Most modern Large Language Models (LLMs) are built behind closed corporate doors. This centralised control often results in black-box training data, limited language representation, and ethical gray areas.
Can the public research sector build competitive, ethical, and fully transparent AI models instead?
The Data Science for Social Impact (DSFSI), AI4D African Languages Lab, and African Institute of Data Science and Artificial Intelligence (AfriDSA) at UP invite you to our upcoming DS4Society Webinar.
Title: Apertus: Democratizing the Foundation for Global Sovereign AI
We will explore Apertus—a fully transparent LLM initiative created through a collaboration between EPFL, ETH Zürich, and CSCS. This initiative showcases how public institutions can challenge corporate dominance in AI by developing models that:
🌐 Support over 1,000 languages.
🔒 Comply strictly with copyright and ethical standards.
📖 Rely exclusively on open, public training data.
Antoine Bosselut is a data engineer and developer relations specialist with deep open-source technology expertise. Background in startups, academia & industry, contributions to a broad range of technical projects like Dribdat, Django, Proxeus, Frictionless Data, CKAN. Prior to joining the EPFL AI Centre to support the Swiss AI Initiative, O lectured at HSLU in the Data Design + Art program, and continues to support teaching and research collaborations with universities and centres of excellence around Switzerland. A life-long learner in Communication Systems, who shares stories of a life-long fascination with protocols, networks and signals at log.alets.ch
Date: Thursday, 13 August 2026
Time: 12:00 PM – 13:00 PM SAST
Location: Online (Link provided after registration)
Who should attend: Postgraduate students, supervisors, scientists, and AI/data researchers.
RSVP Deadline: Tuesday, 11 August 2026
✉️ Enquiries:
dsfsi.info@up.ac.za
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.