Jan 31, 2020

GitHub gets serious about deep learning

GitHub deep learning

GitHub announced a new feature to better match open source contributors with issues that are likely to fit their interests and skill level. GitHub now recommends "good first issues" in open source projects for developers who are hoping to get more involved but are not sure what tasks to complete first.

This is the first deep-learning-enabled product to launch on the Github platform. That marks a massive step forward for GitHub into unchartered territory.

Take one: GitHub first announced an early version of its "good first issues" feature in early 2019, but it placed the burden of triaging and labeling issues—with tags like "good first issue" and "beginner friendly"—on project maintainers. As a result, only 40% of repositories had issues that could be recommended to developers.

Take two: GitHub’s updated recommendation system now works automatically and learns from existing issues. GitHub can recognize patterns in issues that were closed by developers who had never contributed to a repository before. That helps it identify similar issues that could be worthwhile for other beginners.

Why it's a big deal: GitHub’s new feature combines the power of GitHub’s troves of developer data—more than 40 million registered users—with its growing automation capabilities. Add machine learning into the mix, and GitHub is uniquely positioned to make existing developer workflows smarter and create powerful new ones that were not possible before.

GitHub already plans to create a personal set of recommended next issues for developers who have already contributed to a project so that they know what to work on next.

Now that GitHub has taken the first step in expanding its machine learning capabilities, it can begin to replace even more manual workflows.

Want to get more of these in your inbox?

Subscribe for weekly updates from the Software team.