Open-source Artificial Intelligence Model from Switzerland: Functions and Unique Distinctions
In a groundbreaking move, Switzerland has unveiled Apertus, a new player in the AI industry that promises to redefine the landscape of artificial intelligence. Developed in the heart of Europe, Apertus stands as a compelling alternative to the tech giants of Silicon Valley.
Apertus is more than just a new AI model; it's a foundational technology designed to serve the entire Swiss economy and extend its benefits beyond national borders. The project's development is powered by one of the world's leading AI supercomputers, "Alps," located at the Swiss National Supercomputing Centre (CSCS).
The developers of Apertus have made a conscious effort to adhere to Swiss data protection laws and the EU AI Act, ensuring a focus on using only publicly available data. This approach sets Apertus apart from many other AI models, which often rely on proprietary data sources.
To further reinforce its commitment to transparency, Apertus's developers have implemented a system to honour "machine-readable opt-out requests" from websites and to filter out personal data and other unwanted content before training. This system, unique in the AI industry, ensures that user privacy is respected and maintained throughout the development process.
Apertus is an open-source large language model (LLM) and is positioned as a model built for the public good. It's a tool that can be trusted by a society increasingly wary of data exploitation. The model is freely available in two sizes (8 billion and 70 billion parameters) under a permissive open-source license.
The project aims to foster innovation by serving as a foundational technology for various sectors, not just a single product. Its multilingual focus ensures it is globally relevant and serves the unique linguistic landscape of Switzerland and other smaller language communities.
Apertus is a collaborative effort by the Swiss Federal Institutes of Technology in Lausanne (EPFL) and Zurich (ETH Zurich), along with the Swiss National Supercomputing Centre (CSCS). It's a statement of intent, demonstrating a belief that AI should be developed openly, responsibly, and for the benefit of all.
The open and multilingual language model Apertus was cooperatively initiated and funded by a research group led by Mayer, together with the EPFL, ETH Zurich, and the Swiss National Supercomputing Centre CSCS. The model was trained on a massive 15 trillion tokens across more than 1,000 languages, with 40% of the data being non-English, including underrepresented languages like Swiss German and Romansh.
Apertus stands as a compelling alternative to dominant proprietary models like ChatGPT or Google's Gemini, offering a path toward a more transparent and equitable future for artificial intelligence. It provides unprecedented access to its entire development process, including architecture, training data, model weights, and training recipes. This openness allows for inspection and auditing of its creation, making it a black box alternative to proprietary models.
In summary, Apertus, the new open-source AI project from Switzerland, is making waves for its distinct approach to AI development, focusing on transparency, public good, and digital sovereignty. The name Apertus, Latin for "open," perfectly encapsulates the project's mission and vision for a more inclusive and equitable future in AI.
