Thank you to the folks at Sustain for providing the hosting account for CHAOSSCast!

CHAOSScast – Episode 82

In this episode of CHAOSScast, host Dawn Foster brings together Matt Germonprez, Brian Proffitt, and Ashley Wolf to discuss the implications of Artificial Intelligence (AI) on Open Source Program Offices (OSPOs), including policy considerations, the potential for AI-driven contributions to create workload for maintainers, and the quality of contributions. They also touch on the use of AI internally within companies versus contributing back to the open source community, the importance of distinguishing between human and AI contributions, and the potential benefits and challenges AI introduces to open source project health and community metrics. The conversation strikes a balance between optimism for AI’s benefits and caution for its governance, leaving us to ponder the future of open source in an AI-integrated world. Press download to hear more!

[00:03:20] The discussion begins on the role of OSPOs in AI policy making, and Ashley emphasizes the importance of OSPOs in providing guidance on generative AI tools usage and contributions within their organizations.

[00:05:17] Brian observes a conservative reflex towards AI in OSPOs, noting issues around copyright, trust, and the status of AI as not truly open source.

[00:07:10] Matt inquires about aligning different policies from various organizations, like GitHub and Red Hat, with those from the Linux Foundation and Apache Software Foundation regarding generative AI. Brian speaks about Red Hat’s approach to first figure out their policies before seeking alignment with others.

[00:06:45] Ashley appreciates the publicly available AI policies from the Apache and Linux Foundations, noting that GitHub’s policies have been informed by long-term thinking and community feedback.

[00:10:34] Dawn asks about potential internal conflict for GitHub employees given different AI policies at GitHub and other organizations like CNCF and Apache.

[00:12:32] Ashley and Brian talk about what they see as the benefits of AI for OSPOs, and how AI can help scale OSPO support and act as a sounding board for new ideas.

[00:15:32] Matt proposes a scenario where generative AI might increase individual contributions to high-profile projects like Kubernetes for personal gain, potentially burdening maintainers.

[00:18:45] Dawn mentions Daniel Stenberg of cURL who has seen an influx of low-quality issues from AI models, Ashley points out the problem of “drive-by-contributions” and spam, particularly during events like Hacktoberfest, and emphasizes the role of OSPOs in education about responsible contributions, and Brian discusses potential issues with AI contributions leading to homogenization and the increased risk of widespread security vulnerabilities.

[00:22:33] Matt raises another scenario questioning if companies might use generative AI internally as an alternative to open source for smaller issues without contributing back to the community. Ashley states 92% of developers are using AI code generation tools and cautions against creating code in a vacuum, and Brian talks about Red Hat’s approach.

[00:27:18] Dawn discusses the impact of generative AI on companies that are primarily consumers of open source, rarely contributing back, questioning if they might start using AI to make changes instead of contributing. Brian suggests there might be a mixed impact and Ashley optimistically hopes the time saved using AI tools will be redirected to contribute back to open source.

[00:29:49] Brian discusses the state of open source AI, highlighting the lack of a formal definition and ongoing efforts by the OSI and other groups to establish one, and recommends a fascinating article he read from Knowing Machines. Ashley emphasizes the importance of not misusing the term open source for AI until a formal definition is established.

[00:32:42] Matt inquires how metrics can aid in adapting to AI trends in open source, like detecting AI-generated contributions. Brian talks about using signals like time zones to differentiate between corporate contributors and hobbyists, and the potential for tagging contributions from AI for clarity.

[00:35:13] Ashley considers the human aspect of maintainers dealing with an influx of AI-generated contributions and what metrics could indicate a need for additional support, and she mentions the concept of the “Nebraska effect.”

Value Adds (Picks) of the week:

[00:36:59] Dawn’s pick is seeing friends over the 4 day UK Easter holiday, playing board games, eating, and hanging out.
[00:37:21] Brian’s pick is traveling back home to Indiana to see his first ever total solar eclipse and bringing his NC friends along.
[00:38:03] Matt’s pick is reconnecting with colleagues this semester and doing talks at GSU and Syracuse.
[00:38:40] Ashley’s pick is going to the local nursery and acquiring some blueberry plants.