2 years ago

Category: Microsoft, github, artificial-intelligence

Last week, GitHub launched GitHub Copilot, a feature the company refers to as “your AI pair programmer,” and much of the internet immediately erupted in righteous indignation over the new feature, while the rest of it went “whoah, this is really cool.” The gist of the indignation came down to the fact that GitHub Copilot is built using OpenAI Codex, which uses as its training set pretty much all publicly available code on GitHub — including code with viral copyleft licensing, such as GPL. As we discussed last week, however, manypreliminary examinations by those in the legal field seemed to align with that of GitHub, which sees the entire thing as a case of “fair use.”

Meanwhile, yet another piece came out this week, again not only positing that GitHub Copilot is not infringing your copyright but also arguing that the move to restrict it from using copyleft licenses may actually be ill-advised.

In the blog post, Julia Reda, a researcher who focused on copyright reform during her time as a Member of the European Parliament, takes up similar issues to those we saw last week, but also argues that the reaction to GitHub Copilot is antithetical to the end goals of copyleft licenses.