As artificial intelligence (AI) becomes more integrated into daily life, establishing a clear, industry-wide definition of open source in AI is crucial to ensure its responsible use and public benefit. In this video, Mer Joyce, Co-Design Facilitator for the Open Source AI Definition (OSAID) at the Open Source Initiative (OSI), discusses the challenges of defining what open source means for AI, the four requirements for the OSAID, the issues of “open washing”, the importance of community and industry involvement, and OSI’s involvement in the initiative.

Introduction to Mer Joyce and the role of Do Big Good in co-design projects

• Joyce’s professional journey and role as the founder of Do Big Good, a Seattle-based co-design firm specializing in collaborative design and community involvement, is discussed.
• Joyce’s extensive experience includes work on state-level policy initiatives in Washington State, including health and education, as well as involvement in international projects with the Open Knowledge Foundation (OKF) and Open Source Initiative (OSI).
• Do Big Good helps clients make decisions through a collaborative approach where knowledge and power are shared with the community to enhance the decision-making process.

The importance of defining clear open source standards for AI

• The necessity of creating a definition for open source in AI is explained by Joyce, who addresses the confusion caused by the lack of a universally accepted standard and the issue of “open washing,” where proprietary systems are inaccurately marketed as open source.
• OSI decided to extend its expertise in software to develop a definition for AI.
• Joyce emphasizes the importance of having a clear definition to prevent misuse and ensure that the benefits of open source AI are accessible to the public.

Key requirements and principles for open source AI systems

• Four requirements of the Open Source AI definition include the freedoms to study, use, modify, and share the system. This also encompasses the need for open weights, open code, and detailed data information to enable the creation of similar systems.
• Joyce outlines the four key requirements of the Open Source AI definition, known as the OS definition, which aligns with the four freedoms of open source: the rights to study, use, modify, and share the system.
• Joyce explains the significance of open weights, open code, and comprehensive data information is explained by Joyce, highlighting their role in enabling the replication and creation of similar AI systems.
• Joyce emphasizes that the goal is to establish principles that foster broad inclusion within the open source AI community, aiming to create an environment where diverse participants are welcomed and no one is excluded.

Insights into co-design methods used in OSI projects

• Joyce discusses the co-design methodology and process employed by OSI to develop the Open Source AI definition.
• Joyce defines co-design as a collaborative process in which those affected by a problem work together to create solutions, ensuring that knowledge and decision-making power are shared among all participants involved.
• The development of the Open Source AI definition involved a multi-phase approach. This included a series of workshops, both in-person and virtual, as well as asynchronous work where participants contributed remotely to shape and finalize the definition.
• Joyce explains that the co-design process included utilizing the model openness framework from the Linux Foundation to guide the evaluation of AI systems. The involvement of volunteers played a crucial role in validating and testing these systems.

OSI’s goal in defining open source for AI

• Joyce explains that the OSI’s goal is to understand the significance of the Open Source definition as it applies to software and extend this understanding to the AI domain, ensuring that the principles of openness and transparency are maintained.
• The absence of a widely accepted definition in the open source AI space is a major challenge. This lack of clarity has led to inconsistencies and confusion in how open source AI is perceived and applied.
• Joyce outlines the scope of open source in AI by emphasizing three critical components: open weights, open code, and transparent data. Joyce addresses the challenges associated with data openness, including the difficulties in making data sets public.

Industry involvement and endorsements for the Open Source AI project

• Joyce talks about the involvement of the broader industry in the Open Source AI definition project. The most engaged contributors to the project are volunteers who have willingly made their names and affiliations public to ensure transparency in the process.
• Major corporations, including Meta, Google, and Amazon, are active participants in the project. Joyce notes that there is a notable representation of workgroup members from Africa, emphasizing the global reach and diverse participation in the initiative.
• OSI is currently seeking organizational endorsements from technology companies to formalize support and recognition for the Open Source AI definition, aiming to broaden the project’s impact and acceptance within the industry.

Current progress and urgency of defining open source AI

• Joyce discusses the current progress of the project reporting that the Open Source AI definition is nearing finalization, with version 0.0.9 and a release candidate 1 (RC1) expected to be available soon, indicating significant advancement towards completion.
• Joyce emphasizes the urgency of establishing a clear definition for open source AI, given the upcoming implementation of regulations such as the EU AI Act and California’s SB 1047, which will impact the regulatory landscape.
• Joyce underscores the importance of creating a definition that embodies global consultation and adheres to the core values of open source. This approach ensures that the definition is inclusive and reflects a wide range of perspectives and practices.

Public involvement opportunities in shaping the AI definition

• Joyce highlights the opportunities for public engagement in the project, confirming that public feedback is still being accepted for version 0.0.9, and encouraging individuals to submit comments through the OSI website and forum.
• The availability of bi-weekly virtual town hall meetings provides a platform for individuals to ask questions, share feedback, and engage directly with the project team.
• Joyce invites ongoing public involvement, emphasizing that continued participation is crucial for refining and finalizing the definition to ensure it is comprehensive and effective.

Long-term goals and maintenance plans for the AI definition

• The long-term process and objectives for the project include establishing a stable and reliable definition of open source AI that can be utilized in policy-making by October. This milestone aims to provide clarity and consistency in the field.
• Joyce highlights the necessity of implementing a maintenance plan noting that, similar to other open source projects, this plan will ensure the definition stays relevant and accurate over time.
• Joyce recognizes the swift progression of technology and underscores the need for regular updates to the definition to remain aligned with ongoing advancements and shifts in the AI sector.

Guest: Mer Joyce (LinkedIn)
Organization: Open Source Initiative (OSI)  (Twitter)
Show: Let’s Talk

The summary was written by Emily Nicholls.