Open Source vs. Closed Source

A look back at the speech given by Michel-Marie MAUDET, Managing Director of LINAGORA.

He spoke at a round table at MasterDevFrance on the theme of ‘Open Source vs. Closed Source’.

Michel-Marie returned to the notion of truly OpenSource AI. Today, many models are described as Open Source even though they only share their weights (Open Weight), without the training data.

The Open Source Initiative (OSI) has attempted to define what an open source model is, but there is no consensus on this definition. A model can only be considered truly open source if :

• its source code is freely accessible
• Its training data is published;
• There are no restrictions on its use;

Initiatives such as OpenLLM France/Europe are leading the way with the LUCIE model, one of the few truly open source models.