Yannick Bormuth

📢 𝗦𝘄𝗶𝘁𝘇𝗲𝗿𝗹𝗮𝗻𝗱 𝗶𝘀 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝘁𝗵𝗲 𝗼𝗽𝗲𝗻 𝗔𝗜 𝗺𝗼𝗱𝗲𝗹 𝗕𝗶𝗴 𝗧𝗲𝗰𝗵 𝘄𝗼𝗻’𝘁. ETH Zürich, EPFL, and the Swiss National Supercomputing Centre (CSCS) will release a fully open Large Language Model (LLM) – trained on 𝗔𝗹𝗽𝘀, one of Europe’s fastest AI supercomputers. 𝗪𝗵𝗮𝘁 𝘀𝗲𝘁𝘀 𝗶𝘁 𝗮𝗽𝗮𝗿𝘁? ✅ 𝗠𝘂𝗹𝘁𝗶𝗹𝗶𝗻𝗴𝘂𝗮𝗹 𝗯𝘆 𝗱𝗲𝘀𝗶𝗴𝗻 Trained on 1,500+ languages – with strong capabilities in underrepresented communities. ✅ 𝗙𝘂𝗹𝗹𝘆 𝗼𝗽𝗲𝗻 (unlike commercial models) Weights, code, and training data will be public. Licensed under Apache 2.0 for unrestricted reuse – commercial, scientific, or regulatory. ✅ 𝗚𝗹𝗼𝗯𝗮𝗹𝗹𝘆 𝗰𝗼𝗺𝗽𝗲𝘁𝗶𝘁𝗶𝘃𝗲 Available in 8B and 70B parameter sizes – trained on 15 trillion high-quality tokens. ✅ 𝗧𝗿𝘂𝘀𝘁𝗲𝗱 𝗯𝘆 𝗱𝗲𝘀𝗶𝗴𝗻 Developed entirely on Swiss public infrastructure – aligned with Swiss law and the EU AI Act. ✅ 𝗣𝗼𝘄𝗲𝗿𝗲𝗱 𝗯𝘆 𝗔𝗹𝗽𝘀 – 𝗴𝗿𝗲𝗲𝗻, 𝘀𝗼𝘃𝗲𝗿𝗲𝗶𝗴𝗻, 𝗮𝗻𝗱 𝗯𝘂𝗶𝗹𝘁 𝘁𝗼 𝘀𝗰𝗮𝗹𝗲 Trained on Alps, using 100% carbon-neutral electricity and 10,000+ NVIDIA Grace Hopper Superchips – enabled by a 15+ year collaboration between NVIDIA, Hewlett Packard Enterprise, and CSCS. ✅ 𝗪𝗶𝘁𝗵 𝘀𝘁𝗿𝗼𝗻𝗴 𝗿𝗼𝗼𝘁𝘀 𝗶𝗻 𝘁𝗵𝗲 𝗚𝗿𝗲𝗮𝘁𝗲𝗿 𝗭𝘂𝗿𝗶𝗰𝗵 𝗔𝗿𝗲𝗮 😎 Where world-class universities, sovereign infrastructure, public funding, and a dense AI talent pool create one of Europe’s most trusted ecosystems for building technology at scale.🚀 👏 𝗟𝗲𝗱 𝗯𝘆 𝘄𝗼𝗿𝗹𝗱-𝗰𝗹𝗮𝘀𝘀 𝗔𝗜 𝗺𝗶𝗻𝗱𝘀: ETH AI Center’s Imanol Schlag, EPFL AI Center’s Antoine Bosselut and Martin Jaggi. Supported by the Swiss AI Initiative and ELLIS - European Laboratory for Learning and Intelligent Systems 🔗 𝗗𝗶𝘀𝗰𝗼𝘃𝗲𝗿 𝗵𝗼𝘄 𝗚𝗿𝗲𝗮𝘁𝗲𝗿 𝗭𝘂𝗿𝗶𝗰𝗵 𝘀𝘂𝗽𝗽𝗼𝗿𝘁𝘀 𝗴𝗹𝗼𝗯𝗮𝗹 𝗔𝗜 𝗯𝘂𝗶𝗹𝗱𝗲𝗿𝘀: https://okt.to/NCI68o More about the Swiss LLM: https://okt.to/BR2HIV

