A week ago I wrote about AI being used to help modernize Ubuntu’s Error Tracker. Microsoft GitHub Copilot was tasked to help adapt its Cassandra database usage to modern standards. It’s worked in some areas but even for a rather straight forward task, some of the generated functions ended up being “plain wrong” according to the developer involved.
Some Phoronix readers had cheered over last week’s article in arguing that modernizing codebases can be a great use for AI to help update to modern coding standards and removing use of deprecated code. But it’s still not perfect.
Canonical engineer “Skia” provided a status update on this AI-modernization of the Ubuntu Error Tracker. In Thursday’s Ubuntu Foundations Team weekly notes Skia commented:
“Now on the process of reviewing and testing what Copilot produced. It’s not too bad, but also not working out of the box for various reasons, among which it doesn’t have access to a real database, and I didn’t provide the schema in my prompt. Some functions [were] even plain wrong, but the proportion of those is fairly low. See my latest changes in the PR for details.”
So some of the AI generated functions from GitHub Copilot were plain wrong but at least for the most part this AI-driven effort seems to have saved some development time. Those interested in the generated AI slop and corrections and the like for some weekend entertainment can dive in via this GitHub pull request.
