On Monday, Switzerland unveiled an open-source AI model named Apertus, providing an alternative to well-known proprietary systems like ChatGPT from OpenAI and Claude from Anthropic. The announcement was covered by SWI and highlighted by Engadget.
The name Apertus, derived from Latin meaning “open,” reflects the model’s intention to establish a new standard for transparent and trustworthy AI technologies. According to its developers, the model is capable of understanding over 1,800 languages and is available in two configurations—one with 8 billion parameters and another with 70 billion parameters. It has been compared favorably to Meta’s anticipated Llama 3 model set for release in 2024, as reported by SWI.
Apertus has been developed in alignment with European Union copyright legislation and the voluntary AI code of practice. This comes amid concerns from some US AI companies who argue that such regulations could hinder AI innovation and implementation. The training data for Apertus has been sourced exclusively from publicly available materials, respecting opt-out requests from various websites, as stated in the technical report.
3 Comments
