The development of the European speech model

ControlExpert is part of the OpenGPT-X consortium: The development of the European speech model OpenGPT-X has launched

The project aims to develop a big AI speech model for Europe. “International competitors have already understood the enormous disruptive potential for AI speech technologies towards the economy, industry and society as a whole. A European AI speech model such as OpenGPT-X is therefore absolutely necessary to ensure the digital sovereignty and market independence for Europe”, says Dr. Nicolas Flores-Herr, project leader at Fraunhofer IAIS. For the development, special emphasis lies on data protection and European language diversity.

So called big AI speech models such as the American GPT-3 or Chinese Wu Dao 2.0 use enormous data amounts for various AI models which for example write journalistic texts, support in customer care and help to understand texts. The problem: Data protection and the integration of European languages are not guaranteed. “With OpenGPT-X, European companies have the opportunity to access innovative language services other than the English language and to further develop them if necessary without disclosing sensitive company or customer data”, Dr. Flores-Herr explains. “In doing so, it is also possible to realize e.g. individual solutions for regional, small and medium-sized companies.”

For ControlExpert, the participation in the OpenGPT-X initiative means the opportunity to develop new AI concepts for the End-2-End claims management. The digitization experts in automotive claims management process 14 Million documents per year. The use of a big, European speech model promises great benefits in the support of customer care: On the one hand, the understanding of documents is hugely improved by the trained speech model. That helps ControlExpert  to make the claims process of the future more efficient, fast and as easy as possible for the customer. With the use of such models, it will not only be possible to extract information from documents in the future, but also to understand individual texts, such as cover letters, descriptions, notes, etc., and to derive appropriate actions from them. On the other hand,OpenGPT-X will above all also be a driver for automated interaction between customer and insurance: A digital speech assistant will in case of a damage ensure an even faster, more automated settlement of claims. The speech assistant will be able to respond to customer concerns even better and e.g. answer flexibly to inquiries. These optimizations will bring ControlExpert closer to its vision that car drivers worldwide should have their claims settled fairly the same day. “OpenGPT-X will revolutionize the speech processing. We see an enormous potential in the technology for the End-2-End claims management. The fact that we can access huge computing power, data and the know-how of renowned research institutions in the course of the project, will also give our research in artificial intelligence a huge push”, says Dr. Sebastian Schoenen, head of research and development at ControlExpert.

Due to the high technical requirements such as the computing power, high performance speech models such as OpenGPT-X can so far only be implemented by large companies or consortiums. For this purpose, the project was able to win over a variety of well-known partners. While Fraunhofer IAIS together with DFKI as well as the companies Aleph Alpha and Alexander Thamm GmbH will develop the AI models, the Research Center Jülich and Technical University Dresden will provide their high performance computing capacities. IONOS SE will supervise the integration into the GAIA-X structure. ControlExpert’s contribution lies primarily in application development. Real-life use cases contributed by the company will ultimately allow the performance and practical application of the speech model to be verified. Radio and television station WDR acts as another supplier of domain-specific use cases for the big speech model. The AI Federal Association (KI Bundesverband) is in charge of the communication of the project.

The project OpenGPT-X  starts under the management of Fraunhofer Institute for Intelligent Analysis- and Information-Systems IAIS and Integrated Circuits IIS. It is one of the winners of the funding competition Gaia-X. The project Gaia-X aims for building a competitive and trustworthy digital data infrastructure. The goal is to create an open, transparent digital ecosystem where data and services can be provided, brought together and shared in an environment of trust.  

The Federal Ministry of Economics and Climate funds the project OpenGPT-X with ca. 15 Million Euro.