I do not agree with Github's use of copyrighted code as training for Copilot (thelig.ht)

Article summary

The author disagrees with GitHub's use of copyrighted code as training data for their ML-powered Copilot product, citing concerns over unauthorized and unlicensed use of copyrighted work. The author has abandoned GitHub and now hosts their public programming work elsewhere. The issue revolves around the use of copyrighted source code without informing customers of the original license. This raises concerns about the rights of copyright holders and the potential for Copilot to facilitate unauthorized use of copyrighted work.

Main themes

• Copyright and AI
• Code Generation and Licensing
• Open-Source Software and Trust
• AI-Generated Code Quality and Security
• Opting Out and Control Over Copyrighted Work
• Consequences of AI-Generated Code

What commenters say

• The use of copyrighted code to train AI models may be legally permissible, but using the generated code without permission is a different issue.
• Copyright law may not prevent learning from copyrighted data, but it does prevent copying and distributing the data without permission.
• The use of AI-generated code could lead to the spread of low-quality or insecure code, and may be vulnerable to trolling or intentional corruption.
• Some argue that opting out of AI training data is impossible if code is publicly available, while others believe that copyright holders should have control over how their work is used.
• The issue is not just about copyright, but also about the potential for AI-generated code to launder copyrighted work and undermine the original authors' rights.
• There are concerns about the potential consequences of AI-generated code, including the spread of misinformation and the erosion of trust in open-source software.
• Some argue that the benefits of open-source software outweigh the risks, while others believe that the risks of AI-generated code outweigh any potential benefits.