Open-source definition of AI is here, but data remains point of discussion

The Open Source Initiative (OSI) has released its first version of an open-source AI definition aiming to clarify what qualifies as true open-source, emphasizing transparency and accessibility in AI models.

OSI specifies that an AI model must include software for dataset compilation, model parameters, and legally shareable data, while promoting broad usability and compliance with open-source standards to prevent 'openwashing'.

The stringent requirements aim to ensure that only AI models that fulfill specific criteria can be labeled as open-source, promoting a genuine understanding of 'open-source' in the AI community.

Critics raise concerns, particularly regarding the classification of training data as open-source, questioning the implications and interpretations that arise from the new definitions set by the OSI.

Read at Techzine Global