The Open Source Initiative (OSI) has officially launched Version 1.0 of its Open Source AI Definition (OSAID), following two years of development with contributions from experts worldwide. The primary aim of the OSAID is to bring clarity to the concept of open-source AI, by establishing criteria for transparency and accessibility within the field. This new framework seeks to apply traditional open-source principles to Artificial Intelligence (AI), framing it as a system that permits free usage, study, modification, and distribution of the technology.
The OSAID outlines specific requirements that an AI system must meet to be considered open-source. These stipulations are designed to ensure that any AI system is transparent and modifiable throughout its lifecycle. Among the core components are:
Data Information: Detailed documentation is required about the data utilised to train the AI model. This includes information about the data's source, selection process, labelling, and processing methodologies. This level of detail should enable a competent individual to replicate a similar model, although it doesn’t necessarily require that all the raw data is made available.
Code: The complete source code responsible for data processing and model training must be provided. This should be available under OSI-approved licences, fostering transparency in the AI's development process.
Parameters: Access to model parameters and any intermediate training states must be granted under terms approved by OSI. This allows for alteration and transparent adjustment to the technology.
While the OSAID provides these guidelines, it notably refrains from mandating specific licensing mechanisms for data and parameters due to ongoing legal uncertainties surrounding copyright protections. In this regard, the framework allows for these components to be shared under OSI-approved terms rather than rigid licences.
The introduction of OSAID has garnered support from several organisations that view it as a critical step towards establishing uniform practices for open-source AI. Yet, the guidelines have also sparked some debate, particularly concerning safety and ethics, which some experts feel are not explicitly addressed in the current framework.
Seth Khon, an industry observer, expressed his concerns: "I appreciate OSI's effort to promote transparency and openness with the Open Source AI framework. However, I am concerned that it lacks explicit considerations for safety. While open access and modifiability are important, AI poses unique risks that transparency alone does not mitigate. I believe the framework could benefit from integrating safety standards or guidelines to address responsible and ethical use, especially in high-stakes applications."
Despite these discussions, the OSI's release of the OSAID is seen as a foundational move to support both innovation and accountability in AI development. By providing clear standards, the OSI aims to encourage ethical AI advancement while facilitating effective oversight for developers and entities operating within the technological sphere.
Source: Noah Wire Services