A groundbreaking artificial intelligence model, dubbed \"Apertus\" (Latin for \"open\"), has been introduced by a collective of prominent Swiss academic and scientific organizations, including EPFL, ETH Zurich, and the Swiss National Supercomputing Centre (CSCS). This new creation is envisioned as a foundational element for future research and practical applications, emphasizing complete transparency in its architecture, training methodologies, and data sources. Unlike other AI systems that might only disclose partial details, Apertus makes every aspect of its development publicly accessible, underscoring a commitment to openness.
Apertus offers immense potential for developers and organizations to craft diverse AI-powered solutions, ranging from sophisticated chatbots to advanced translation tools and specialized educational platforms. The model is available in two distinct versions: an 8-billion-parameter model and a more expansive 70-billion-parameter variant. Both versions are released under a permissive open-source license, facilitating their use in academic research, educational initiatives, and commercial ventures. This initiative aims to serve as a blueprint for the creation of reliable, independent, and inclusive AI systems, with continuous updates and improvements planned by a dedicated team of engineers and researchers. Furthermore, the model's comprehensive training dataset, comprising 15 trillion tokens across over 1,000 languages, significantly includes non-English languages, such as Swiss German and Romansh, highlighting its dedication to multilingualism and inclusivity.
The introduction of Apertus signifies a pivotal shift towards AI as a public utility, akin to essential infrastructure like roads or electricity. Its development meticulously adhered to stringent Swiss data protection regulations, copyright laws, and the transparency mandates of the EU AI Act. The training data was carefully curated from publicly available information, with strict filtering to eliminate personal data and honor opt-out requests, and ethical guidelines were applied to exclude undesirable content. This forward-thinking approach showcases how powerful generative AI can also be transparent and openly accessible, fostering innovation while ensuring responsible development. The journey of Apertus is ongoing, with future enhancements planned to broaden its capabilities, refine its efficiency, and create specialized tools for sectors like law, healthcare, climate science, and education, all while maintaining its core principles of openness and trustworthiness for global public good.
This pioneering endeavor by Switzerland demonstrates a profound commitment to harnessing artificial intelligence for the betterment of society. By championing transparency, accessibility, and ethical considerations, Apertus embodies the spirit of collaborative innovation. It serves as a beacon, illustrating that advanced technological development can, and should, be pursued with integrity and a focus on collective welfare, ensuring that the benefits of AI are widely shared and responsibly managed for a more equitable and advanced future.
