Project and lecture videos

The world is becoming more and more information-rich, sometimes of good and sometimes of bad quality. Large Language Models (LLMs) like ChatGPT are trained on this information to answer our questions and assist us. To improve their answers, LLMs can use methods like Retrieval Augmented Generation (RAG), which attaches external information to the question to generate better answers. However, the most valuable and interesting information is often not freely accessible. So, how can large language models deliver better answers by incorporating protected information?

If we combine LLMs and data spaces using the concept of retrieval-augmented generation, questions can be sent to a data space, asking each data provider how likely it is that their data set will improve the answer. The data sets are offered to the user together with a probability calculation of how helpful the data is. The user can accept the terms of use (e.g. by purchasing the data) and then gain access to the raw data. The LLM on the consumer side then generates an answer to the original question from the new data.

By linking data spaces with LLMs via the concept of RAG, business cases can be realised that were not feasible in the past. LLM-based data spaces for searching in protected data are a development of the Fraunhofer ISST in cooperation with Huawei using the technology Boot-X.

What comes after the proof of concept? How we can put digital innovations on the road | March 6, 2024 | 12:20 p.m. | Transformation Stage | 

There is a need for action in the implementation of digital strategies in Germany. We have no problem developing innovations in this country. But we also have too many reasons for not putting them on the road.

That's why the digital transformation in Germany is - to be honest - sobering. In too many industries, processes are still based on Excel, e-mail and telephone

Boris Otto is calling for a tangible "master plan for digital transformation" and already has the most important cornerstones ready. In this keynote, he explains why we need eco-systems instead of ego-systems, why we should view software not as a by-product but as a value-adding factor and why innovation potential cannot unfold if we always fear in advance what might not work.

Companies can benefit from cross-company data sharing and strategic data management in a variety of ways. The IEDS (Incentives and Economics of Data Sharing) project, financed by the German Federal Ministry of Education and Research (BMBF), is looking at the benefits that may be realized and how firms should proceed both internally and externally.

Visit www.ieds-projekt.de/en for further details.

We're looking forward to the exchange!

Across the economy, companies are using digital technologies to pivot into innovative services, accelerate growth, and transform business processes. Data assets play an important role in this modern economy. To share this data, new technology is needed that supports openness, interoperability, extensibility, sovereignty and transparency. The Eclipse Dataspace Connector is that technology. The open-source project is supported by Fraunhofer Institute for Software and Systems Engineering ISST, Daimler TSS, BMW Group, Deutsche Telekom, Microsoft, Amazon AWS, SAP, ZF Friedrichshafen and GAIA-X AISBL as well as the International Data Spaces Association.

The Dataspace Connector (DSC) is an open source software for sovereign data exchange. Even after the data has been exchanged, the data provider retains control over what happens to the data. Based on the International Data Spaces reference architecture, the DSC allows data to be extended with policies and conditions and processed. The DSC is a core component for building data spaces according to the DIN SPEC 27070 standard. The Dataspace Connector is licensed under Apache 2.0. and is absolutely free for personal or commercial use.

Further information: www.dataspace-connector.io

The Shared Digital Twin was developed based on the RIOTANA IoT architecture created by Fraunhofer ISST. With RIOTANA, informative performance figures can be generated from the raw data of ongoing processes (such as vibrations, temperature, or friction) in real time. The administration shell standardized by the industry 4.0 platform is used as the digital twin, so the framework and interfaces of the twin are clearly defined and uniform.

Shared however means that the data do not remain only in the company but are exchanged in network structures. Here the solution uses IDS connectors promoted by Fraunhofer and numerous industry partners within the framework of the International Data Spaces (IDS) initiative, with the goal of enabling the standardized, sovereign exchange of data between various companies. By using International Data Spaces, control of the data is always retained by the company that provides the data by attaching usage policies to the data that are shared with cooperation partners. Both companies that share the digital twin can also enrich the data with information.

Thus the Shared Digital Twin with the administration shell and IDS connectors is based on two standards that are already being used by industry. What database is used in the twin can be decided according to the respective use case.

Immersion - always what? Sounds a bit like emulsion, but it has nothing to do with that. Immersion means as much as immersion. In the context of VR, immersion means being immersed in the virtual environment, which can be achieved particularly well with VR glasses. The positive effects of immersion are also used in health research, for example to treat phobias, train nursing staff or for rehabilitation. Because virtual reality can help to reduce fears, strengthen self-confidence and gain security - and finally we can transfer the movements into the real world. This is what we are researching at the Fraunhofer ISST in the area of "Serious Games".

SMITH stands for Smart Medical Information Technology for Healthcare. It describes one of four publicly funded projects under the medical information sciences initiative of the Federal Ministry of Education and Research (BMBF), with the goal of improving individual patient care by efficiently making data available for clinical research.

A reference architecture was developed for this purpose, allowing the participating university hospitals and their data integration centers to exchange data using a central platform: the SMITH service platform. The SMITH marketplace developed by our researchers is one of several applications serving as the central point of access for the end user. It permits researchers to access and use data of all data integration centers that are relevant for their research.

Watch the video to find out exactly how these processes work and learn about the technical background.

What is neurogaming? How can neurogaming help people, and what are we researching at Fraunhofer ISST regarding neurogaming?

Neurogaming is a mix of two disciplines: neuroscience and digital gaming. It focuses on game interaction using brain waves. The data generated when the game’s brain-computer interface measures the brain’s electrical activity are interpreted by the digital game. Thus the brain activity controls what happens in the game, and the brain becomes the game controller.

Neurogaming is a very recent field of research with great potential. A great deal of work remains to be done. Currently we are researching how we can support children and adolescents in psychotherapy treatment through digital games in uncomplicated and utilizable settings.

The first application developed as part of a master thesis is "Enlightend". This is an initial prototype, a technologically uncomplicated setup for children and adolescents with concentration and attention deficits. It is intended to help players improve their power of concentration, use it purposefully, and maintain it.

In his lecture, Prof. Dr.-Ing Boris Otto will show why we need a data economy — in other words, a balance between the community's interest in sharing the data of each individual and the individual's justified desire for protection and self-determination of personal data. He also discusses how this balancing act can be supported technologically with the help of data spaces and what contribution data spaces make to data sovereignty in the context of the European data infrastructure initiative Gaia-X.

»The Value of Data - über den Wert einer strategischen Ressource«

»Wir brauchen eine Kosten- und Leistungsrechnung für Daten. Und die Fähigkeit, Daten zu teilen, ohne die Kontrolle darüber zu verlieren«, sagt Prof. Dr.-Ing. Boris Otto während eines Science Slams im Rahmen der Night of Innovations auf der Hannover Messe (1.4.2019). Wie das gehen soll und warm das wichtig ist, erklärt er im Video.