Language technologies such as voice assistants or chatbots already support many people and companies in making their everyday lives more efficient. Through the use of artificial intelligence (AI), machine processing of natural language now functions at a very high level. Language models can already take over qualitatively demanding tasks such as creating newspaper reports or conducting customer dialogues. In order to open up the possibilities of innovation for the European market and especially for small and medium-sized enterprises, a large and powerful language model for Europe is being created in the OpenGPT-X research project.
With the new language AI, intelligent speech applications are to be created by application partners that significantly improve existing solutions and offer companies room for new products and business models. In addition to innovation and competitiveness, OpenGPT-X is also designed to preserve data privacy. The OpenGPT-X speech applications will be made available via the European cloud initiative Gaia-X, which is currently being created according to European values and security standards, also with the participation of DFKI. In future, the new language AI will be made available in the three Gaia-X domains of mobility, finance and media.
Due to the high technical requirements in terms of computing power, high-performance language models can so far only be implemented by large companies or consortia. For this reason, the project consortium includes many renowned partners: While Fraunhofer IAIS will develop the AI models together with DFKI and the companies Aleph Alpha and Alexander Thamm GmbH, the Jülich Research Centre and the Technical University Dresden will provide their high-performance computing capacities. Together with 1&1 IONOS SE, the DFKI will oversee the integration into the Gaia-X infrastructure. The project will receive domain-specific data for the development of the models from ControlExpert and WDR, among others, who will also develop use cases for the project.
For the DFKI research area Speech and Language Technology, the OpenGPT-X project complements the activities already underway at the European level, which focus on language equality in the digital space and the development of an infrastructure for the multilingual Europe: While the European Language Grid (ELG) as a platform for language technologies and companies promotes exchange in the sector both on an economic and research level, the European Language Equality (ELE) project is developing an agenda whose goal is to achieve digital language equity in Europe - by 2030 at the latest. Digital language equality means that all European languages should be supported by language technology to a comparable extent.
Prof. Dr. Georg Rehm, project leader of OpenGPT-X at DFKI: "The development of large language models on other continents in recent years has proven the potential that language AI holds for applications in everyday life. Due to the diversity of languages and the high demand for data protection, the development of large language models for Europe is not only a special but also an urgent challenge and at the same time a great opportunity for the European community. We are pleased to contribute our long experience in research on language technologies in Europe to OpenGPT-X."
The OpenGPT-X project started in January 2022 and will run for three years. The total budget is €19 million and the funding volume is €14.9 million.
The OpenGPT-X consortium
Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. (Fraunhofer IAIS and Fraunhofer IIS; consortium leadership), Aleph Alpha GmbH, Alexander Thamm GmbH, ControlExpert GmbH, Deutsches Forschungszentrum für Künstliche Intelligenz GmbH, Forschungszentrum Jülich GmbH, KI Bundesverband e.V., Technische Universität Dresden, Westdeutscher Rundfunk Köln Anstalt des öffentlichen Rechts, IONOS SE