models, which are flexible, reusable AI models which have been trained on large amounts and a broad range of data. Currently, the United States and China are in a leading position with respect to the development and practical implementation of such large AI models. The authors of the study examine the demand and potential use cases for AI foundation models especially in Germany and Europe. They explain how joint action of the federal state, science and economy could contribute to building an AI ecosystem, including suitable high-performance computing facilities, which meets the growing demand. Joining forces for the development of large AI models in Germany will lay the groundwork for upholding ethical standards, strengthening the competitive position of the European economy, and ensuring digital sovereignty.
The LEAM initiative
Download the study “Große KI-Modelle für Deutschland” (in German)