905

16 Kommentare

We’re thrilled to welcome Yannis Chronis, newly appointed Tenure Track Assistant Professor, to the Department of Computer Science. His research focuses on making databases and data processing more efficient on modern and emerging hardware platforms. In this interview, Yannis talks about the goals he aims to pursue at ETH Zurich, the growing importance of data management in light of AI advancements, and the responsibilities these technologies place on computer scientists. Read full interview: https://lnkd.in/drUiuq6Z Have a great start, Yannis! #datamanagement #computerscience #dataprocessing #artificialintelligence

386

19 Kommentare

Working toward greater balance 📊 A look at the Gender Equality Monitoring 2024 at UZH shows a success: ✨ By the end of last year, the proportion of women among professors reached a record 30.3%. 👩‍🎓 Assistant professorships played a key role: 50% were awarded to women researchers – another record 🏆. 💡 There is still room for improvement in the highest-ranking professorship category, the full professorships. 🔍 Based on the data collected in the Gender Equality Monitoring, two levers can be identified where UZH could take action to further increase the proportion of female professors 📈. 👉 Read more: https://lnkd.in/dkavj338

171

11 Kommentare

🚀 Switzerland’s first large-scale, open, multilingual language model is here: Apertus   EPFL, ETH Zürich, and CSCS have just released #Apertus, a milestone in generative AI built with transparency, diversity, and accessibility at its core.   So, what makes Apertus unique❓ 🔓 Fully open & transparent:  Apertus is Latin for “open”, highlighting that its architecture, training data, and model weights are fully accessible and documented. 🌍 Multilingual by design: Trained on 15 trillion tokens across 1,000+ languages, with 40% non-English data. This includes underrepresented languages such as Swiss German and Romansh. 📈 Scalable access: Available in two versions (8 billion and 70 billion parameters) via Hugging Face and Swisscom’s sovereign Swiss AI platform. 🤝 Built for the public good: Open-source licensing enables use in research, education, and commercial applications. 🌐 Global availability: People outside Switzerland can also access Apertus through the Public AI Inference Utility, making it part of a worldwide movement for public, open AI.   With Apertus, Switzerland positions itself at the forefront of open and responsible AI, turning cutting-edge research into a foundation for innovation across society, academia, and industry.   Read more in this article: https://lnkd.in/drnMUjSt   Learn how to access Apertus: https://lnkd.in/ey6mypX9   #WhySwitzerland #OpenSourceAI EPFL AI Center ETH AI Center Swiss {ai} Weeks Presence Switzerland swisstech Deep Tech Nation Switzerland Swissnex Basel Area Business & Innovation Greater Geneva Bern area (GGBa) Greater Zurich Area St.GallenBodenseeArea Lucerne Business Economic Promotion Valais Standort Promotion in Obwalden Standortförderung Kanton Aargau Brigitte Fäh Martin Jaggi Thomas Schulthess Imanol Schlag Daniel Dobos Joshua Tan

97

3 Kommentare

🚀 From Vision to Impact: 5 Years of the ETH AI Center Five years ago, the ETH AI Center was founded with a clear mission: 👉 We foster research excellence, industry innovation, and AI entrepreneurship to promote trustworthy, accessible, and inclusive AI systems. Today, it stands as a thriving community of over 130 professors and more than 70 active fellows (with 25 alumni), uniting disciplines across ETH Zurich — and beyond. Through the Doctoral and Postdoctoral Fellowship Programs, co-supervised across diverse research areas, we’ve built an environment where fellows not only become faculty and innovators but also launch startups and pioneer interdisciplinary breakthroughs. Together with partners like EPFL and CSCS, the Center continues to bridge research and real-world impact — from projects such as Apertus to a growing network of AI researchers driving progress across Switzerland. Our initiatives — including but not limited to Swiss AI Initiative, ICAIN, AI Launchpad, Talent Kick, AI House Davos, the AI+X Summit, the Zurich AI Festival, and the AI Challenge for Teens — reflect our commitment to advancing trustworthy, inclusive, and impactful AI for science, industry, and society. #ETHZurich #AICenter #AIInnovation #AIResearch #TrustworthyAI

