Mopping Up the Slop: How bad is it, and what can we do?
MCLD 2002 | Sun 09 Aug 4:30 p.m.–5:15 p.m.
Presented by
-
Dr. Kaylea Champion
@kaylea@social.coop
https://kayleachampion.com
Dr. Kaylea Champion studies how people cooperate to build groovy public goods like GNU/Linux and Wikipedia, including what gets maintained and the risks we face. She is an Assistant Professor in Computing and Software Systems at the University of Washington Bothell, where she teaches software engineering with an emphasis on cybersecurity, evidence-based practices, ethics, and free/libre software. Dr. Champion is eager to build connections with communities, and has previously given talks at FOSSY, Quantum Wednesday, Norwescon, WorldCon, PyConUS, the OpenJS Summit, Wikimania, MozFest, Women in Data Science, PuPPY, DebConf, and SeaGL. A Linux user since 1994, she enjoys tromping through the woods, smashing goblins, conjuring data, and cooking for a crowd.
Dr. Kaylea Champion
@kaylea@social.coop
https://kayleachampion.com
Dr. Kaylea Champion studies how people cooperate to build groovy public goods like GNU/Linux and Wikipedia, including what gets maintained and the risks we face. She is an Assistant Professor in Computing and Software Systems at the University of Washington Bothell, where she teaches software engineering with an emphasis on cybersecurity, evidence-based practices, ethics, and free/libre software. Dr. Champion is eager to build connections with communities, and has previously given talks at FOSSY, Quantum Wednesday, Norwescon, WorldCon, PyConUS, the OpenJS Summit, Wikimania, MozFest, Women in Data Science, PuPPY, DebConf, and SeaGL. A Linux user since 1994, she enjoys tromping through the woods, smashing goblins, conjuring data, and cooking for a crowd.
Abstract
The commercialization of generative AI has introduced new challenges to the ways we organize and run our communities, including a rise in automated contributions, synthetic engagement, and automated attacks. AI slop -- especially "plausible" slop, that is, contributions and bug reports that seem sincere and well-founded but are ultimately fabricated and misleading -- are a substantial source of frustration for maintainers. How widespread is AI slop in open source? How can we detect it and respond? In this talk, I offer an update on my project investigating the problem, and the multiple follow-up efforts and new projects this problem is inspiring.
The commercialization of generative AI has introduced new challenges to the ways we organize and run our communities, including a rise in automated contributions, synthetic engagement, and automated attacks. AI slop -- especially "plausible" slop, that is, contributions and bug reports that seem sincere and well-founded but are ultimately fabricated and misleading -- are a substantial source of frustration for maintainers. How widespread is AI slop in open source? How can we detect it and respond? In this talk, I offer an update on my project investigating the problem, and the multiple follow-up efforts and new projects this problem is inspiring.