Fritz

For improving the capabilities of general-purpose AI models and for extending their applicability to domains where the temporal dimension – among several others – is of importance, we will target the development of the next generation of Multimodal Space-Time Foundation Models (MSTFMs). These will combine spatio-temporal understanding, which is important even for modalities such as the

visual one that have already been introduced in large generative models, with the effective management of new time-relevant modalities that are yet to be supported in foundation models, such as industrial time series data, remote sensing data and health-related measurements. Real and synthetic data, to mitigate data scarcity, will be leveraged for training general-purpose MSTFMs and for further adapting them for specific downstream tasks. Real data used for training will include data directly provided by members of the consortium as well as data from relevant European Data Spaces, while complementary synthetic data will be generated by exploiting existing generative AI  capabilities as well as new ones developed in the project. European HPC infrastructure is directly included in the consortium to ensure the availability of the necessary computing resources.

Funding Body

Funding Body

European Union

That will address major challenges hampering the deployment of AI technology. These grand challenges are fundamental in nature. Addressing them in a sustainable manner requires a lighthouse rooted in scientific excellence and rigorous methods. We will develop a strategic research agenda which is supported by research programmes that focus on “technical robustness and safety”,“privacy preserving techniques and infrastructures” and “human agency and oversight”. Furthermore, we focus our efforts to detect, prevent, and mitigate threats and enable recovery from harm by 3 grand challenges: “Robustness guarantees and certification”,“Private and robust collaborative learning at scale” and “Human-in-the-loop decision making: Integrated governance to ensure meaningful oversight” that cut across 6 use cases: health, autonomous driving, robotics, cybersecurity, multi-media, and document intelligence. Throughout our project, we seek to integrate robust technical approaches with legal and ethical principles supported by meaningful and effective governance architectures to nurture and sustain the development and deployment of AI technology that serves and promotes foundational European values. Our initiative builds on and expands the internationally recognized, highly successful and fully operational network of excellence ELLIS. We build on its 3 pillars: research programmes, a set of research units, and a PhD/PostDoc programme, thereby connecting a network of over 100 organizations and more than 337 ELLIS Fellows and Scholars (113 ERC grants) committed to shared standards of excellence. Not only will we establish a virtual center of excellence, but all our activities will also be inclusive and open to input, interactions, and collaboration of AI researchers and industrial partners in order to drive the entire field forward.

Funding Body

Funding Body

European Commission

Ever since the last Coronavirus epidemic caused by SARS-CoV-1, plans and tools for the containment of epidemics are being developed. However, an appropriate early warning system for local health authorities addressing this need on a regional, targeted level is not available. In the current SARS-CoV-2 pandemic, the need for such a system becomes increasingly obvious. The heterogeneity of different regions and localized outbreaks require a locally adapted monitoring and evaluation of infection dynamics.

Early recognition of an emerging epidemic is a crucial part of a successful intervention. The comparison of Germany to other European nations illustrates how crucial a timely implementation of non-pharmaceutical interventions is for the containment of an epidemic. Hence, continuous monitoring of infection processes is indispensable. For strategic planning of political interventions, epidemiological modelling and scenario calculations for forecasting and evaluation of interventions and scenarios have shown their importance. The accuracy of such forecasts largely depends on the robustness and broadness of the underlying data. Further, there is a need for an intelligible presentation of often complex simulation results without oversimplification of their interpretation and inherent uncertainty.

In this proposal, we develop a platform that integrates data streams from various sources in a privacy preserving manner. For their analysis, a variety of methods from machine learning to epidemiological modeling are employed to detect local outbreaks early on and enable an evaluation for different assumptions and on different scales. These models will be integrated into automatized workflows and presented in an interactive web application with custom scenario simulations. The platform will be based on insights gained by retrospective and prospective evaluation of the COVID-19 pandemic, using SARS-CoV-2 as a blueprint for the prevention and containment of future respiratory virus epidemics. The platform will be transferred to the Academy for Public Health Services and optimized in pilot projects with selected local health authorities under real-world conditions.

Funding Body

Funding Body

Helmholtz-Gemeinschaft e.V.; Impuls- und Vernetzungsfonds

Partner

Project Partner

Helmholtz-Zentrum für Infektionsforschung (HZI)

Project Partner

Deutsches Zentrum für Luft- und Raumfahrt (DLR)

Project Partner

Forschungszentrum Jülich

Project Partner

Helmholtz-Zentrum für Umweltforschung (UFZ)

Subproject: Federated Analytics and Machine Learning, Privacy Models, and Benchmarking: CISPA

The goal of PrivateAIM is to develop a federated platform for machine learning (ML) and data analysis within the framework of the Medical Informatics Initiative (MII), in which analyses are brought to the data rather than the data being brought to the analyses. Methods that enable distributed data processing in the Data Integration Centers (DIZ) established by the MII are important for several reasons:

1. Patient data may only be used without consent if anonymity is ensured;

2. Federated technologies can help connect the MII with other health data networks.

However, the mechanisms currently established within the MII have significant limitations and, for example, are not suitable for complex ML and data science tasks. Moreover, federated platforms developed in other contexts are:

1. complicated to set up and operate,

2. support only a limited number of analysis or ML methods,

3. do not implement modern privacy-preserving technologies, and

4. are not scalable or mature.

The PrivateAIM consortium, supported by all MII consortia, brings together experts to develop the next generation of federated analysis and ML platforms for the MII. The Federated Learning and Analysis Methods Platform (FLAME) will combine state-of-the-art federated methods with innovative privacy models for multimodal data. Using components for monitoring and controlling the level of protection, these will be integrated into a distributed infrastructure that can be easily adopted by the Data Integration Centers.

The practical implementation will be supported by addressing challenges at the intersection of technology and law, developing concepts for operation by hospital IT departments, and coordinating with ethics committees and data protection officers.

Funding Body

Funding Body

Bundesministerium für Forschung, Technologie und Raumfahrt

Partner

Project Partner

Universität zu Köln

Universität zu Köln

Project Partner

Universität Leipzig

Universität Leipzig

Project Partner

Deutsches Krebsforschungszentrum (DKFZ)

Project Partner

Universität Ulm

Universität Ulm

Project Partner

Universität Heidelberg

Universität Heidelberg

Project Partner

Albert-Ludwigs-Universität Freiburg

Albert-Ludwigs-Universität Freiburg

Project Partner

Charité - Universitätsmedizin Berlin

Charité - Universitätsmedizin Berlin

Project Partner

Technische Universität München

Technische Universität München

Project Partner

Eberhard Karls Universität Tübingen

Eberhard Karls Universität Tübingen

Project Partner

TMF - Technologie- und Methodenplattform für die vernetzte medizinische Forschung e. V. Berlin

TMF - Technologie- und Methodenplattform für die vernetzte medizinische Forschung e. V. Berlin

Project Partner

Friedrich-Alexander-Universität Erlangen-Nürnberg

Friedrich-Alexander-Universität Erlangen-Nürnberg

Project Partner

Ludwig-Maximilians-Universität München

Ludwig-Maximilians-Universität München