Following the successful completion of DEEP-EST and the launch of the prototype at the Jülich Supercomputing Centre, the DEEP projects face a new challenge: how to design programming environments that can support future Exascale systems with a wide variety of different workloads.
Enter DEEP-SEA (DEEP – Software for Exascale Architectures). DEEP-SEA started in April 2021 with a duration of 3 years. The project provides the programming environment for future European Exascale systems, adapting all levels of the software stack to support highly heterogeneous compute and memory configurations. It will also enable code optimisation across existing and future architectures and systems. The software stack includes low-level drivers, computation and communication libraries, resource management and programming abstractions with associated runtime systems and tools.
European Digital Innovation Hubs Network (EDIH)
The University of Iceland, or Háskóli Íslands, situated in Reykjavík, was founded in 1911 and is the country's leading academic institution. It offers a broad range of study programs, including practical fields like High-Performance Computing (HPC), Artificial Intelligence (AI), and Computer Science.
These disciplines at the university have seen remarkable growth, backed by state-of-the-art facilities and renowned faculty, positioning Iceland on the global map of technological advancement. Alongside its modern contributions, the university remains a guardian of Iceland's distinct linguistic and cultural legacies.
The TrustLLM project is an initiative funded by the European Union Horizon program. Its primary objective is to develop a Large Language Model (LLM) specifically tailored for Germanic languages. With a particular emphasis on supporting lower-resource languages, the project seeks to bridge the gap in linguistic representation within the realm of advanced AI models.
The project has set forth several key objectives:
- Improving the Factual Correctness of LLMs: Ensuring that the generated content is accurate and reliable.
- Providing a Modular and Transferable Framework: This will facilitate the training of multilingual LLMs, enabling seamless integration of various languages.
- Sustainable and Trustworthy Development: The LLMs developed under this project will align with European values, ensuring their ethical and responsible use.
- Establishing a European Ecosystem: This ecosystem will cater to the evolving needs of LLMs, fostering collaboration and innovation.
- Assessment through Demonstrations and Benchmarking: By using real-world use cases, the project aims to showcase the capabilities of the developed LLMs and measure their performance against established benchmarks.
- Large Scale Data Management: Efficiently handling and processing vast amounts of linguistic data to train these models.
As of the latest update, the grant application for TrustLLM has been accepted, and the project is gearing up to commence its operations. The anticipated start date is no later than November 1st. In line with its objectives, the project is on the lookout for talented individuals, including potential PhD candidates, to join its ranks.