Apertus – Open Foundation Model for Sovereign AI (apertvs.ai)

Article summary

The Apertus project is a fully open foundation model for sovereign AI, developed by the Swiss AI Initiative. It provides open weights, open data, and open science, and is compliant with EU AI Act requirements. The model is competitive with top open models and is multilingual, trained on over 1000 languages. Apertus aims to provide a global foundation for building sovereign AI.

Main themes

• Sovereign AI
• Open Source Models
• AI Ethics
• Model Performance
• Data Licensing
• AI Adoption

What commenters say

• The Apertus model is not useful due to its previous version's poor performance and potential copyright issues.
• The open nature of the Apertus model is a significant contribution to science, despite potential limitations in its performance.
• Some argue that training AI models on scraped, unlicensed data is unethical, even if it is legal, and that 'vegan' models trained on licensed or out-of-copyright data are needed.
• Others believe that the benefits of open models, such as Apertus, outweigh the potential drawbacks, and that they can be a rising tide that lifts all boats.
• There is a perceived need for more transparency and accountability in AI development, with some arguing that the current state of AI is driven by a 'tech elite' that prioritizes profit over ethics.
• The use of AI is becoming increasingly widespread, but opinions on its impact and usefulness are divided, with some seeing it as a necessary tool and others as a threat to livelihoods and creativity.
• Some commenters argue that the focus on sovereign AI and open models is misguided, and that the real issue is the lack of transparency and accountability in AI development, rather than the models themselves.
• The development of 'vegan' models, such as those trained on licensed or out-of-copyright data, is seen as a potential solution to the ethical concerns surrounding AI development.