157

4 Kommentare

Is the new open-source large language model developed by ETH Zürich and EPFL “just another LLM” or a real game changer?   Let’s start with the “why”. The launch of Deepseek, and most recently Kimi K2, has once again sparked a discussion about the costs and the accessibility of large language models. More affordable models could indeed be a game changer for public institutions and smaller businesses who have fewer resources. In addition, digital sovereignty becomes more and more important - many users want to know who they share their data with.   With their new large language model, which will be released in late summer, ETH and EPFL are launching the first LLM “built for public use across science, government, and industry.”    The model is transparent by design with the source code, model weights, and training data being publicly available. It is released under an open-source license and supports over 1,000 languages. Users will also be able to modify it according to their needs.    So far, the facts. But what’s the impact going to be?   With 70 billion parameters, it might not be the largest LLM. Thus, it’s certainly not built to compete with the big commercial models on every single aspect. Yet, its transparent design goes well beyond what other open-source models offer.    Also, its adaptability is a big plus. Users will be able to finetune it to their needs thanks to its open architecture. This allows them to experiment with new use cases without having to invest big sums beforehand. Ultimately, the data provided by a user will not be used to train the model, which eases data privacy concerns.      That’s why I see a strong use case for public and research institutions where trust and data privacy play a particularly important role. The new Swiss-made LLM is certainly an important step towards democratizing the use of AI.    I am looking forward to its release in the coming weeks - and am more than excited to test it!      #swissai #llm   https://lnkd.in/drbiDAHQ   

80

𝐒𝐰𝐢𝐭𝐳𝐞𝐫𝐥𝐚𝐧𝐝’𝐬 𝐝𝐞𝐞𝐩 𝐭𝐞𝐜𝐡 𝐥𝐚𝐧𝐝𝐬𝐜𝐚𝐩𝐞 𝐬𝐡𝐢𝐧𝐞𝐬! With strengths to celebrate and gaps to tackle. Based on our latest spider graph, here’s our take: ➕ Innovation & Academia are our powerhouses: Switzerland 🇨🇭 is a global leader in patents and boasts top-tier universities that are engines for innovation and spinouts. ➖ Capital & Ecosystem need a boost: While early-stage funding is strong, we're losing out on late-stage capital and need a more connected, specialized ecosystem to help startups scale. 🎯 The goal: To create a cohesive "Deep Tech Nation" where excellence in innovation is matched by an ecosystem that can turn great ideas into global success stories. To be the #1 Deep Tech Nation by 2033, Switzerland must do more than invent. It must host and scale innovation. We’ve got the IP, the science, and growing capital. Now we need scale muscles and a bigger stage. Read the full analysis by Amara Cespedes to understand where we're strong, where we're slow, and what to do about it. Link in the comments! 👇

165

10 Kommentare

GSEM Senior Lecturer Dr. Thomas Maillart, in collaboration with armasuisse Science + Technology, has received funding of CHF 60,000 for the project “AI-Driven Scientific Foresight: Modeling Concept Evolution for Predicting Emerging Breakthroughs.” This project strengthens research ties in knowledge organization and scientific forecasting and aims to develop artificial intelligence methods to anticipate the evolution of scientific concepts and predict emerging breakthroughs across disciplines.   For more information on the ongoing research projects of the GSEM faculty: https://lnkd.in/d59GvPh   #GSEM #GenevaAdvantage #ScientificForesight #ArtificialIntelligence Université de Genève – armasuisse Wissenschaft und Technologie

88

13 Kommentare